Merge remote-tracking branch 'origin/main' into litellm_lit_8281_langfuse_otel_trace_fields

.circleci/config.yml

@@ -12,6 +12,9 @@ parameters:
   migration_source_sha:
     type: string
     default: ""
+  routing_parity_base:
+    type: string
+    default: ""
 orbs:
   codecov: codecov/codecov@4.0.1
   node: circleci/node@5.1.0 # Add this line to declare the node orb
@@ -176,6 +179,9 @@ commands:
       image:
         type: string
         default: postgres:14@sha256:6a70deda415ec296f977890e11aba04a0db9f632a362e3fce45e845e3db74f26
+      server_args:
+        type: string
+        default: ""
     steps:
       - run:
           name: Start PostgreSQL
@@ -186,7 +192,7 @@ commands:
               -e POSTGRES_PASSWORD=postgres \
               -e POSTGRES_DB=<< parameters.db_name >> \
               -p 5432:5432 \
-              << parameters.image >>
+              << parameters.image >> << parameters.server_args >>
       - wait_for_service:
           url: tcp://localhost:5432
           timeout: "60"
@@ -3108,6 +3114,10 @@ jobs:
     parameters:
       suite:
         type: string
+      mode:
+        type: enum
+        enum: [standard, replica]
+        default: standard
     machine:
       image: ubuntu-2204:2024.04.1
     resource_class: large
@@ -3142,18 +3152,19 @@ jobs:
                 command: cd ui/litellm-dashboard && NEXT_TELEMETRY_DISABLED=1 npm run build
       - start_postgres:
           image: postgres:16@sha256:e17e86066e5ef83e0952a9347f5c792b7ece00972e2aa787a6986f471b3dd3d5
+          server_args: "-c shared_preload_libraries=pg_stat_statements -c pg_stat_statements.track=all -c pg_stat_statements.max=20000"
       - start_redis
       - run:
           name: Run owned integration contracts
-          command: bash .circleci/scripts/run_integration.sh << parameters.suite >>
+          command: bash .circleci/scripts/run_integration.sh << parameters.suite >> << parameters.mode >>
           no_output_timeout: 15m
       - run:
           name: Stop owned database and Redis
           when: always
           command: |
-            mkdir -p test-results/integration-<< parameters.suite >>
-            docker logs postgres-db > test-results/integration-<< parameters.suite >>/postgres.log 2>&1 || true
-            docker logs redis-cache > test-results/integration-<< parameters.suite >>/redis.log 2>&1 || true
+            mkdir -p test-results/services-<< parameters.suite >>-<< parameters.mode >>
+            docker logs postgres-db > test-results/services-<< parameters.suite >>-<< parameters.mode >>/postgres.log 2>&1 || true
+            docker logs redis-cache > test-results/services-<< parameters.suite >>-<< parameters.mode >>/redis.log 2>&1 || true
             docker rm -f postgres-db redis-cache
             test -z "$(docker ps -aq --filter name=postgres-db --filter name=redis-cache)"
       - store_test_results:
@@ -3161,6 +3172,76 @@ jobs:
       - store_artifacts:
           path: test-results
 
+  routing_parity:
+    parameters:
+      suite:
+        type: string
+    machine:
+      image: ubuntu-2204:2024.04.1
+    resource_class: large
+    working_directory: ~/project
+    steps:
+      - setup_litellm_test_deps
+      - run:
+          name: Check out base product code
+          environment:
+            ROUTING_PARITY_BASE: << pipeline.parameters.routing_parity_base >>
+          command: |
+            [[ "$ROUTING_PARITY_BASE" =~ ^[0-9a-f]{40}$ ]] || exit 1
+            git fetch --depth 1 origin "$ROUTING_PARITY_BASE"
+            git rm -r -f --quiet litellm enterprise litellm-proxy-extras
+            git checkout "$ROUTING_PARITY_BASE" -- litellm enterprise litellm-proxy-extras
+            git reset --quiet
+            test -f litellm/rust_bridge/_native.abi3.so
+      - start_postgres:
+          image: postgres:16@sha256:e17e86066e5ef83e0952a9347f5c792b7ece00972e2aa787a6986f471b3dd3d5
+          server_args: "-c shared_preload_libraries=pg_stat_statements -c pg_stat_statements.track=all -c pg_stat_statements.max=20000"
+      - start_redis
+      - run:
+          name: Run base side
+          command: bash .circleci/scripts/run_integration.sh << parameters.suite >> parity base
+          no_output_timeout: 15m
+      - run:
+          name: Stop base database and Redis
+          when: always
+          command: |
+            mkdir -p test-results/services-<< parameters.suite >>-parity-base
+            docker logs postgres-db > test-results/services-<< parameters.suite >>-parity-base/postgres.log 2>&1 || true
+            docker logs redis-cache > test-results/services-<< parameters.suite >>-parity-base/redis.log 2>&1 || true
+            docker rm -f postgres-db redis-cache
+            test -z "$(docker ps -aq --filter name=postgres-db --filter name=redis-cache)"
+      - run:
+          name: Check out head product code
+          command: |
+            git rm -r -f --quiet litellm enterprise litellm-proxy-extras
+            git checkout "$CIRCLE_SHA1" -- litellm enterprise litellm-proxy-extras
+            git reset --quiet
+            test -f litellm/rust_bridge/_native.abi3.so
+      - start_postgres:
+          image: postgres:16@sha256:e17e86066e5ef83e0952a9347f5c792b7ece00972e2aa787a6986f471b3dd3d5
+          server_args: "-c shared_preload_libraries=pg_stat_statements -c pg_stat_statements.track=all -c pg_stat_statements.max=20000"
+      - start_redis
+      - run:
+          name: Run head side
+          command: bash .circleci/scripts/run_integration.sh << parameters.suite >> parity head
+          no_output_timeout: 15m
+      - run:
+          name: Stop head database and Redis
+          when: always
+          command: |
+            mkdir -p test-results/services-<< parameters.suite >>-parity-head
+            docker logs postgres-db > test-results/services-<< parameters.suite >>-parity-head/postgres.log 2>&1 || true
+            docker logs redis-cache > test-results/services-<< parameters.suite >>-parity-head/redis.log 2>&1 || true
+            docker rm -f postgres-db redis-cache
+            test -z "$(docker ps -aq --filter name=postgres-db --filter name=redis-cache)"
+      - run:
+          name: Compare routing parity
+          command: PYTHONPATH="$PWD/tests" .venv/bin/python -m integration._support.routing check test-results/parity-<< parameters.suite >>/base test-results/parity-<< parameters.suite >>/head
+      - store_test_results:
+          path: test-results
+      - store_artifacts:
+          path: test-results
+
   unit:
     machine:
       image: ubuntu-2204:2024.04.1
@@ -3222,10 +3303,24 @@ workflows:
           cron: "17 0,6,12,18 * * *"
           filters:
             branches:
-              only: litellm_internal_staging
+              only: main
     jobs: *migration_jobs
+  routing_parity:
+    when:
+      not:
+        equal: ["", << pipeline.parameters.routing_parity_base >>]
+    jobs:
+      - routing_parity:
+          name: routing-parity-<< matrix.suite >>
+          matrix:
+            parameters:
+              suite: [management, accounting, database, providers, extensions, cost, mcp]
   integration:
-    unless: << pipeline.parameters.run_migration_tests >>
+    unless:
+      or:
+        - << pipeline.parameters.run_migration_tests >>
+        - not:
+            equal: ["", << pipeline.parameters.routing_parity_base >>]
     jobs:
       - integration_contracts:
           name: integration-<< matrix.suite >>
@@ -3237,8 +3332,23 @@ workflows:
               only:
                 - main
                 - /litellm_.*/
+      - integration_contracts:
+          name: integration-<< matrix.suite >>-replica
+          matrix:
+            parameters:
+              suite: [management, database]
+              mode: [replica]
+          filters:
+            branches:
+              only:
+                - main
+                - /litellm_.*/
   build_and_test:
-    unless: << pipeline.parameters.run_migration_tests >>
+    unless:
+      or:
+        - << pipeline.parameters.run_migration_tests >>
+        - not:
+            equal: ["", << pipeline.parameters.routing_parity_base >>]
     jobs:
       - using_litellm_on_windows:
           filters: &main_branches

.circleci/scripts/classify_changes.sh

@@ -14,12 +14,12 @@ while IFS= read -r file || [ -n "$file" ]; do
   [ -n "$file" ] || continue
   case "$file" in
     *.md | *.mdx) : ;;
-    pyproject.toml | */pyproject.toml | uv.lock | uv.toml | .python-version | rust-toolchain.toml | litellm-rust/* | litellm/__init__.py | litellm/proxy/proxy_server.py | litellm/*mcp* | tests/*mcp* | litellm/integrations/arize/* | tests/base_sdk_tests/* | scripts/check_mcp_sdk_install.py | .github/workflows/test-mcp-dependency-resolution.yml | .github/actions/detect-changes/* | .github/actions/setup-uv-with-retries/* | .github/actions/cache-cargo-build/* | .github/scripts/detect_changes.sh | .github/scripts/uv_sync_with_retries.sh | .circleci/scripts/classify_changes.sh | tests/test_litellm/test_circleci_path_filter.py | tests/test_litellm/test_detect_changes.py)
+    pyproject.toml | */pyproject.toml | uv.lock | uv.toml | .python-version | rust-toolchain.toml | litellm-rust/* | litellm/__init__.py | litellm/proxy/proxy_server.py | litellm/*mcp* | tests/*mcp* | litellm/integrations/arize/* | tests/base_sdk_tests/* | scripts/check_mcp_sdk_install.py | .github/workflows/test-mcp-dependency-resolution.yml | .github/actions/detect-changes/* | .github/actions/setup-uv-with-retries/* | .github/actions/cache-cargo-build/* | .github/scripts/detect_changes.sh | .github/scripts/uv_sync_with_retries.sh | .circleci/scripts/classify_changes.sh | tests/unit/test_circleci_path_filter.py | tests/unit/test_detect_changes.py)
       has_mcp_dependencies=true ;;
   esac
   case "$file" in
     tests/e2e/*/*.py) : ;;
-    tests/e2e/*.py | tests/code_coverage_tests/test_provider_cache.py | tests/code_coverage_tests/test_provider_replay_harness.py | tests/test_litellm/test_circleci_path_filter.py | .circleci/* | pyproject.toml | uv.lock)
+    tests/e2e/*.py | tests/code_coverage_tests/test_provider_cache.py | tests/code_coverage_tests/test_provider_replay_harness.py | tests/unit/test_circleci_path_filter.py | .circleci/* | pyproject.toml | uv.lock)
       has_provider_harness=true ;;
   esac
   case "$file" in
@@ -31,7 +31,7 @@ while IFS= read -r file || [ -n "$file" ]; do
   case "$file" in
     model_prices_and_context_window.json | litellm/model_prices_and_context_window_backup.json | model_prices_and_context_window.schema.json)
       has_cost_map=true ;;
-    tests/test_litellm/* | tests/proxy_unit_tests/*) : ;;
+    tests/test_litellm/* | tests/proxy_unit_tests/* | tests/unit/proxy/*) : ;;
     *) outside_cost_map_set=true ;;
   esac
 done

.circleci/scripts/path_filter.sh

@@ -11,7 +11,7 @@ run_full() {
 
 [ -n "${CIRCLE_PULL_REQUEST:-}" ] || run_full "not a pull request"
 
-candidate_bases="main litellm_internal_staging litellm_oss_staging"
+candidate_bases="${PATH_FILTER_BASE_BRANCH:-main}"
 merge_base=""
 for base in $candidate_bases; do
   git fetch --quiet origin "$base" 2>/dev/null || continue

.circleci/scripts/prepare_replica_roles.py

@@ -0,0 +1,52 @@
+from __future__ import annotations
+
+import os
+from typing import Final
+from urllib.parse import urlsplit, urlunsplit
+
+import psycopg
+
+DATABASE_URL: Final = os.environ["DATABASE_URL"]
+
+
+def postgres_url() -> str:
+    parsed: Final = urlsplit(DATABASE_URL)
+    return urlunsplit(parsed._replace(path="/postgres"))
+
+
+def main() -> None:
+    with psycopg.connect(postgres_url(), autocommit=True) as admin:
+        admin.execute("CREATE EXTENSION IF NOT EXISTS pg_stat_statements")
+        admin.execute("CREATE ROLE litellm_writer LOGIN PASSWORD 'litellm-writer' NOSUPERUSER")
+        admin.execute("CREATE ROLE litellm_reader LOGIN PASSWORD 'litellm-reader' NOSUPERUSER NOINHERIT")
+        admin.execute("ALTER ROLE litellm_reader SET default_transaction_read_only = on")
+        admin.execute("ALTER DATABASE circle_test OWNER TO litellm_writer")
+        admin.execute("GRANT CONNECT ON DATABASE circle_test TO litellm_reader")
+    with psycopg.connect(DATABASE_URL, autocommit=True) as admin:
+        admin.execute("GRANT USAGE ON SCHEMA public TO litellm_reader")
+        admin.execute(
+            "ALTER DEFAULT PRIVILEGES FOR ROLE litellm_writer IN SCHEMA public GRANT SELECT ON TABLES TO litellm_reader"
+        )
+        admin.execute("GRANT SELECT ON ALL TABLES IN SCHEMA public TO litellm_reader")
+
+    parsed: Final = urlsplit(DATABASE_URL)
+    reader_url: Final = urlunsplit(
+        parsed._replace(netloc=f"litellm_reader:litellm-reader@{parsed.hostname}:{parsed.port}")
+    )
+    writer_url: Final = urlunsplit(
+        parsed._replace(netloc=f"litellm_writer:litellm-writer@{parsed.hostname}:{parsed.port}")
+    )
+    with psycopg.connect(reader_url, autocommit=True) as reader:
+        assert reader.execute("SHOW transaction_read_only").fetchone() == ("on",)
+        try:
+            reader.execute("CREATE TABLE integration_readonly_probe (id int)")
+        except psycopg.errors.ReadOnlySqlTransaction:
+            pass
+        else:
+            raise AssertionError("litellm_reader executed a write statement")
+    with psycopg.connect(writer_url, autocommit=True) as writer:
+        assert writer.execute("SELECT current_user").fetchone() == ("litellm_writer",)
+
+
+if __name__ == "__main__":
+    main()

.circleci/scripts/run_integration.sh

@@ -7,7 +7,15 @@ if [ "${GITHUB_ACTIONS:-}" = true ]; then
 fi
 
 suite="${1:?integration suite required}"
-results="test-results/integration-${suite}"
+mode="${2:-standard}"
+side="${3:-}"
+if [ "$mode" = replica ]; then
+  results="test-results/integration-${suite}-replica"
+elif [ "$mode" = parity ]; then
+  results="test-results/parity-${suite}/${side:?parity side required}"
+else
+  results="test-results/integration-${suite}"
+fi
 mkdir -p "$results"
 integration_identity="$(.venv/bin/python -c 'import uuid; print(uuid.uuid4().hex)')"
 upstream_pid=""
@@ -80,6 +88,18 @@ export INTEGRATION_ORDER_SEED="$INTEGRATION_SEED"
 
 uv run --no-sync prisma generate --schema litellm/proxy/schema.prisma > "$results/prisma-generate.log" 2>&1
 
+export INTEGRATION_PROXY_DATABASE_URL=""
+export INTEGRATION_PROXY_READ_REPLICA_URL=""
+export INTEGRATION_ROUTING=""
+if [ "$mode" = replica ] || [ "$mode" = parity ]; then
+  .venv/bin/python .circleci/scripts/prepare_replica_roles.py > "$results/prepare-replica-roles.log" 2>&1
+  export INTEGRATION_PROXY_DATABASE_URL="postgresql://litellm_writer:litellm-writer@127.0.0.1:5432/circle_test"
+  export INTEGRATION_PROXY_READ_REPLICA_URL="postgresql://litellm_reader:litellm-reader@127.0.0.1:5432/circle_test"
+fi
+if [ "$mode" = parity ]; then
+  export INTEGRATION_ROUTING=capture
+fi
+
 sudo iptables -N integration_only
 guard_created=true
 sudo iptables -A integration_only -o lo -j ACCEPT
@@ -137,8 +157,12 @@ start_proxy() {
   else
     cost_map_env=("LITELLM_LOCAL_MODEL_COST_MAP=True")
   fi
+  local -a database_env=("DATABASE_URL=${INTEGRATION_PROXY_DATABASE_URL:-$DATABASE_URL}")
+  if [ -n "$INTEGRATION_PROXY_READ_REPLICA_URL" ]; then
+    database_env+=("DATABASE_URL_READ_REPLICA=$INTEGRATION_PROXY_READ_REPLICA_URL")
+  fi
   setsid env -i PATH="$PATH" HOME="$HOME" PYTHONPATH="$PYTHONPATH" INTEGRATION_RUN_ID="$integration_identity" \
-    DATABASE_URL="$DATABASE_URL" REDIS_HOST="$REDIS_HOST" REDIS_PORT="$REDIS_PORT" \
+    "${database_env[@]}" REDIS_HOST="$REDIS_HOST" REDIS_PORT="$REDIS_PORT" \
     INTEGRATION_UPSTREAM_URL="$INTEGRATION_UPSTREAM_URL" \
     LITELLM_MASTER_KEY="$LITELLM_MASTER_KEY" LITELLM_SALT_KEY="$LITELLM_SALT_KEY" LITELLM_UI_PATH="$LITELLM_UI_PATH" PROXY_BASE_URL="http://127.0.0.1:$port" \
     LITELLM_MODE=PRODUCTION STORE_MODEL_IN_DB=True "${cost_map_env[@]}" \
@@ -195,6 +219,9 @@ env -i PATH="$PATH" HOME="$HOME" PYTHONPATH="$PYTHONPATH" \
   INTEGRATION_SEED="$INTEGRATION_SEED" \
   INTEGRATION_ORDER_SEED="$INTEGRATION_ORDER_SEED" \
   LITELLM_LOCAL_MODEL_COST_MAP=True AWS_EC2_METADATA_DISABLED=true DO_NOT_TRACK=1 \
+  INTEGRATION_PROXY_DATABASE_URL="$INTEGRATION_PROXY_DATABASE_URL" \
+  INTEGRATION_PROXY_READ_REPLICA_URL="$INTEGRATION_PROXY_READ_REPLICA_URL" \
+  INTEGRATION_ROUTING="$INTEGRATION_ROUTING" \
   .venv/bin/python tests/integration/run.py "$suite" --results "$results"
 
 if [ "${INTEGRATION_COVERAGE:-0}" = 1 ]; then

.circleci/scripts/unit_selection.sh

@@ -0,0 +1,170 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+flag="${1:?usage: unit_selection.sh <codecov flag>}"
+
+legacy_flags=(
+  caching-local
+  enterprise-package
+  enterprise-routing
+  integrations
+  llm-other-providers
+  llm-vertex-ai
+  mcp-integration
+  misc
+  proxy-db-auth-checks
+  proxy-db-budgets
+  proxy-db-custom-logging
+  proxy-db-db-and-spend
+  proxy-db-endpoints-and-responses
+  proxy-db-guardrails-hooks
+  proxy-db-jwt-and-keys
+  proxy-db-key-generation
+  proxy-db-logging-misc
+  proxy-db-proxy-runtime
+  proxy-db-proxy-server-core
+  proxy-db-proxy-utils
+  proxy-extras
+  proxy-infra
+  responses-caching-types
+)
+
+legacy_paths() {
+  case "$1" in
+    caching-local) echo tests/unit/caching ;;
+    enterprise-package)
+      echo tests/unit/enterprise/integrations
+      echo tests/unit/enterprise/proxy/auth
+      echo tests/unit/enterprise/proxy/guardrails
+      echo tests/unit/enterprise/proxy/hooks
+      echo tests/unit/enterprise/proxy/management_endpoints
+      echo tests/unit/enterprise/proxy/test_audit_logging_endpoints.py
+      echo tests/unit/enterprise/enterprise_callbacks/test_prometheus_logging_callbacks.py ;;
+    enterprise-routing)
+      echo tests/unit/google_genai
+      echo tests/unit/enterprise/enterprise_callbacks/send_emails
+      echo tests/unit/enterprise/proxy/test_afile_retrieve_returns_unified_id.py
+      echo tests/unit/enterprise/proxy/test_batch_retrieve_input_file_id.py
+      echo tests/unit/enterprise/proxy/test_batch_retrieve_registers_missing_output_file_id.py
+      echo tests/unit/enterprise/proxy/test_batch_retrieve_returns_unified_input_file_id.py
+      echo tests/unit/enterprise/proxy/test_batch_update_db_managed_output_file_id.py
+      echo tests/unit/enterprise/proxy/test_deleted_file_returns_403_not_404.py
+      echo tests/unit/enterprise/proxy/test_enterprise_routes.py
+      echo tests/unit/enterprise/proxy/test_file_deletion_blocking.py
+      echo tests/unit/enterprise/proxy/test_managed_files_access_check.py
+      echo tests/unit/enterprise/proxy/test_managed_files_hook.py ;;
+    integrations) echo tests/unit/integrations ;;
+    llm-other-providers) find tests/unit/llms -name 'test_*.py' -not -path 'tests/unit/llms/vertex_ai/*' ;;
+    llm-vertex-ai) echo tests/unit/llms/vertex_ai ;;
+    mcp-integration)
+      echo tests/unit/experimental_mcp_client
+      echo tests/unit/proxy/_experimental/mcp_server
+      echo tests/unit/responses/mcp
+      echo tests/mcp_tests/test_proxy_mcp_e2e.py ;;
+    misc)
+      find tests/unit -maxdepth 1 -name 'test_*.py'
+      echo tests/unit/test_router
+      echo tests/unit/a2a_protocol
+      echo tests/unit/batches
+      echo tests/unit/chat_completions
+      echo tests/unit/completion_extras
+      echo tests/unit/containers
+      echo tests/unit/embeddings
+      echo tests/unit/endpoints
+      echo tests/unit/files
+      echo tests/unit/images
+      echo tests/unit/interactions
+      echo tests/unit/messages
+      echo tests/unit/rag
+      echo tests/unit/rerank_api
+      echo tests/unit/secret_managers
+      echo tests/unit/vector_stores
+      echo tests/unit/videos ;;
+    proxy-db-auth-checks)
+      echo tests/unit/proxy/auth/test_auth_checks.py
+      echo tests/unit/proxy/auth/test_user_api_key_auth.py
+      echo tests/unit/proxy/test_deprecated_key_grace_period.py ;;
+    proxy-db-budgets)
+      echo tests/unit/proxy/auth/test_default_end_user_budget_simple.py
+      echo tests/unit/proxy/hooks/test_unit_test_max_model_budget_limiter.py
+      echo tests/unit/proxy/test_zero_cost_model_budget_bypass.py ;;
+    proxy-db-custom-logging)
+      echo tests/unit/proxy/test_custom_callback_input.py
+      echo tests/unit/proxy/test_custom_logger_s3_gcs.py ;;
+    proxy-db-db-and-spend)
+      echo tests/unit/proxy/common_utils/test_proxy_encrypt_decrypt.py
+      echo tests/unit/proxy/db/db_transaction_queue/test_e2e_pod_lock_manager.py
+      echo tests/unit/proxy/db/test_update_daily_tag_spend.py
+      echo tests/unit/proxy/test_db_schema_changes.py
+      echo tests/unit/proxy/test_prisma_client_backoff_retry.py
+      echo tests/unit/proxy/test_update_spend.py
+      echo tests/unit/skills/test_skills_db.py ;;
+    proxy-db-endpoints-and-responses)
+      echo tests/unit/proxy/auth/test_models_fallback_endpoint.py
+      echo tests/unit/proxy/common_utils/test_check_batch_cost.py
+      echo tests/unit/proxy/common_utils/test_check_responses_cost.py
+      echo tests/unit/proxy/common_utils/test_realtime_cache.py
+      echo tests/unit/proxy/google_endpoints/test_gemini_agents_endpoints.py
+      echo tests/unit/proxy/google_endpoints/test_google_endpoint_routing.py
+      echo tests/unit/proxy/google_endpoints/test_google_gemini_proxy_request.py
+      echo tests/unit/proxy/public_endpoints/test_blog_posts_endpoint.py
+      echo tests/unit/proxy/response_polling/test_response_polling_handler.py
+      echo tests/unit/proxy/test_custom_tokenizer_bug.py
+      echo tests/unit/proxy/test_get_favicon.py
+      echo tests/unit/proxy/test_get_image.py
+      echo tests/unit/proxy/test_prompt_test_endpoint.py
+      echo tests/unit/proxy/test_reducto_ocr_route.py
+      echo tests/unit/proxy/test_response_polling_pre_call_checks.py
+      echo tests/unit/proxy/test_ui_path_detection.py ;;
+    proxy-db-guardrails-hooks)
+      echo tests/unit/proxy/hooks/test_banned_keyword_list.py
+      echo tests/unit/proxy/test_proxy_setting_guardrails.py
+      echo tests/unit/proxy/test_unit_test_proxy_hooks.py ;;
+    proxy-db-jwt-and-keys)
+      echo tests/unit/proxy/auth/test_jwt.py
+      echo tests/unit/proxy/management_endpoints/test_jwt_key_mapping.py
+      echo tests/unit/proxy/test_proxy_custom_auth.py ;;
+    proxy-db-key-generation) echo tests/unit/proxy/management_endpoints/test_key_generate_prisma.py ;;
+    proxy-db-logging-misc)
+      echo tests/unit/proxy/management_helpers/test_audit_logs_proxy.py
+      echo tests/unit/proxy/spend_tracking/test_search_api_logging.py
+      echo tests/unit/proxy/test_proxy_reject_logging.py ;;
+    proxy-db-proxy-runtime)
+      echo tests/unit/proxy/auth/test_multipart_bypass_repro.py
+      echo tests/unit/proxy/auth/test_proxy_routes.py
+      echo tests/unit/proxy/middleware/test_request_size_limit_middleware.py
+      echo tests/unit/proxy/test_proxy_config_unit_test.py
+      echo tests/unit/proxy/test_proxy_token_counter.py
+      echo tests/unit/proxy/test_server_root_path.py ;;
+    proxy-db-proxy-server-core)
+      echo tests/unit/proxy/test_aproxy_startup.py
+      echo tests/unit/proxy/test_proxy_server.py ;;
+    proxy-db-proxy-utils) echo tests/unit/proxy/test_proxy_utils.py ;;
+    proxy-extras) echo tests/unit/litellm_proxy_extras ;;
+    proxy-infra) echo tests/unit/gateway ;;
+    responses-caching-types) echo tests/unit/types ;;
+    *) echo "unit_selection.sh: unknown flag $1" >&2; exit 1 ;;
+  esac
+}
+
+expand() {
+  while read -r path; do
+    if [ -d "$path" ]; then
+      find "$path" -name 'test_*.py'
+    elif [ -f "$path" ]; then
+      echo "$path"
+    else
+      echo "unit_selection.sh: $path does not exist" >&2
+      exit 1
+    fi
+  done
+}
+
+if [ "$flag" = unit ]; then
+  comm -23 \
+    <(find tests/unit -name 'test_*.py' | sort) \
+    <(for legacy in "${legacy_flags[@]}"; do legacy_paths "$legacy"; done | expand | sort)
+  exit 0
+fi
+
+legacy_paths "$flag" | expand | sort

.circleci/tests.yml

@@ -0,0 +1,409 @@
+version: 2.1
+
+commands:
+  wait_for_service:
+    parameters:
+      url:
+        type: string
+      timeout:
+        type: string
+        default: "60"
+    steps:
+      - run:
+          name: "Wait for << parameters.url >>"
+          command: |
+            TIMEOUT=<< parameters.timeout >>
+            URL="<< parameters.url >>"
+            ELAPSED=0
+            echo "Waiting up to ${TIMEOUT}s for ${URL} ..."
+            if echo "$URL" | grep -q '^tcp://'; then
+              HOST=$(echo "$URL" | sed 's|tcp://||' | cut -d: -f1)
+              PORT=$(echo "$URL" | sed 's|tcp://||' | cut -d: -f2)
+              while ! bash -c "echo > /dev/tcp/$HOST/$PORT" 2>/dev/null; do
+                sleep 2; ELAPSED=$((ELAPSED+2))
+                if [ "$ELAPSED" -ge "$TIMEOUT" ]; then echo "Timed out"; exit 1; fi
+              done
+            else
+              while ! curl -sf --max-time 5 "$URL" > /dev/null 2>&1; do
+                sleep 2; ELAPSED=$((ELAPSED+2))
+                if [ "$ELAPSED" -ge "$TIMEOUT" ]; then echo "Timed out"; exit 1; fi
+              done
+            fi
+            echo "Service ready after ${ELAPSED}s"
+  install_uv:
+    steps:
+      - run:
+          name: Install uv (pinned 0.10.9)
+          command: |
+            curl -LsSf -o /tmp/uv-install.sh https://astral.sh/uv/0.10.9/install.sh
+            echo "7fc46e39cb97290b57169c0c813a17970585ac519139f19006453c99b5f2f45f  /tmp/uv-install.sh" | sha256sum -c -
+            env UV_NO_MODIFY_PATH=1 sh /tmp/uv-install.sh
+            rm -f /tmp/uv-install.sh
+            echo 'export PATH="$HOME/.local/bin:$PATH"' >> "$BASH_ENV"
+            export PATH="$HOME/.local/bin:$PATH"
+  install_rust:
+    steps:
+      - run:
+          name: Install Rust (rustup 1.28.2, toolchain 1.98.0)
+          command: |
+            case "$(uname -m)" in
+              x86_64)
+                RUSTUP_TRIPLE=x86_64-unknown-linux-gnu
+                RUSTUP_SHA256=20a06e644b0d9bd2fbdbfd52d42540bdde820ea7df86e92e533c073da0cdd43c
+                ;;
+              aarch64)
+                RUSTUP_TRIPLE=aarch64-unknown-linux-gnu
+                RUSTUP_SHA256=e3853c5a252fca15252d07cb23a1bdd9377a8c6f3efa01531109281ae47f841c
+                ;;
+              *)
+                echo "install_rust: unsupported architecture $(uname -m)" >&2
+                exit 1
+                ;;
+            esac
+            curl -sSLf -o /tmp/rustup-init \
+              "https://static.rust-lang.org/rustup/archive/1.28.2/${RUSTUP_TRIPLE}/rustup-init"
+            echo "${RUSTUP_SHA256}  /tmp/rustup-init" | sha256sum -c -
+            chmod +x /tmp/rustup-init
+            /tmp/rustup-init -y --no-modify-path --profile minimal --default-toolchain 1.98.0
+            rm -f /tmp/rustup-init
+            echo 'export PATH="$HOME/.cargo/bin:$PATH"' >> "$BASH_ENV"
+            export PATH="$HOME/.cargo/bin:$PATH"
+            rustc --version
+            cargo --version
+  install_codecov_cli:
+    steps:
+      - run:
+          name: Install Codecov CLI (pinned v11.3.1)
+          when: always
+          command: |
+            curl -sSLf -o /tmp/codecov https://cli.codecov.io/v11.3.1/linux/codecov
+            curl -sSLf -o /tmp/codecov.SHA256SUM https://cli.codecov.io/v11.3.1/linux/codecov.SHA256SUM
+            [ "$(cat /tmp/codecov.SHA256SUM)" = "ca1d64196d2d34771084afe76ea657d581bf628e31d993ff8e52ea09cc88a56d  codecov" ]
+            (cd /tmp && sha256sum -c codecov.SHA256SUM)
+            chmod +x /tmp/codecov
+            mkdir -p "$HOME/.local/bin"
+            mv /tmp/codecov "$HOME/.local/bin/codecov"
+  setup_litellm_enterprise_pip:
+    steps:
+      - run:
+          name: "Install local version of litellm-enterprise"
+          command: |
+            uv run --no-sync python -c "import litellm_enterprise; print('litellm-enterprise OK:', litellm_enterprise.__file__)"
+  setup_test_deps:
+    steps:
+      - install_uv
+      - install_rust
+      - restore_cache:
+          keys:
+            - v3-integration-uv-cache-{{ checksum "uv.lock" }}
+      - run:
+          name: Install Dependencies
+          command: |
+            uv sync --frozen --all-groups --all-extras --python 3.12
+      - setup_litellm_enterprise_pip
+      - save_cache:
+          paths:
+            - ~/.cache/uv
+          key: v3-integration-uv-cache-{{ checksum "uv.lock" }}
+      - run:
+          name: Generate Prisma client
+          command: uv run --no-sync prisma generate --schema litellm/proxy/schema.prisma
+  skip_unless_relevant:
+    parameters:
+      category:
+        type: string
+        default: backend
+      base_ref:
+        type: string
+        default: ""
+      pull_request_url:
+        type: string
+        default: ""
+    steps:
+      - run:
+          name: "Skip job when no << parameters.category >>-relevant files changed"
+          command: |
+            export CIRCLE_PULL_REQUEST="${CIRCLE_PULL_REQUEST:-<< parameters.pull_request_url >>}"
+            export PATH_FILTER_BASE_BRANCH="<< parameters.base_ref >>"
+            [ -n "$PATH_FILTER_BASE_BRANCH" ] || unset PATH_FILTER_BASE_BRANCH
+            bash .circleci/scripts/path_filter.sh << parameters.category >>
+  start_postgres:
+    parameters:
+      db_name:
+        type: string
+        default: circle_test
+      image:
+        type: string
+        default: postgres:14@sha256:6a70deda415ec296f977890e11aba04a0db9f632a362e3fce45e845e3db74f26
+    steps:
+      - run:
+          name: Start PostgreSQL
+          command: |
+            docker run -d \
+              --name postgres-db \
+              -e POSTGRES_USER=postgres \
+              -e POSTGRES_PASSWORD=postgres \
+              -e POSTGRES_DB=<< parameters.db_name >> \
+              -p 5432:5432 \
+              << parameters.image >>
+      - wait_for_service:
+          url: tcp://localhost:5432
+          timeout: "60"
+  start_redis:
+    steps:
+      - run:
+          name: Start Redis
+          command: |
+            docker run -d \
+              --name redis-cache \
+              -p 6379:6379 \
+              redis:7-alpine@sha256:7aec734b2bb298a1d769fd8729f13b8514a41bf90fcdd1f38ec52267fbaa8ee6
+      - wait_for_service:
+          url: tcp://localhost:6379
+          timeout: "60"
+
+jobs:
+  unit:
+    parameters:
+      flag:
+        type: string
+        default: unit
+      shards:
+        type: integer
+        default: 6
+      workers:
+        type: integer
+        default: 4
+      dist:
+        type: string
+        default: loadscope
+      base_ref:
+        type: string
+        default: ""
+      pull_request_url:
+        type: string
+        default: ""
+      legacy_mcp_peer:
+        type: boolean
+        default: false
+      reruns:
+        type: integer
+        default: 0
+    machine:
+      image: ubuntu-2204:2024.04.1
+    resource_class: large
+    working_directory: ~/project
+    parallelism: << parameters.shards >>
+    environment:
+      COVERAGE_CORE: sysmon
+      LITELLM_LOCAL_MODEL_COST_MAP: "True"
+    steps:
+      - checkout
+      - skip_unless_relevant:
+          base_ref: << parameters.base_ref >>
+          pull_request_url: << parameters.pull_request_url >>
+      - setup_test_deps
+      - when:
+          condition: << parameters.legacy_mcp_peer >>
+          steps:
+            - run:
+                name: Install MCP SDK1 peer
+                command: |
+                  uv venv --python 3.12 .venv-mcp-peer
+                  uv pip install --python .venv-mcp-peer 'mcp==1.28.1' 'langchain-mcp-adapters==0.2.1'
+                  echo "export MCP_TEST_PEER_PYTHON=$PWD/.venv-mcp-peer/bin/python" >> "$BASH_ENV"
+      - run:
+          name: "Run << parameters.flag >> shard"
+          no_output_timeout: 20m
+          command: |
+            mkdir -p test-results/<< parameters.flag >>
+            selection="$(bash .circleci/scripts/unit_selection.sh << parameters.flag >>)" || { echo "unit_selection.sh failed for << parameters.flag >>"; exit 1; }
+            [ -n "${selection}" ] || { echo "unit_selection.sh produced no files for << parameters.flag >>"; exit 1; }
+            shard="$(printf '%s\n' "${selection}" | circleci tests split --split-by=timings --timings-type=filename)" || { echo "circleci tests split failed for << parameters.flag >>"; exit 1; }
+            [ -n "${shard}" ] || { echo "shard ${CIRCLE_NODE_INDEX} received no << parameters.flag >> files; nothing to run"; exit 0; }
+            mapfile -t files < <(printf '%s\n' "${shard}")
+            xdist_args=()
+            if [ "<< parameters.workers >>" -gt 0 ]; then xdist_args=(-n << parameters.workers >> --dist=<< parameters.dist >>); fi
+            rerun_args=(-p no:rerunfailures)
+            if [ "<< parameters.reruns >>" -gt 0 ]; then rerun_args=(--reruns << parameters.reruns >> --reruns-delay 1 --rerun-except "from pytest-timeout"); fi
+            test_env=(PATH="$PATH" HOME="$HOME" CI=true COVERAGE_CORE="$COVERAGE_CORE" LITELLM_LOCAL_MODEL_COST_MAP="$LITELLM_LOCAL_MODEL_COST_MAP")
+            if [ -n "${MCP_TEST_PEER_PYTHON:-}" ]; then test_env+=(MCP_TEST_PEER_PYTHON="$MCP_TEST_PEER_PYTHON"); fi
+            set +e
+            env -i "${test_env[@]}" \
+              uv run --no-sync pytest "${files[@]}" "${rerun_args[@]}" -p no:pytest-retry --timeout=90 "${xdist_args[@]}" --tb=short --durations=20 -o junit_family=xunit1 --junitxml=test-results/<< parameters.flag >>/junit.xml --cov=./litellm --cov=./enterprise/litellm_enterprise --cov-report=xml:coverage.xml --cov-config=pyproject.toml
+            status=$?
+            set -e
+            if [ "$status" -eq 5 ]; then echo "pytest collected no tests from the shard; passing"; exit 0; fi
+            exit "$status"
+      - install_codecov_cli
+      - run:
+          name: Upload coverage
+          when: always
+          command: |
+            [ -f coverage.xml ] || { echo "no coverage.xml produced; skipping upload"; exit 0; }
+            codecov upload-process --disable-search -f coverage.xml -F << parameters.flag >> -C "$CIRCLE_SHA1" -n "<< parameters.flag >>-${CIRCLE_NODE_INDEX}-${CIRCLE_BUILD_NUM}" --git-service github
+      - store_test_results:
+          path: test-results
+      - store_artifacts:
+          path: test-results
+      - store_artifacts:
+          path: coverage.xml
+  documentation:
+    machine:
+      image: ubuntu-2204:2024.04.1
+    resource_class: large
+    working_directory: ~/project
+    steps:
+      - checkout
+      - setup_test_deps
+      - run:
+          name: Checkout litellm-docs
+          command: rm -rf docs/my-website && git clone --depth 1 https://github.com/BerriAI/litellm-docs.git docs/my-website
+      - run:
+          name: Run documentation validation
+          command: |
+            uv run --no-sync python ./tests/documentation_tests/test_env_keys.py
+            uv run --no-sync python ./tests/documentation_tests/test_router_settings.py
+            uv run --no-sync python ./tests/documentation_tests/test_api_docs.py
+            uv run --no-sync python ./tests/documentation_tests/test_circular_imports.py
+  integration:
+    parameters:
+      suite:
+        type: string
+      base_ref:
+        type: string
+        default: ""
+      pull_request_url:
+        type: string
+        default: ""
+    machine:
+      image: ubuntu-2204:2024.04.1
+    resource_class: large
+    working_directory: ~/project
+    steps:
+      - checkout
+      - skip_unless_relevant:
+          base_ref: << parameters.base_ref >>
+          pull_request_url: << parameters.pull_request_url >>
+      - setup_test_deps
+      - start_postgres:
+          image: postgres:16@sha256:e17e86066e5ef83e0952a9347f5c792b7ece00972e2aa787a6986f471b3dd3d5
+      - start_redis
+      - run:
+          name: Run owned integration contracts
+          command: env -i PATH="$PATH" HOME="$HOME" CIRCLE_SHA1="$CIRCLE_SHA1" CIRCLE_WORKFLOW_ID="$CIRCLE_WORKFLOW_ID" bash .circleci/scripts/run_integration.sh << parameters.suite >>
+          no_output_timeout: 15m
+      - run:
+          name: Stop owned database and Redis
+          when: always
+          command: |
+            mkdir -p test-results/integration-<< parameters.suite >>
+            docker logs postgres-db > test-results/integration-<< parameters.suite >>/postgres.log 2>&1 || true
+            docker logs redis-cache > test-results/integration-<< parameters.suite >>/redis.log 2>&1 || true
+            docker rm -f postgres-db redis-cache
+            test -z "$(docker ps -aq --filter name=postgres-db --filter name=redis-cache)"
+      - store_test_results:
+          path: test-results
+      - store_artifacts:
+          path: test-results
+
+workflows:
+  tests:
+    when: (pipeline.event.name == "push" and pipeline.git.branch == "main") or pipeline.event.name == "api" or (pipeline.event.name == "pull_request" and (pipeline.event.github.pull_request.base.ref == "main" or pipeline.event.github.pull_request.base.ref starts-with "litellm_"))
+    jobs:
+      - unit:
+          base_ref: << pipeline.event.name == "pull_request" and pipeline.event.github.pull_request.base.ref or "" >>
+          pull_request_url: << pipeline.event.name == "pull_request" and pipeline.event.github.pull_request.url or "" >>
+      - unit:
+          name: unit-<< matrix.flag >>
+          shards: 1
+          workers: 2
+          reruns: 2
+          matrix:
+            parameters:
+              flag: [caching-local, proxy-extras, enterprise-routing]
+          base_ref: << pipeline.event.name == "pull_request" and pipeline.event.github.pull_request.base.ref or "" >>
+          pull_request_url: << pipeline.event.name == "pull_request" and pipeline.event.github.pull_request.url or "" >>
+      - unit:
+          name: unit-mcp-integration
+          flag: mcp-integration
+          shards: 1
+          workers: 2
+          legacy_mcp_peer: true
+          base_ref: << pipeline.event.name == "pull_request" and pipeline.event.github.pull_request.base.ref or "" >>
+          pull_request_url: << pipeline.event.name == "pull_request" and pipeline.event.github.pull_request.url or "" >>
+      - unit:
+          name: unit-<< matrix.flag >>
+          shards: 1
+          reruns: 2
+          matrix:
+            parameters:
+              flag:
+                - enterprise-package
+                - proxy-infra
+                - responses-caching-types
+                - proxy-db-auth-checks
+                - proxy-db-jwt-and-keys
+                - proxy-db-proxy-server-core
+                - proxy-db-proxy-runtime
+                - proxy-db-custom-logging
+                - proxy-db-logging-misc
+                - proxy-db-db-and-spend
+                - proxy-db-guardrails-hooks
+                - proxy-db-budgets
+                - proxy-db-endpoints-and-responses
+          base_ref: << pipeline.event.name == "pull_request" and pipeline.event.github.pull_request.base.ref or "" >>
+          pull_request_url: << pipeline.event.name == "pull_request" and pipeline.event.github.pull_request.url or "" >>
+      - unit:
+          name: unit-llm-vertex-ai
+          flag: llm-vertex-ai
+          shards: 2
+          workers: 1
+          reruns: 2
+          base_ref: << pipeline.event.name == "pull_request" and pipeline.event.github.pull_request.base.ref or "" >>
+          pull_request_url: << pipeline.event.name == "pull_request" and pipeline.event.github.pull_request.url or "" >>
+      - unit:
+          name: unit-llm-other-providers
+          flag: llm-other-providers
+          shards: 3
+          reruns: 2
+          base_ref: << pipeline.event.name == "pull_request" and pipeline.event.github.pull_request.base.ref or "" >>
+          pull_request_url: << pipeline.event.name == "pull_request" and pipeline.event.github.pull_request.url or "" >>
+      - unit:
+          name: unit-integrations
+          flag: integrations
+          shards: 2
+          reruns: 3
+          base_ref: << pipeline.event.name == "pull_request" and pipeline.event.github.pull_request.base.ref or "" >>
+          pull_request_url: << pipeline.event.name == "pull_request" and pipeline.event.github.pull_request.url or "" >>
+      - unit:
+          name: unit-misc
+          flag: misc
+          shards: 2
+          reruns: 2
+          base_ref: << pipeline.event.name == "pull_request" and pipeline.event.github.pull_request.base.ref or "" >>
+          pull_request_url: << pipeline.event.name == "pull_request" and pipeline.event.github.pull_request.url or "" >>
+      - unit:
+          name: unit-proxy-db-proxy-utils
+          flag: proxy-db-proxy-utils
+          shards: 1
+          reruns: 2
+          dist: worksteal
+          base_ref: << pipeline.event.name == "pull_request" and pipeline.event.github.pull_request.base.ref or "" >>
+          pull_request_url: << pipeline.event.name == "pull_request" and pipeline.event.github.pull_request.url or "" >>
+      - unit:
+          name: unit-proxy-db-key-generation
+          flag: proxy-db-key-generation
+          shards: 1
+          workers: 0
+          reruns: 2
+          base_ref: << pipeline.event.name == "pull_request" and pipeline.event.github.pull_request.base.ref or "" >>
+          pull_request_url: << pipeline.event.name == "pull_request" and pipeline.event.github.pull_request.url or "" >>
+      - documentation
+      - integration:
+          name: integration-<< matrix.suite >>
+          matrix:
+            parameters:
+              suite: [sdk]
+          base_ref: << pipeline.event.name == "pull_request" and pipeline.event.github.pull_request.base.ref or "" >>
+          pull_request_url: << pipeline.event.name == "pull_request" and pipeline.event.github.pull_request.url or "" >>

.githooks/pre-push

@@ -8,7 +8,6 @@
 #
 # Protected branches (always allowed):
 #   - main
-#   - litellm_internal_staging
 #   - dependabot/*
 #   - gh-readonly-queue/*
 #
@@ -22,7 +21,7 @@ ZERO_OID_SHA256="000000000000000000000000000000000000000000000000000000000000000
 ALLOWED_TYPES="feature|bugfix|hotfix|release|chore"
 BRANCH_PATTERN="^(${ALLOWED_TYPES})/.+"
 
-PROTECTED_NAMES="main litellm_internal_staging"
+PROTECTED_NAMES="main"
 PROTECTED_PREFIXES="dependabot/ gh-readonly-queue/"
 
 is_protected() {
@@ -78,8 +77,7 @@ if [ -n "$invalid" ]; then
     chore/bump-deps
     hotfix/auth-bypass
 
-  Protected (always allowed): main, litellm_internal_staging,
-  dependabot/*, gh-readonly-queue/*.
+  Protected (always allowed): main, dependabot/*, gh-readonly-queue/*.
 
 See https://conventional-branch.github.io/
 

.github/ci-coverage-allowlist.yml

@@ -10,12 +10,13 @@ test_paths:
     paths:
       - tests/rust-python-harness
   - reason: >-
-      What is left of the caching suite in tests/local_testing that runs nowhere. Every job that
-      globs that directory either deselects it (local_testing_part1 and part2 carry `-k "... and
-      not caching and not cache"`) or keeps only another keyword (langfuse, router, assistants),
-      and no job names these files the way redis_caching_unit_tests names test_dual_cache.py.
-      The gap was eight files and 118 tests when measured 2026-08-20; the five keyless ones now
-      run in the caching-local shard, leaving these three. Measured 2026-08-21 with no provider
+      Live-provider caching cases in tests/local_testing that remain outside CI. Jobs that
+      glob that directory either deselect them (local_testing_part1 and part2 carry `-k "... and
+      not caching and not cache"`) or keep only another keyword (langfuse, router, assistants).
+      Separately, test-redis-compat.yml selects two IAM cluster authentication tests in
+      test_caching.py by node ID. It does not run that file's other tests.
+      The gap was eight files and 118 tests when measured 2026-08-20; the five keyless files now
+      run in the caching-local shard, leaving live cases in these three. Measured 2026-08-21 with no provider
       credentials and no Redis: test_caching.py needs both (37 of 65 fail without them),
       test_disk_cache_unit_tests.py needs OPENAI_API_KEY for 2 of its 4, and
       test_gcs_cache_unit_tests.py needs GCS credentials for all 4. They want the keyless/live

.github/merge-smoke-tests.json

@@ -0,0 +1,15 @@
+{
+  "cases": {
+    "CHAT-JSON": "tests/unit/llms/openai/test_openai.py::test_acompletion_returns_json_reply_over_injected_transport",
+    "CHAT-TEXT-STREAM": "tests/unit/llms/openai/test_openai.py::test_acompletion_streams_text_deltas_over_injected_transport",
+    "CHAT-TOOL-STREAM": "tests/unit/llms/openai/test_openai.py::test_acompletion_streams_tool_call_arguments_over_injected_transport",
+    "MODEL-ALLOW": "tests/test_litellm/proxy/auth/test_auth_checks.py::test_can_object_call_model_allows_listed_model_for_key",
+    "MODEL-DENY": "tests/test_litellm/proxy/auth/test_auth_checks.py::test_can_object_call_model_denials_return_forbidden[key-key_model_access_denied]",
+    "COST-EXPLICIT": "tests/unit/test_cost_calculator.py::test_completion_cost_charges_explicit_per_token_rates_over_registered_ones",
+    "COST-ZERO": "tests/unit/test_cost_calculator.py::test_completion_cost_is_zero_when_explicit_rates_are_zero",
+    "LOG-CONTENT-ON": "tests/test_litellm/litellm_core_utils/test_litellm_logging.py::test_standard_logging_payload_keeps_message_content_when_message_logging_is_on",
+    "LOG-CONTENT-OFF": "tests/test_litellm/litellm_core_utils/test_litellm_logging.py::test_standard_logging_payload_redacts_message_content_when_message_logging_is_off",
+    "CALLBACK-SUCCESS": "tests/test_litellm/litellm_core_utils/test_litellm_logging.py::test_async_success_handler_delivers_standard_logging_payload_to_custom_logger",
+    "CALLBACK-FAILURE": "tests/test_litellm/litellm_core_utils/test_litellm_logging.py::test_async_failure_handler_delivers_failure_payload_to_custom_logger"
+  }
+}

.github/pull_request_template.md

@@ -1,10 +1,13 @@
 <!-- The whole description's target audience is humans, not AI agents: write it in plain, simple,
      everyday engineering language, extremely parsable and readable at a glance. This goes double for
-     the TLDR, User Flow, and Caveats sections -->
+     the TLDR, User Flow, and Caveats sections
+     Drop every section you have nothing to put in, heading included: a bare "## Relevant issues" or
+     "## Affected release" with nothing under it must not appear in the final description -->
 
 ## TLDR
 
-<!-- Fill in the bullets below and keep each one short and concrete: one line per bullet, roughly 10 words max -->
+<!-- Fill in the bullets below and keep each one short and concrete: one line per bullet, roughly 10 words max
+     If the PR intentionally changes what existing users see or how a screen behaves, add a line under the bullets that starts "Intentional product change:" describing what changes, why, and what users lose. Reviewers must never have to infer a deliberate UX change from the diff -->
 
 Problem this solves:
 
@@ -21,11 +24,13 @@ How it solves it:
 <!-- Two ordered lists, Before and After, walking the same end user through the same task, written strictly from that user's seat
      Read the linked issue, ticket, or customer thread first so the flow reflects the real application and the routes its users actually hit; don't invent a generic scenario
      Lead each list with one plain sentence saying where the flow fails (Before) or succeeds (After), then number the steps
+     Keep it tight: aim for 3 to 5 steps per list, one line each, roughly 20 words max, and never pad a shorter flow with filler steps to hit the count. Cover the one path the PR changes and fold variants (case, other field, second endpoint) into a clause on the step they belong to rather than their own steps. The example below is the target length
      Every step is something the user does or observes: the HTTP method and full URL they hit, what they sent, and what visibly came back (status code, error text, the shape of an ID). UI steps name the page URL and what is on screen
      No LiteLLM internals: never name functions, files, DB tables, config classes, hooks, callbacks, or code paths. "The upload hands back an ID that looks like OpenAI's own `file-abc123` instead of the scrambled one the gateway returned" is right, "no managed-file row was registered" is wrong
      Keep the two lists step-for-step identical until they diverge, so the changed step is obvious
      If the bug had a security or authorization consequence, end each list with what another user could or could no longer do
-     Regenerate this section whenever new commits change the PR's behavior, so it never describes an older revision
+     Regenerate this section, screenshots included, whenever new commits change the PR's behavior, so it never describes an older revision
+     If the PR changes what an Admin UI page shows, embed a before and an after screenshot of that page right after its list, taken at the same URL on the same data, with the rows, fields, or controls that changed boxed in red so a reader spots the difference without reading the steps. These are the UI screenshots for Screenshots / Proof of Fix too: embed them once here and have that section's Before and After steps point back to them instead of repeating the images
 
 Example:
 
@@ -45,15 +50,15 @@ After: the same request comes back with real token counts, so the dashboard show
 
 ## Relevant issues
 
-<!-- e.g., "Fixes #000" -->
+<!-- e.g., "Fixes #000". Drop the section if there is none -->
 
 ## Affected release
 
-<!-- Only for a fix to a regression in a released or rc version (perf, memory, crash, or behavior): name the version it regressed in, e.g. "regression in v1.100.0" or "since v1.101.0-rc.1", and add the `backport-stable` label so the fix is cherry-picked onto the rc line before the stable is tagged. Leave the section blank otherwise -->
+<!-- Only for a fix to a regression in a released or rc version (perf, memory, crash, or behavior): name the version it regressed in, e.g. "regression in v1.100.0" or "since v1.101.0-rc.1", and add the `backport-stable` label so the fix is cherry-picked onto the rc line before the stable is tagged. Drop the section otherwise -->
 
 ## Linear ticket
 
-<!-- if you are an internal contributor, add "Resolves " followed by the Linear ticket e.g., "Resolves LIT-1234" to link the Linear ticket to the GitHub PR. If you don't have one, leave the section blank rather than guessing -->
+<!-- if you are an internal contributor, add "Resolves " followed by the Linear ticket e.g., "Resolves LIT-1234" to link the Linear ticket to the GitHub PR. If you don't have one, drop the section rather than guessing -->
 
 ## Pre-Submission checklist
 
@@ -134,7 +139,7 @@ If you're seeing a delay in your PR being merged, ping the LiteLLM Team on [Slac
      human reader
      If you assumed something instead of testing it, e.g. "only reproduces with X on" or "no
      user-observable behavior difference", list it here too with what breaks if it is wrong
-     Leave this section empty if there are none -->
+     Drop this section if there are none -->
 
 ## QA runbook
 

.github/scripts/assert_ci_coverage.py

@@ -34,7 +34,6 @@ GLOB_CHARS = frozenset("*?")
 # tests has to be named by some shard or it runs nowhere. A child listed here is
 # itself decomposed one level deeper and is checked through its own entry.
 SHARDED_ROOTS: tuple[str, ...] = (
-    "tests/proxy_unit_tests",
     "tests/test_litellm",
     "tests/test_litellm/proxy",
 )
@@ -120,6 +119,13 @@ def _invoked_test_tokens(scalars: Iterable[Scalar]) -> frozenset[str]:
     )
 
 
+def _unit_selection_tokens(repo_root: pathlib.Path = REPO_ROOT) -> frozenset[str]:
+    script: Final = repo_root / ".circleci/scripts/unit_selection.sh"
+    if not script.is_file():
+        return frozenset()
+    return frozenset(match.group(0).rstrip("/") for match in TEST_TOKEN_RE.finditer(_uncommented(script.read_text())))
+
+
 def _built_dockerfile_tokens(scalars: Iterable[Scalar]) -> frozenset[str]:
     return frozenset(
         match.group(0)
@@ -611,7 +617,10 @@ def main() -> int:
     scalars = _all_scalars()
 
     integration_paths, ownership_findings = _integration_ownership()
-    test_findings = _uncovered_tests(allowlist, _invoked_test_tokens(scalars) | integration_paths) + ownership_findings
+    test_findings = (
+        _uncovered_tests(allowlist, _invoked_test_tokens(scalars) | _unit_selection_tokens() | integration_paths)
+        + ownership_findings
+    )
     dockerfile_findings = _uncovered_dockerfiles(allowlist, _built_dockerfile_tokens(scalars))
     stale_findings = _stale_allowlist_paths(allowlist, test_files=_test_files(), dockerfiles=_dockerfiles())
 

.github/scripts/read_rc_version.py

@@ -0,0 +1,44 @@
+#!/usr/bin/env python3
+"""Print `version=X.Y.0` from [project].version in pyproject.toml for $GITHUB_OUTPUT.
+
+Usage
+-----
+    python3 read_rc_version.py [path/to/pyproject.toml] >> "$GITHUB_OUTPUT"
+
+Exit code 1 with a `::error::` line on stderr when the version is not an X.Y.0 release.
+"""
+
+from __future__ import annotations
+
+import pathlib
+import re
+import sys
+from typing import Final
+
+if sys.version_info >= (3, 11):
+    import tomllib
+else:
+    import tomli as tomllib
+
+RELEASE_VERSION: Final = re.compile(r"[0-9]+\.[0-9]+\.0")
+
+
+def read_version(pyproject: pathlib.Path) -> str:
+    with pyproject.open("rb") as f:
+        return tomllib.load(f)["project"]["version"]
+
+
+def main(argv: list[str]) -> int:
+    pyproject: Final = pathlib.Path(argv[1]) if len(argv) > 1 else pathlib.Path("pyproject.toml")
+    version: Final = read_version(pyproject)
+    if RELEASE_VERSION.fullmatch(version) is None:
+        print(  # noqa: T201  # the ::error:: line to stderr is the workflow's failure signal
+            f"::error::pyproject.toml version {version} is not an X.Y.0 release version", file=sys.stderr
+        )
+        return 1
+    print(f"version={version}")  # noqa: T201  # stdout line is appended to $GITHUB_OUTPUT
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main(sys.argv))

.github/scripts/run_merge_smoke.py

@@ -0,0 +1,493 @@
+#!/usr/bin/env python3
+"""Merge smoke harness: bounded checks run inside a loopback-only Linux network namespace."""
+
+# ruff: noqa: T201  # CLI harness: stdout/stderr lines are the reported result
+
+from __future__ import annotations
+
+import argparse
+import contextlib
+import http.client
+import json
+import os
+import secrets
+import signal
+import socket
+import subprocess
+import sys
+import time
+from collections import Counter
+from collections.abc import Sequence
+from dataclasses import dataclass, field
+from pathlib import Path
+from types import MappingProxyType
+from typing import Final, NoReturn, TextIO, cast
+
+import pytest
+
+EXPECTED_CASES: Final = (
+    "CHAT-JSON",
+    "CHAT-TEXT-STREAM",
+    "CHAT-TOOL-STREAM",
+    "MODEL-ALLOW",
+    "MODEL-DENY",
+    "COST-EXPLICIT",
+    "COST-ZERO",
+    "LOG-CONTENT-ON",
+    "LOG-CONTENT-OFF",
+    "CALLBACK-SUCCESS",
+    "CALLBACK-FAILURE",
+)
+
+
+@dataclass(frozen=True, slots=True)
+class CheckResult:
+    ok: bool
+    detail: str = ""
+
+
+@dataclass(slots=True)
+class _Args:
+    command: str = ""
+    no_child: bool = False
+    expect: str = ""
+    litellm_bin: str | None = None
+    lite_bin: str | None = None
+    diagnostics_dir: str = ""
+    ready_deadline: float = 120.0
+    shutdown_deadline: float = 20.0
+    poll_interval: float = 0.5
+    manifest: str = ""
+    rootdir: str | None = None
+
+
+def fail(reason: str) -> NoReturn:
+    print(f"merge-smoke: FAIL {reason}", file=sys.stderr)
+    sys.exit(1)
+
+
+def ok(step: str) -> None:
+    print(f"merge-smoke: OK {step}")
+
+
+def tail(path: Path, lines: int = 20) -> str:
+    try:
+        return "\n".join(path.read_text(errors="replace").splitlines()[-lines:])
+    except OSError as exc:
+        return f"<cannot read {path}: {exc}>"
+
+
+def cmd_verify_isolation(args: _Args) -> int:
+    if os.geteuid() == 0:
+        fail("verify-isolation must run unprivileged (geteuid()==0)")
+    try:
+        socket.create_connection(("192.0.2.1", 9), timeout=3)
+    except OSError as exc:
+        print(f"external connect blocked as expected: errno={exc.errno} {exc}")
+    else:
+        fail("external TCP connect to 192.0.2.1:9 succeeded; namespace is not isolated")
+    listener: Final = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
+    listener.bind(("127.0.0.1", 0))
+    listener.listen(1)
+    port: Final = cast(int, listener.getsockname()[1])
+    client: Final = socket.create_connection(("127.0.0.1", port), timeout=5)
+    accepted: Final = listener.accept()
+    accepted[0].close()
+    client.close()
+    listener.close()
+    print(f"loopback connect ok on 127.0.0.1:{port}")
+    if not args.no_child:
+        proc: Final = subprocess.run(
+            [sys.executable, str(Path(__file__).resolve()), "verify-isolation", "--no-child"],
+            timeout=30,
+            capture_output=True,
+            text=True,
+        )
+        if proc.returncode != 0:
+            fail(f"child process did not inherit isolation: {proc.stderr.strip()}")
+        print("child process inherits isolation")
+    ok("verify-isolation")
+    return 0
+
+
+def cmd_interpreter(args: _Args) -> int:
+    print(sys.version)
+    print(sys.executable)
+    actual: Final = f"{sys.version_info.major}.{sys.version_info.minor}"
+    if actual != args.expect:
+        fail(f"interpreter is {actual}, expected {args.expect}")
+    ok(f"interpreter {actual}")
+    return 0
+
+
+def _run_cli(argv: Sequence[str], label: str) -> CheckResult:
+    try:
+        proc: Final = subprocess.run(list(argv), timeout=120, capture_output=True, text=True)
+    except subprocess.TimeoutExpired:
+        return CheckResult(ok=False, detail=f"{label} timed out after 120s")
+    sys.stdout.write(proc.stdout)
+    sys.stderr.write(proc.stderr)
+    if proc.returncode != 0:
+        return CheckResult(ok=False, detail=f"{label} exited {proc.returncode}")
+    return CheckResult(ok=True)
+
+
+def cmd_cli(args: _Args) -> int:
+    venv_bin: Final = Path(sys.executable).parent
+    litellm_bin: Final = Path(args.litellm_bin) if args.litellm_bin else venv_bin / "litellm"
+    lite_bin: Final = Path(args.lite_bin) if args.lite_bin else venv_bin / "lite"
+    commands: Final = (
+        ("import litellm", [sys.executable, "-c", "import litellm"]),
+        ("litellm --version", [str(litellm_bin), "--version"]),
+        ("lite version", [str(lite_bin), "version"]),
+    )
+    for label, argv in commands:
+        result = _run_cli(argv, label)
+        if not result.ok:
+            fail(result.detail)
+        ok(label)
+    return 0
+
+
+def _free_port() -> int:
+    sock: Final = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
+    sock.bind(("127.0.0.1", 0))
+    port: Final = cast(int, sock.getsockname()[1])
+    sock.close()
+    return port
+
+
+_CONFIG_TEMPLATE: Final = """model_list:
+  - model_name: smoke-model
+    litellm_params:
+      model: openai/smoke-model
+      api_base: http://127.0.0.1:9/v1
+      api_key: synthetic-key
+general_settings:
+  master_key: os.environ/LITELLM_MASTER_KEY
+"""
+
+
+def _listen_inode(port: int) -> str | None:
+    target: Final = f"{port:04X}"
+    for table in ("/proc/net/tcp", "/proc/net/tcp6"):
+        try:
+            rows = Path(table).read_text().splitlines()[1:]
+        except OSError:
+            continue
+        for row in rows:
+            cols = row.split()
+            if len(cols) > 9 and cols[3] == "0A" and cols[1].rsplit(":", 1)[-1] == target:
+                return cols[9]
+    return None
+
+
+def _ancestors(pid: int) -> frozenset[int]:
+    chain: Final[set[int]] = set()
+    pending: Final[list[int]] = [pid]
+    while pending:
+        current = pending.pop()
+        if current <= 0 or current in chain:
+            continue
+        chain.add(current)
+        try:
+            stat = Path(f"/proc/{current}/stat").read_text()
+        except OSError:
+            continue
+        pending.append(int(stat.rpartition(")")[2].split()[1]))
+    return frozenset(chain)
+
+
+def _socket_owner_pid(inode: str) -> int | None:
+    for proc_dir in Path("/proc").iterdir():
+        if not proc_dir.name.isdigit():
+            continue
+        fd_dir = proc_dir / "fd"
+        try:
+            for fd in fd_dir.iterdir():
+                try:
+                    if os.readlink(fd) == f"socket:[{inode}]":
+                        return int(proc_dir.name)
+                except OSError:
+                    continue
+        except OSError:
+            continue
+    return None
+
+
+def _verify_port_owner(port: int, proc: subprocess.Popen[bytes]) -> CheckResult:
+    inode: Final = _listen_inode(port)
+    if inode is None:
+        return CheckResult(ok=False, detail=f"no LISTEN socket found for port {port} in /proc/net/tcp")
+    owner: Final = _socket_owner_pid(inode)
+    if owner is None:
+        return CheckResult(ok=False, detail=f"no process owns the listen socket inode {inode} for port {port}")
+    if owner != proc.pid and proc.pid not in _ancestors(owner):
+        return CheckResult(
+            ok=False, detail=f"port {port} owned by pid {owner} outside the launched process group {proc.pid}"
+        )
+    if proc.poll() is not None:
+        return CheckResult(ok=False, detail=f"proxy exited with code {proc.returncode} after readiness")
+    return CheckResult(ok=True)
+
+
+def cmd_proxy_startup(args: _Args) -> int:
+    diagnostics: Final = Path(args.diagnostics_dir)
+    diagnostics.mkdir(parents=True, exist_ok=True)
+    venv_bin: Final = Path(sys.executable).parent
+    litellm_bin: Final = Path(args.litellm_bin) if args.litellm_bin else venv_bin / "litellm"
+    port: Final = _free_port()
+    master_key: Final = "sk-smoke-" + secrets.token_hex(16)
+    config_path: Final = diagnostics / "config.yaml"
+    config_path.write_text(_CONFIG_TEMPLATE)
+    log_path: Final = diagnostics / "proxy.log"
+    result_path: Final = diagnostics / "result.json"
+    outcome: Final[dict[str, object]] = {
+        "port": port,
+        "time_to_ready_s": None,
+        "shutdown_s": None,
+        "readiness": None,
+        "outcome": "failed",
+    }
+    log_file: Final = log_path.open("w")
+    env: Final = {
+        **os.environ,
+        "LITELLM_MASTER_KEY": master_key,
+        "LITELLM_LOCAL_MODEL_COST_MAP": "True",
+    }
+    started: Final = time.monotonic()
+    proc: Final = subprocess.Popen(
+        [str(litellm_bin), "--config", str(config_path), "--host", "127.0.0.1", "--port", str(port)],
+        stdout=log_file,
+        stderr=subprocess.STDOUT,
+        start_new_session=True,
+        env=env,
+    )
+    body: str | None = None
+    last_status: int | None = None
+    while time.monotonic() - started < args.ready_deadline:
+        if proc.poll() is not None:
+            log_file.close()
+            result_path.write_text(json.dumps(outcome))
+            fail(f"proxy exited early with code {proc.returncode}\n{tail(log_path)}")
+        try:
+            conn = http.client.HTTPConnection("127.0.0.1", port, timeout=5)
+            conn.request("GET", "/health/readiness")
+            resp = conn.getresponse()
+            last_status = resp.status
+            candidate = resp.read().decode()
+            conn.close()
+        except (http.client.HTTPException, ConnectionError, OSError):
+            time.sleep(args.poll_interval)
+            continue
+        if last_status == 200:
+            body = candidate
+            break
+        time.sleep(args.poll_interval)
+    outcome["time_to_ready_s"] = round(time.monotonic() - started, 3)
+    if body is None:
+        _terminate(proc, log_file)
+        result_path.write_text(json.dumps(outcome))
+        detail = f"last status {last_status}" if last_status is not None else "no response"
+        fail(f"readiness not reached within {args.ready_deadline}s ({detail})\n{tail(log_path)}")
+    outcome["readiness"] = body
+    try:
+        readiness = cast(object, json.loads(body))
+    except json.JSONDecodeError:
+        readiness = None
+    if readiness != {"status": "healthy", "db": "Not connected"}:
+        _terminate(proc, log_file)
+        result_path.write_text(json.dumps(outcome))
+        fail(f"unexpected readiness body: {body}")
+    owner_check: Final = _verify_port_owner(port, proc)
+    if not owner_check.ok:
+        _terminate(proc, log_file)
+        result_path.write_text(json.dumps(outcome))
+        fail(owner_check.detail)
+    shutdown_started: Final = time.monotonic()
+    os.killpg(proc.pid, signal.SIGTERM)
+    try:
+        proc.wait(timeout=args.shutdown_deadline)
+    except subprocess.TimeoutExpired:
+        os.killpg(proc.pid, signal.SIGKILL)
+        proc.wait(timeout=10)
+        outcome["shutdown_s"] = round(time.monotonic() - shutdown_started, 3)
+        log_file.close()
+        result_path.write_text(json.dumps(outcome))
+        fail(f"forced kill after {args.shutdown_deadline}s\n{tail(log_path)}")
+    outcome["shutdown_s"] = round(time.monotonic() - shutdown_started, 3)
+    try:
+        os.killpg(proc.pid, 0)
+    except ProcessLookupError:
+        pass
+    else:
+        os.killpg(proc.pid, signal.SIGKILL)
+        log_file.close()
+        result_path.write_text(json.dumps(outcome))
+        fail("process group survived SIGTERM")
+    log_file.close()
+    outcome["outcome"] = "ok"
+    result_path.write_text(json.dumps(outcome))
+    ok(f"proxy-startup ready={outcome['time_to_ready_s']}s shutdown={outcome['shutdown_s']}s")
+    return 0
+
+
+def _terminate(proc: subprocess.Popen[bytes], log_file: TextIO) -> None:
+    with contextlib.suppress(ProcessLookupError):
+        os.killpg(proc.pid, signal.SIGTERM)
+    try:
+        proc.wait(timeout=10)
+    except subprocess.TimeoutExpired:
+        with contextlib.suppress(ProcessLookupError):
+            os.killpg(proc.pid, signal.SIGKILL)
+        with contextlib.suppress(subprocess.TimeoutExpired):
+            proc.wait(timeout=10)
+    log_file.close()
+
+
+def _load_manifest(path: Path) -> MappingProxyType[str, str]:
+    def no_duplicates(pairs: list[tuple[object, object]]) -> dict[object, object]:
+        seen: dict[object, object] = {}
+        for key, value in pairs:
+            if key in seen:
+                raise ValueError(f"duplicate key in manifest: {key}")
+            seen[key] = value
+        return seen
+
+    raw_value: object = cast(object, json.loads(path.read_text(), object_pairs_hook=no_duplicates))
+    if not isinstance(raw_value, dict):
+        raise ValueError("manifest must be an object")
+    loaded: Final = cast(dict[object, object], raw_value)
+    cases_value: object = loaded.get("cases")
+    if not isinstance(cases_value, dict):
+        raise ValueError("manifest must be an object with a 'cases' object")
+    cases_any: Final = cast(dict[object, object], cases_value)
+    cases: Final = {k: v for k, v in cases_any.items() if isinstance(k, str) and isinstance(v, str)}
+    if len(cases) != len(cases_any):
+        raise ValueError("manifest 'cases' must map string ids to string node ids")
+    return MappingProxyType(cases)
+
+
+@dataclass(slots=True, eq=False)
+class _Recorder:
+    collect_failed: list[str] = field(default_factory=list)
+    collected: tuple[str, ...] = ()
+    reports: dict[str, list[tuple[str, str, bool]]] = field(default_factory=dict)
+
+    def pytest_collectreport(self, report: pytest.CollectReport) -> None:
+        if report.failed:
+            self.collect_failed.append(report.nodeid)
+
+    def pytest_collection_finish(self, session: pytest.Session) -> None:
+        self.collected = tuple(item.nodeid for item in session.items)
+
+    def pytest_runtest_logreport(self, report: pytest.TestReport) -> None:
+        self.reports.setdefault(report.nodeid, []).append((report.when, report.outcome, hasattr(report, "wasxfail")))
+
+
+def cmd_pytest(args: _Args) -> int:
+    try:
+        cases: Final = _load_manifest(Path(args.manifest))
+    except (OSError, ValueError, json.JSONDecodeError) as exc:
+        fail(f"manifest invalid: {exc}")
+    if tuple(cases) != EXPECTED_CASES:
+        fail(f"manifest case ids must be exactly {list(EXPECTED_CASES)} in order, got {list(cases)}")
+    node_ids: Final = tuple(cases.values())
+    if len(set(node_ids)) != len(node_ids):
+        fail("manifest node ids are not unique")
+    argv: Final = [
+        *node_ids,
+        "-p",
+        "no:cacheprovider",
+        "-p",
+        "no:xdist",
+        "-p",
+        "no:rerunfailures",
+        "-p",
+        "no:randomly",
+        "-rA",
+        "-q",
+        *(["--rootdir", args.rootdir] if args.rootdir else []),
+    ]
+
+    recorder: Final = _Recorder()
+    code: Final = pytest.main(argv, plugins=[recorder])
+    name_of: Final = MappingProxyType({node_id: case_id for case_id, node_id in cases.items()})
+    problems: Final[list[str]] = []
+    if code != 0:
+        problems.append(f"pytest exit code {code}")
+    for failed_id in recorder.collect_failed:
+        problems.append(f"collection failed: {name_of.get(failed_id, failed_id)}")
+    expected: Final = Counter(node_ids)
+    collected: Final = Counter(recorder.collected)
+    for node_id in expected - collected:
+        problems.append(f"missing case {name_of[node_id]} ({node_id})")
+    for node_id in collected - expected:
+        problems.append(f"unexpected test collected: {node_id}")
+    for node_id, count in collected.items():
+        if count > 1:
+            problems.append(f"duplicated test id: {node_id}")
+    if len(recorder.collected) != len(EXPECTED_CASES):
+        problems.append(f"collected {len(recorder.collected)} tests, expected {len(EXPECTED_CASES)}")
+    rows: Final[list[tuple[str, bool]]] = []
+    for case_id, node_id in cases.items():
+        reports = recorder.reports.get(node_id, [])
+        case_ok = (
+            bool(reports)
+            and all(outcome == "passed" and not wasxfail for _, outcome, wasxfail in reports)
+            and {when for when, _, _ in reports} >= {"setup", "call", "teardown"}
+        )
+        rows.append((case_id, case_ok))
+        if not reports:
+            problems.append(f"{case_id} ({node_id}) produced no runtest reports")
+            continue
+        for when, outcome, wasxfail in reports:
+            if outcome != "passed":
+                problems.append(f"{case_id} ({node_id}) {when} outcome={outcome}")
+            if wasxfail:
+                problems.append(f"{case_id} ({node_id}) {when} was xfail/xpass")
+        missing_phases = {"setup", "call", "teardown"} - {when for when, _, _ in reports}
+        for phase in sorted(missing_phases):
+            problems.append(f"{case_id} ({node_id}) missing {phase} report")
+    for case_id, passed in rows:
+        print(f"{case_id}  {'PASS' if passed else 'FAIL'}  {cases[case_id]}")
+    if problems:
+        for problem in problems:
+            print(f"merge-smoke: {problem}", file=sys.stderr)
+        fail("pytest verdict failed")
+    ok("pytest 11 cases")
+    return 0
+
+
+def main() -> int:
+    parser: Final = argparse.ArgumentParser(description=__doc__)
+    subs: Final = parser.add_subparsers(dest="command", required=True)
+    p_iso: Final = subs.add_parser("verify-isolation")
+    p_iso.add_argument("--no-child", action="store_true")
+    p_interp: Final = subs.add_parser("interpreter")
+    p_interp.add_argument("--expect", required=True)
+    p_cli: Final = subs.add_parser("cli")
+    p_cli.add_argument("--litellm-bin", default=None)
+    p_cli.add_argument("--lite-bin", default=None)
+    p_proxy: Final = subs.add_parser("proxy-startup")
+    p_proxy.add_argument("--diagnostics-dir", required=True)
+    p_proxy.add_argument("--litellm-bin", default=None)
+    p_proxy.add_argument("--ready-deadline", type=float, default=120)
+    p_proxy.add_argument("--shutdown-deadline", type=float, default=20)
+    p_proxy.add_argument("--poll-interval", type=float, default=0.5)
+    p_test: Final = subs.add_parser("pytest")
+    p_test.add_argument("--manifest", required=True)
+    p_test.add_argument("--rootdir", default=None)
+    args: Final = parser.parse_args(namespace=_Args())
+    handlers: Final = {
+        "verify-isolation": cmd_verify_isolation,
+        "interpreter": cmd_interpreter,
+        "cli": cmd_cli,
+        "proxy-startup": cmd_proxy_startup,
+        "pytest": cmd_pytest,
+    }
+    return handlers[args.command](args)
+
+
+if __name__ == "__main__":
+    sys.exit(main())

.github/scripts/verify_linux_native_wheel.py

@@ -214,7 +214,7 @@ def main(
     native_module: Final = load_native_module(native_path)
     native_module_loads: Final = native_module is not None
     panic_test_hook_absent: Final = native_module is not None and not hasattr(native_module, "_panic_for_test")
-    native_size_limit: Final = 35_000_000
+    native_size_limit: Final = 40_000_000
     native_size_within_limit: Final = native_member.file_size <= native_size_limit
     validations: Final = (
         (f"Python tag is {EXPECTED_PYTHON_TAG}", python_tag == EXPECTED_PYTHON_TAG),

.github/workflows/_test-unit-base.yml

@@ -13,6 +13,16 @@ on:
           have its path existence-checked like any other token.
         required: true
         type: string
+      unit-flag:
+        description: >-
+          Codecov flag of the `.circleci/tests.yml` job that now owns part of
+          this shard. The shard also runs the files
+          `.circleci/scripts/unit_selection.sh` lists for the flag, on every
+          event, because the CircleCI pipeline is manual-only while the tests
+          migrate.
+        required: false
+        type: string
+        default: ""
       workers:
         description: "Number of pytest-xdist workers"
         required: false
@@ -92,6 +102,7 @@ jobs:
       pull-requests: read
     outputs:
       decision: ${{ steps.changes.outputs.decision }}
+      has-coverage: ${{ steps.tests.outputs.has-coverage }}
 
     steps:
       - uses: actions/checkout@08eba0b27e820071cde6df949e0beb9ba4906955 # v4.3.0
@@ -160,10 +171,12 @@ jobs:
           uv run --no-sync prisma generate --schema litellm/proxy/schema.prisma
 
       - name: Run tests
+        id: tests
         if: steps.changes.outputs.decision != 'skip'
         timeout-minutes: ${{ inputs.timeout-minutes }}
         env:
           TEST_PATH: ${{ inputs.test-path }}
+          UNIT_FLAG: ${{ inputs.unit-flag }}
           MAX_FAILURES: ${{ inputs.max-failures }}
           WORKERS: ${{ inputs.workers }}
           RERUNS: ${{ inputs.reruns }}
@@ -171,9 +184,18 @@ jobs:
           DIST: ${{ inputs.dist }}
           COVERAGE_CORE: sysmon
         run: |
+          echo "has-coverage=false" >> "$GITHUB_OUTPUT"
+          selection="${TEST_PATH}"
+          if [ -n "${UNIT_FLAG}" ]; then
+            selection="${TEST_PATH} $(bash .circleci/scripts/unit_selection.sh "${UNIT_FLAG}" | tr '\n' ' ')"
+          fi
+          if [ -z "${selection// /}" ]; then
+            echo "shard selection is empty; nothing to run"
+            exit 0
+          fi
           pytest_args=()
           existing_paths=0
-          for token in ${TEST_PATH:?}; do
+          for token in ${selection}; do
             case "${token}" in
               -*) pytest_args+=("${token}") ;;
               *)
@@ -187,7 +209,7 @@ jobs:
             esac
           done
           if [ "${existing_paths}" -eq 0 ]; then
-            echo "No path in TEST_PATH exists (${TEST_PATH}); nothing to run"
+            echo "No path in the selection exists (${selection}); nothing to run"
             exit 0
           fi
           xdist_args=()
@@ -209,8 +231,11 @@ jobs:
             --cov-config=pyproject.toml
           status=$?
           set -e
+          if [ -f coverage.xml ]; then
+            echo "has-coverage=true" >> "$GITHUB_OUTPUT"
+          fi
           if [ "$status" -eq 5 ]; then
-            echo "pytest collected no tests from ${TEST_PATH}; passing"
+            echo "pytest collected no tests from ${selection}; passing"
             exit 0
           fi
           exit "$status"
@@ -226,7 +251,7 @@ jobs:
   upload-coverage:
     name: Upload coverage to Codecov
     needs: run
-    if: always() && needs.run.outputs.decision != 'skip'
+    if: always() && needs.run.outputs.decision != 'skip' && needs.run.outputs.has-coverage == 'true'
     runs-on: ubuntu-latest
     permissions:
       contents: read

.github/workflows/check-ui-api-types.yml

@@ -4,8 +4,6 @@ on:
   pull_request:
     branches:
       - main
-      - litellm_internal_staging
-      - litellm_oss_staging
       - "litellm_**"
 
 permissions:

.github/workflows/ci-coverage.yml

@@ -4,13 +4,10 @@ on:
   pull_request:
     branches:
       - main
-      - litellm_internal_staging
-      - litellm_oss_staging
       - "litellm_**"
   push:
     branches:
       - main
-      - litellm_internal_staging
 
 permissions:
   contents: read

.github/workflows/codeql.yml

@@ -67,12 +67,18 @@ jobs:
       # further up the stack are modified. The suppression is scoped to this one
       # file/rule pair via SARIF post-filtering so every other callsite of
       # py/weak-sensitive-data-hashing in the repository continues to be analyzed.
-      - name: Filter SARIF (OCI sha256)
+      # The same query fires on the HIBP k-anonymity lookup in
+      # litellm/proxy/auth/password_policy.py, where the password's SHA-1 is only
+      # a lookup key into the haveibeenpwned range API (the protocol mandates
+      # SHA-1) and the digest itself never leaves the proxy beyond its first 5
+      # characters.
+      - name: Filter SARIF (OCI sha256, HIBP sha1)
         if: matrix.language == 'python'
         uses: advanced-security/filter-sarif@2da736ff05ef065cb2894ac6892e47b5eac2c3c0 # v1.1
         with:
           patterns: |
             -litellm/llms/oci/common_utils.py:py/weak-sensitive-data-hashing
+            -litellm/proxy/auth/password_policy.py:py/weak-sensitive-data-hashing
           input: sarif-results/python.sarif
           output: sarif-results/python.sarif
 

.github/workflows/codspeed.yml

@@ -4,7 +4,6 @@ on:
   push:
     branches:
       - main
-      - litellm_internal_staging
     paths:
       - "litellm/**"
       - "tests/benchmarks/**"
@@ -17,7 +16,6 @@ on:
   pull_request:
     branches:
       - main
-      - litellm_internal_staging
     paths:
       - "litellm/**"
       - "tests/benchmarks/**"

.github/workflows/compat-matrix-image.yml

@@ -4,6 +4,7 @@ on:
   pull_request:
     paths:
       - tests/e2e/claude_code/cron_vm/**
+      - tests/e2e/claude_code/pr_gate_version_resolver.py
       - .github/workflows/compat-matrix-image.yml
   workflow_dispatch:
 
@@ -28,6 +29,14 @@ jobs:
       - name: Build the Render cron image
         run: docker build -f tests/e2e/claude_code/cron_vm/Dockerfile -t compat-matrix:${{ github.sha }} tests/e2e
 
-      - name: Run the pinned binaries as the cron user
+      - name: Resolve and install the Claude Code CLI as the cron user
         run: |
-          docker run --rm compat-matrix:${{ github.sha }} bash -c 'set -e; whoami; claude --version; gh --version; uv --version'
+          docker run --rm compat-matrix:${{ github.sha }} bash -c '
+            set -euo pipefail
+            whoami
+            gh --version
+            uv --version
+            version="$(uv run --no-project --python 3.12 python /opt/litellm/tests/e2e/claude_code/pr_gate_version_resolver.py)"
+            /opt/litellm/tests/e2e/claude_code/cron_vm/install_claude_code.sh "${version}" /tmp/claude-cli
+            /tmp/claude-cli/claude --version
+          '

.github/workflows/cost-map-guard.yml

@@ -4,8 +4,6 @@ on: # zizmor: ignore[dangerous-triggers] runs the base branch's code only; the P
   pull_request_target:
     branches:
       - main
-      - litellm_internal_staging
-      - litellm_oss_staging
       - "litellm_**"
 
 permissions:

.github/workflows/create-rc-branch.yml

@@ -0,0 +1,66 @@
+name: Create RC Branch
+
+on:
+  schedule:
+    - cron: "0 3 * * 5"
+      timezone: "America/Los_Angeles"
+  workflow_dispatch:
+
+permissions: {}
+
+jobs:
+  create-rc-branch:
+    name: Create RC Branch
+    if: github.event_name != 'schedule' || github.repository == 'BerriAI/litellm'
+    runs-on: ubuntu-latest
+    permissions:
+      contents: write
+    steps:
+      - name: Require main
+        env:
+          REF: ${{ github.ref }}
+        run: |
+          if [ "$REF" != "refs/heads/main" ]; then
+            echo "::error::rc branches are cut from refs/heads/main only, got $REF"
+            exit 1
+          fi
+
+      - uses: actions/checkout@08eba0b27e820071cde6df949e0beb9ba4906955 # v4.3.0
+        with:
+          persist-credentials: false
+
+      - name: Read release version
+        id: version
+        run: python3 .github/scripts/read_rc_version.py >> "$GITHUB_OUTPUT"
+
+      - name: Create rc branch
+        env:
+          VERSION: ${{ steps.version.outputs.version }}
+        uses: actions/github-script@60a0d83039c74a4aee543508d2ffcb1c3799cdea # v7.0.1
+        with:
+          script: |
+            const branchName = `rc/${process.env.VERSION}`;
+            const ref = `heads/${branchName}`;
+
+            const existing = await github.rest.git.getRef({
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+              ref,
+            }).catch((error) => {
+              if (error.status === 404) {
+                return null;
+              }
+              throw error;
+            });
+            if (existing !== null) {
+              core.setFailed(`Branch ${branchName} already exists at ${existing.data.object.sha}; leaving it untouched`);
+              return;
+            }
+
+            await github.rest.git.createRef({
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+              ref: `refs/${ref}`,
+              sha: context.sha,
+            });
+            core.info(`Created branch ${branchName} at ${context.sha}`);

.github/workflows/create_daily_staging_branch.yml

@@ -1,49 +0,0 @@
-name: Create Daily Staging Branch
-
-on:
-  schedule:
-    - cron: "0 0,12 * * *" # Runs every 12 hours at midnight and noon UTC
-  workflow_dispatch: # Allow manual trigger
-
-jobs:
-  create-staging-branch:
-    if: github.repository == 'BerriAI/litellm'
-    runs-on: ubuntu-latest
-    permissions:
-      contents: write
-
-    steps:
-      - name: Create daily staging branch
-        env:
-          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-        run: |
-          BRANCH_NAME="litellm_oss_staging_$(date +'%m_%d_%Y')"
-          echo "Creating branch: $BRANCH_NAME"
-          if gh api "repos/${{ github.repository }}/git/ref/heads/$BRANCH_NAME" --silent 2>/dev/null; then
-            echo "Branch $BRANCH_NAME already exists. Skipping creation."
-            exit 0
-          fi
-          MAIN_SHA=$(gh api "repos/${{ github.repository }}/git/ref/heads/main" --jq '.object.sha')
-          gh api "repos/${{ github.repository }}/git/refs" -f ref="refs/heads/$BRANCH_NAME" -f sha="$MAIN_SHA" --silent
-          echo "Successfully created branch: $BRANCH_NAME at $MAIN_SHA"
-
-  create-internal-dev-branch:
-    if: github.repository == 'BerriAI/litellm'
-    runs-on: ubuntu-latest
-    permissions:
-      contents: write
-
-    steps:
-      - name: Create internal dev branch
-        env:
-          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-        run: |
-          BRANCH_NAME="litellm_internal_dev_$(date +'%m_%d_%Y')"
-          echo "Creating branch: $BRANCH_NAME"
-          if gh api "repos/${{ github.repository }}/git/ref/heads/$BRANCH_NAME" --silent 2>/dev/null; then
-            echo "Branch $BRANCH_NAME already exists. Skipping creation."
-            exit 0
-          fi
-          MAIN_SHA=$(gh api "repos/${{ github.repository }}/git/ref/heads/main" --jq '.object.sha')
-          gh api "repos/${{ github.repository }}/git/refs" -f ref="refs/heads/$BRANCH_NAME" -f sha="$MAIN_SHA" --silent
-          echo "Successfully created branch: $BRANCH_NAME at $MAIN_SHA"

.github/workflows/guard-fork-dependencies.yml

@@ -4,8 +4,6 @@ on:
   pull_request:
     branches:
       - main
-      - litellm_internal_staging
-      - litellm_oss_staging
       - "litellm_**"
     paths:
       - "uv.lock"

.github/workflows/image-scan.yml

@@ -4,7 +4,6 @@ on:
   pull_request:
     branches:
       - main
-      - litellm_internal_staging
       - litellm_oss_branch
       - "litellm_**"
     paths:

.github/workflows/osv-scan.yml

@@ -4,8 +4,6 @@ on:
   pull_request:
     branches:
       - main
-      - litellm_internal_staging
-      - litellm_oss_staging
       - "litellm_**"
   schedule:
     - cron: "23 6 * * *"

.github/workflows/publish-basedpyright-base-counts.yml

@@ -1,6 +1,6 @@
 name: Publish basedpyright base counts
 
-# Every commit on main or litellm_internal_staging can become a future merge-base.
+# Every commit on main can become a future merge-base.
 # Publishing its per-rule basedpyright counts as an artifact lets
 # scripts/type_check_gate.py download them in seconds instead of paying a
 # 60-110s second basedpyright pass on every fresh worktree or moved merge-base.
@@ -11,7 +11,6 @@ on:
   push:
     branches:
       - main
-      - litellm_internal_staging
   workflow_dispatch:
     inputs:
       ref:

.github/workflows/test-code-quality.yml

@@ -4,13 +4,10 @@ on:
   pull_request:
     branches:
       - main
-      - litellm_internal_staging
-      - litellm_oss_staging
       - "litellm_**"
   push:
     branches:
       - main
-      - litellm_internal_staging
 
 permissions:
   contents: read
@@ -83,6 +80,9 @@ jobs:
       - name: test_e2e_changed_gate
         run: uv run --no-sync pytest -q --noconftest -p no:cacheprovider -c /dev/null tests/code_coverage_tests/test_e2e_changed_gate.py tests/code_coverage_tests/test_e2e_idp_stack.py
 
+      - name: Check merge smoke harness
+        run: uv run --no-sync pytest -q --noconftest -p no:cacheprovider -c /dev/null tests/code_coverage_tests/test_merge_smoke.py
+
       - name: router_code_coverage
         run: uv run --no-sync python ./tests/code_coverage_tests/router_code_coverage.py
 

.github/workflows/test-linting.yml

@@ -4,8 +4,6 @@ on:
   pull_request:
     branches:
       - main
-      - litellm_internal_staging
-      - litellm_oss_staging
       - "litellm_**"
 
 permissions:
@@ -182,6 +180,18 @@ jobs:
             echo "No changed tests/e2e Python files; skipping."
           fi
 
+      - name: Run the claude_code harness unit tests
+        if: steps.changes.outputs.decision != 'skip'
+        run: |
+          if ! git diff --name-only --diff-filter=ACMRD "$GATE_BASE_SHA" HEAD -- ':(glob)tests/e2e/claude_code/**/*.py' ':(glob)tests/e2e/*.py' tests/e2e/claude_code/cron_vm/install_claude_code.sh pyproject.toml uv.lock .github/workflows/test-linting.yml | grep -q .; then
+            echo "No changed claude_code harness files; skipping."
+            exit 0
+          fi
+          retry() { "$@" || { sleep 15; "$@"; } || { sleep 30; "$@"; }; }
+          CLAUDE_VERSION="$(retry uv run --no-sync python tests/e2e/claude_code/pr_gate_version_resolver.py)"
+          tests/e2e/claude_code/cron_vm/install_claude_code.sh "$CLAUDE_VERSION" "$RUNNER_TEMP/claude-cli"
+          PATH="$RUNNER_TEMP/claude-cli:$PATH" uv run --no-sync pytest -q --noconftest -o addopts= -o pythonpath=tests/e2e -p no:rerunfailures tests/e2e/claude_code/_*_unit_tests
+
       - name: Check for circular imports
         if: steps.changes.outputs.decision != 'skip'
         run: |

.github/workflows/test-litellm-ui-build.yml

@@ -7,8 +7,6 @@ on:
   pull_request:
     branches:
       - main
-      - litellm_internal_staging
-      - litellm_oss_staging
       - "litellm_**"
 
 concurrency:

.github/workflows/test-litellm-ui-lint.yml

@@ -6,8 +6,6 @@ on:
   pull_request:
     branches:
       - main
-      - litellm_internal_staging
-      - litellm_oss_staging
       - "litellm_**"
 
 concurrency:

.github/workflows/test-litellm-ui-unit.yml

@@ -7,13 +7,10 @@ on:
   pull_request:
     branches:
       - main
-      - litellm_internal_staging
-      - litellm_oss_staging
       - "litellm_**"
   push:
     branches:
       - main
-      - litellm_internal_staging
 
 concurrency:
   group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}

.github/workflows/test-mcp-dependency-resolution.yml

@@ -4,8 +4,6 @@ on:
   pull_request:
     branches:
       - main
-      - litellm_internal_staging
-      - litellm_oss_staging
       - "litellm_**"
 
 permissions:

.github/workflows/test-merge-smoke.yml

@@ -0,0 +1,95 @@
+name: Merge smoke checks
+
+on:
+  pull_request:
+    branches: [main, litellm_internal_staging, litellm_oss_staging, "litellm_**"]
+  workflow_dispatch:
+
+permissions:
+  contents: read
+
+concurrency:
+  group: merge-smoke-${{ github.event.pull_request.number || github.run_id }}
+  cancel-in-progress: true
+
+jobs:
+  dashboard-build:
+    name: Dashboard build
+    runs-on: ubuntu-24.04
+    timeout-minutes: 30
+    steps:
+      - name: Checkout
+        uses: actions/checkout@08eba0b27e820071cde6df949e0beb9ba4906955 # v4.3.0
+        with:
+          persist-credentials: false
+
+      - name: Build the dashboard stage
+        run: docker build --target ui-builder -f Dockerfile .
+
+  core-checks:
+    name: Core checks (Python ${{ matrix.python-version }})
+    runs-on: ubuntu-24.04
+    timeout-minutes: 30
+    strategy:
+      fail-fast: false
+      matrix:
+        python-version: ["3.10", "3.11", "3.12", "3.13", "3.14"]
+    env:
+      LITELLM_LOCAL_MODEL_COST_MAP: "True"
+    steps:
+      - name: Checkout
+        uses: actions/checkout@08eba0b27e820071cde6df949e0beb9ba4906955 # v4.3.0
+        with:
+          persist-credentials: false
+
+      - name: Set up Python
+        uses: actions/setup-python@a26af69be951a213d495a4c3e4e4022e16d87065 # v5.6.0
+        with:
+          python-version: ${{ matrix.python-version }}
+
+      - name: Set up uv
+        uses: ./.github/actions/setup-uv-with-retries
+        with:
+          version: "0.10.9"
+
+      - name: Install dependencies
+        run: .github/scripts/uv_sync_with_retries.sh --frozen --extra proxy --extra cli --group dev --group proxy-dev --python ${{ matrix.python-version }}
+
+      - name: Create the loopback-only network namespace
+        run: |
+          sudo ip netns add smoke
+          sudo ip netns exec smoke ip link set lo up
+          cat > "${RUNNER_TEMP}/in-netns" <<'WRAP'
+          #!/usr/bin/env bash
+          set -euo pipefail
+          exec sudo --preserve-env=LITELLM_LOCAL_MODEL_COST_MAP ip netns exec smoke setpriv --reuid "$(id -u)" --regid "$(id -g)" --init-groups -- env HOME="${HOME}" PATH="${PATH}" "$@"
+          WRAP
+          chmod +x "${RUNNER_TEMP}/in-netns"
+          echo "IN_NETNS=${RUNNER_TEMP}/in-netns" >> "${GITHUB_ENV}"
+
+      - name: Verify namespace isolation
+        run: $IN_NETNS .venv/bin/python .github/scripts/run_merge_smoke.py verify-isolation
+
+      - name: Verify interpreter version
+        run: $IN_NETNS .venv/bin/python .github/scripts/run_merge_smoke.py interpreter --expect ${{ matrix.python-version }}
+
+      - name: Import and CLI checks
+        run: $IN_NETNS .venv/bin/python .github/scripts/run_merge_smoke.py cli
+
+      - name: Proxy startup check
+        run: $IN_NETNS .venv/bin/python .github/scripts/run_merge_smoke.py proxy-startup --diagnostics-dir "${RUNNER_TEMP}/smoke-diagnostics"
+
+      - name: Run curated smoke cases
+        run: $IN_NETNS .venv/bin/python .github/scripts/run_merge_smoke.py pytest --manifest .github/merge-smoke-tests.json
+
+      - name: Upload smoke diagnostics
+        if: always()
+        uses: actions/upload-artifact@4cec3d8aa04e39d1a68397de0c4cd6fb9dce8ec1 # v4.6.1
+        with:
+          name: merge-smoke-diagnostics-py${{ matrix.python-version }}
+          path: ${{ runner.temp }}/smoke-diagnostics
+          if-no-files-found: ignore
+
+      - name: Remove the network namespace
+        if: always()
+        run: sudo ip netns delete smoke

.github/workflows/test-postgres.yml

@@ -4,13 +4,10 @@ on:
   pull_request:
     branches:
       - main
-      - litellm_internal_staging
-      - litellm_oss_staging
       - "litellm_**"
   push:
     branches:
       - main
-      - litellm_internal_staging
   workflow_dispatch:
 
 permissions:

.github/workflows/test-redis-compat.yml

@@ -4,14 +4,17 @@ on:
   pull_request:
     branches:
       - main
-      - litellm_internal_staging
-      - litellm_oss_staging
       - "litellm_**"
     paths:
       - "litellm/_redis.py"
       - "litellm/_redis_credential_provider.py"
-      - "tests/test_litellm/test_redis.py"
+      - "litellm/caching/redis_cache.py"
+      - "litellm/caching/evicted_client_closer.py"
+      - "tests/unit/test_redis.py"
+      - "tests/local_testing/test_caching.py"
       - "tests/test_litellm/caching/test_redis_connection_pool.py"
+      - "tests/test_litellm/caching/test_redis_cluster_cache.py"
+      - "tests/test_litellm/caching/test_evicted_client_closer.py"
       - ".github/workflows/test-redis-compat.yml"
       - "pyproject.toml"
       - "uv.lock"
@@ -28,6 +31,9 @@ jobs:
     name: "redis-py ${{ matrix.redis-version }}"
     runs-on: ubuntu-latest
     timeout-minutes: 15
+    permissions:
+      contents: read
+      id-token: write
 
     strategy:
       fail-fast: false
@@ -57,7 +63,7 @@ jobs:
 
       - name: Install dependencies
         run: |
-          .github/scripts/uv_sync_with_retries.sh --frozen --group ci --group proxy-dev --extra google --extra proxy --extra semantic-router
+          .github/scripts/uv_sync_with_retries.sh --frozen --group ci --group proxy-dev --extra google --extra proxy --extra extra_proxy --extra semantic-router
 
       - name: Pin redis-py to the matrix version
         env:
@@ -66,12 +72,35 @@ jobs:
           uv pip install "redis==${REDIS_VERSION:?}"
           uv run --no-sync python -c "import redis; assert redis.__version__ == '${REDIS_VERSION:?}', redis.__version__; print('redis-py', redis.__version__)"
 
+      - name: Build Redis for cluster authentication tests
+        run: |
+          curl --fail --location --retry 3 https://download.redis.io/releases/redis-7.2.16.tar.gz -o "$RUNNER_TEMP/redis-7.2.16.tar.gz"
+          echo "960a8ec15e34ff40e57ff16837b26b33bd81f2da6d24497bb63de532a323a18e  $RUNNER_TEMP/redis-7.2.16.tar.gz" | sha256sum --check
+          tar -xzf "$RUNNER_TEMP/redis-7.2.16.tar.gz" -C "$RUNNER_TEMP"
+          make -C "$RUNNER_TEMP/redis-7.2.16" -j2 MALLOC=libc OPTIMIZATION=-O1 redis-server
+          echo "$RUNNER_TEMP/redis-7.2.16/src" >> "$GITHUB_PATH"
+
       - name: Run redis unit tests
         run: |
+          redis-server --version
           uv run --no-sync pytest \
-            tests/test_litellm/test_redis.py \
+            tests/unit/test_redis.py \
             tests/test_litellm/caching/test_redis_connection_pool.py \
+            tests/test_litellm/caching/test_redis_cluster_cache.py \
+            tests/test_litellm/caching/test_evicted_client_closer.py \
+            tests/local_testing/test_caching.py::test_sync_cluster_authenticates_with_azure_credentials \
+            tests/local_testing/test_caching.py::test_sync_cluster_authenticates_with_gcp_credentials \
             --tb=short -vv \
             --reruns 2 \
             --reruns-delay 1 \
-            --durations=20
+            --durations=20 \
+            --cov=./litellm --cov-report=xml:coverage-redis.xml
+
+      - name: Upload Redis coverage
+        if: matrix.redis-version == '5.3.1'
+        uses: codecov/codecov-action@75cd11691c0faa626561e295848008c8a7dddffe # v5.5.4
+        with:
+          use_oidc: true
+          files: coverage-redis.xml
+          flags: redis-compat
+          fail_ci_if_error: false

.github/workflows/test-rust.yml

@@ -29,8 +29,6 @@ on:
   pull_request:
     branches:
       - main
-      - litellm_internal_staging
-      - litellm_oss_staging
       - "litellm_**"
     paths:
       - "litellm-rust/**"

.github/workflows/test-semgrep.yml

@@ -4,8 +4,6 @@ on:
   pull_request:
     branches:
       - main
-      - litellm_internal_staging
-      - litellm_oss_staging
       - "litellm_**"
 
 permissions:

.github/workflows/test-terraform-modules.yml

@@ -9,8 +9,6 @@ on:
   pull_request:
     branches:
       - main
-      - litellm_internal_staging
-      - litellm_oss_staging
       - "litellm_**"
     paths:
       - "terraform/litellm/aws/**"

.github/workflows/test-terraform-provider.yml

@@ -8,8 +8,6 @@ on:
   pull_request:
     branches:
       - main
-      - litellm_internal_staging
-      - litellm_oss_staging
       - "litellm_**"
     paths:
       - "terraform/provider/**"

.github/workflows/test-unit-documentation.yml

@@ -4,13 +4,10 @@ on:
   pull_request:
     branches:
       - main
-      - litellm_internal_staging
-      - litellm_oss_staging
       - "litellm_**"
   push:
     branches:
       - main
-      - litellm_internal_staging
 
 permissions:
   contents: read

.github/workflows/test-unit-proxy-db.yml

@@ -4,13 +4,10 @@ on:
   pull_request:
     branches:
       - main
-      - litellm_internal_staging
-      - litellm_oss_staging
       - "litellm_**"
   push:
     branches:
       - main
-      - litellm_internal_staging
 
 permissions:
   contents: read
@@ -23,6 +20,13 @@ concurrency:
 # rather than alphabetical letter ranges. Adding a new test file means adding it
 # to whichever group it belongs to, not reshuffling slices.
 #
+# `.circleci/tests.yml` runs each group's files on same-repo events under the
+# `proxy-db-<group>` Codecov flag; `.circleci/scripts/unit_selection.sh` holds
+# the file lists. That pipeline is manual-only while the tests migrate, so
+# `unit-flag` makes the shard run that list on every event. `test-path` keeps
+# the files that still reach real providers and never left
+# tests/proxy_unit_tests.
+#
 # Design targets:
 #   * Every shard runs in <= 7 minutes of wall-clock on the default runner.
 #     Most of a shard's time is pytest plugin load + xdist worker imports +
@@ -61,7 +65,7 @@ jobs:
   proxy-db:
     needs: assert-shard-coverage
     # Display only the semantic shard name in the checks UI instead of GHA's
-    # default "proxy-db (key-generation, tests/proxy_unit_tests/…, 0, loadscope, 20)"
+    # default "proxy-db (key-generation, tests/unit/proxy/…, 0, loadscope, 20)"
     # which includes every matrix field and gets truncated past the test-path.
     name: ${{ matrix.test-group }}
     permissions:
@@ -74,132 +78,93 @@ jobs:
         include:
           # Must run serially — event-loop conflict with the logging worker.
           - test-group: key-generation
-            test-path: "tests/proxy_unit_tests/test_key_generate_prisma.py"
+            test-path: ""
+            unit-flag: proxy-db-key-generation
             workers: 0
             dist: loadscope
             timeout: 20
 
           # ---- auth: split into 2 shards ----
           - test-group: auth-checks
-            test-path: >-
-              tests/proxy_unit_tests/test_auth_checks.py
-              tests/proxy_unit_tests/test_user_api_key_auth.py
-              tests/proxy_unit_tests/test_deprecated_key_grace_period.py
+            test-path: ""
+            unit-flag: proxy-db-auth-checks
             workers: 4
             dist: loadscope
             timeout: 15
           - test-group: jwt-and-keys
-            test-path: >-
-              tests/proxy_unit_tests/test_jwt.py
-              tests/proxy_unit_tests/test_jwt_key_mapping.py
-              tests/proxy_unit_tests/test_proxy_custom_auth.py
-              tests/proxy_unit_tests/test_key_generate_dynamodb.py
+            test-path: ""
+            unit-flag: proxy-db-jwt-and-keys
             workers: 4
             dist: loadscope
             timeout: 15
 
           # ---- test_proxy_utils.py, single shard, worksteal distribution ----
           - test-group: proxy-utils
-            test-path: "tests/proxy_unit_tests/test_proxy_utils.py"
+            test-path: ""
+            unit-flag: proxy-db-proxy-utils
             workers: 4
             dist: worksteal
             timeout: 15
 
           # ---- proxy server: split into 2 shards ----
           - test-group: proxy-server-core
-            test-path: >-
-              tests/proxy_unit_tests/test_proxy_server.py
-              tests/proxy_unit_tests/test_aproxy_startup.py
+            test-path: "tests/proxy_unit_tests/test_proxy_server_gemini_pass_through.py"
+            unit-flag: proxy-db-proxy-server-core
             workers: 4
             dist: loadscope
             timeout: 15
           - test-group: proxy-runtime
-            test-path: >-
-              tests/proxy_unit_tests/test_proxy_config_unit_test.py
-              tests/proxy_unit_tests/test_proxy_routes.py
-              tests/proxy_unit_tests/test_server_root_path.py
-              tests/proxy_unit_tests/test_proxy_pass_user_config.py
-              tests/proxy_unit_tests/test_proxy_token_counter.py
-              tests/proxy_unit_tests/test_request_size_limit_middleware.py
-              tests/proxy_unit_tests/test_multipart_bypass_repro.py
+            test-path: ""
+            unit-flag: proxy-db-proxy-runtime
             workers: 4
             dist: loadscope
             timeout: 15
 
           # ---- logging: split into 2 shards ----
           - test-group: custom-logging
-            test-path: >-
-              tests/proxy_unit_tests/test_custom_callback_input.py
-              tests/proxy_unit_tests/test_custom_logger_s3_gcs.py
-              tests/proxy_unit_tests/test_proxy_custom_logger.py
+            test-path: "tests/proxy_unit_tests/test_proxy_custom_logger.py"
+            unit-flag: proxy-db-custom-logging
             workers: 4
             dist: loadscope
             timeout: 15
           - test-group: logging-misc
-            test-path: >-
-              tests/proxy_unit_tests/test_proxy_reject_logging.py
-              tests/proxy_unit_tests/test_audit_logs_proxy.py
-              tests/proxy_unit_tests/test_search_api_logging.py
+            test-path: ""
+            unit-flag: proxy-db-logging-misc
             workers: 4
             dist: loadscope
             timeout: 15
 
           - test-group: db-and-spend
-            test-path: >-
-              tests/proxy_unit_tests/test_prisma_client_backoff_retry.py
-              tests/proxy_unit_tests/test_db_schema_changes.py
-              tests/proxy_unit_tests/test_e2e_pod_lock_manager.py
-              tests/proxy_unit_tests/test_skills_db.py
-              tests/proxy_unit_tests/test_update_daily_tag_spend.py
-              tests/proxy_unit_tests/test_update_spend.py
-              tests/proxy_unit_tests/test_proxy_encrypt_decrypt.py
+            test-path: ""
+            unit-flag: proxy-db-db-and-spend
             workers: 4
             dist: loadscope
             timeout: 15
 
           # ---- guardrails + budget + hooks: split into 2 ----
           - test-group: guardrails-hooks
-            test-path: >-
-              tests/proxy_unit_tests/test_proxy_setting_guardrails.py
-              tests/proxy_unit_tests/test_banned_keyword_list.py
-              tests/proxy_unit_tests/test_unit_test_proxy_hooks.py
+            test-path: ""
+            unit-flag: proxy-db-guardrails-hooks
             workers: 4
             dist: loadscope
             timeout: 15
           - test-group: budgets
-            test-path: >-
-              tests/proxy_unit_tests/test_default_end_user_budget_simple.py
-              tests/proxy_unit_tests/test_unit_test_max_model_budget_limiter.py
-              tests/proxy_unit_tests/test_zero_cost_model_budget_bypass.py
+            test-path: ""
+            unit-flag: proxy-db-budgets
             workers: 4
             dist: loadscope
             timeout: 15
 
           - test-group: endpoints-and-responses
-            test-path: >-
-              tests/proxy_unit_tests/test_blog_posts_endpoint.py
-              tests/proxy_unit_tests/test_models_fallback_endpoint.py
-              tests/proxy_unit_tests/test_google_endpoint_routing.py
-              tests/proxy_unit_tests/test_google_gemini_proxy_request.py
-              tests/proxy_unit_tests/test_gemini_agents_endpoints.py
-              tests/proxy_unit_tests/test_get_favicon.py
-              tests/proxy_unit_tests/test_get_image.py
-              tests/proxy_unit_tests/test_reducto_ocr_route.py
-              tests/proxy_unit_tests/test_ui_path_detection.py
-              tests/proxy_unit_tests/test_prompt_test_endpoint.py
-              tests/proxy_unit_tests/test_check_batch_cost.py
-              tests/proxy_unit_tests/test_check_responses_cost.py
-              tests/proxy_unit_tests/test_response_polling_handler.py
-              tests/proxy_unit_tests/test_response_polling_pre_call_checks.py
-              tests/proxy_unit_tests/test_realtime_cache.py
-              tests/proxy_unit_tests/test_proxy_exception_mapping.py
-              tests/proxy_unit_tests/test_custom_tokenizer_bug.py
+            test-path: "tests/proxy_unit_tests/test_proxy_exception_mapping.py"
+            unit-flag: proxy-db-endpoints-and-responses
             workers: 4
             dist: loadscope
             timeout: 15
     uses: ./.github/workflows/_test-unit-base.yml
     with:
       test-path: ${{ matrix.test-path }}
+      unit-flag: ${{ matrix.unit-flag }}
       workers: ${{ matrix.workers }}
       reruns: 2
       timeout-minutes: ${{ matrix.timeout }}

.github/workflows/test-unit.yml

@@ -4,13 +4,10 @@ on:
   pull_request:
     branches:
       - main
-      - litellm_internal_staging
-      - litellm_oss_staging
       - "litellm_**"
   push:
     branches:
       - main
-      - litellm_internal_staging
   workflow_dispatch:
 
 permissions:
@@ -34,10 +31,14 @@ concurrency:
 # number, so a partially-specified entry would fail the call rather than fall
 # back to the default.
 #
-# tests/proxy_unit_tests keeps its own caller (test-unit-proxy-db.yml): it is
-# already a matrix and carries a shard-coverage guard that reads that file by
-# name. Folding it in here is a follow-up, together with generalising that guard
-# into assert_ci_coverage.py.
+# tests/unit/proxy keeps its own caller (test-unit-proxy-db.yml): it is already
+# a matrix and carries a shard-coverage guard that reads that file by name.
+# Folding it in here is a follow-up, together with generalising that guard into
+# assert_ci_coverage.py.
+#
+# `unit-flag` names the `.circleci/tests.yml` job that now runs part of the
+# shard under the same Codecov flag. That pipeline is manual-only while the
+# tests migrate, so the shard also runs those files on every event.
 jobs:
   unit:
     name: ${{ matrix.shard }}
@@ -51,7 +52,8 @@ jobs:
         include:
           - shard: mcp-integration
             artifact-name: mcp-integration
-            test-path: "tests/mcp_tests tests/test_litellm/experimental_mcp_client"
+            test-path: "tests/mcp_tests"
+            unit-flag: mcp-integration
             workers: 2
             reruns: 0
             timeout-minutes: 20
@@ -68,10 +70,9 @@ jobs:
           - shard: enterprise-routing
             artifact-name: enterprise-routing
             test-path: >-
-              tests/test_litellm/enterprise
-              tests/test_litellm/google_genai
               tests/test_litellm/router_utils
               tests/test_litellm/router_strategy
+            unit-flag: enterprise-routing
             workers: 2
             reruns: 2
             timeout-minutes: 20
@@ -79,7 +80,8 @@ jobs:
 
           - shard: integrations
             artifact-name: integrations
-            test-path: "tests/test_litellm/integrations"
+            test-path: ""
+            unit-flag: integrations
             workers: 2
             reruns: 3
             timeout-minutes: 20
@@ -88,6 +90,7 @@ jobs:
           - shard: Vertex AI
             artifact-name: llm-vertex-ai
             test-path: "tests/test_litellm/llms/vertex_ai"
+            unit-flag: llm-vertex-ai
             workers: 1
             reruns: 2
             timeout-minutes: 20
@@ -96,6 +99,7 @@ jobs:
           - shard: All Other Providers
             artifact-name: llm-other-providers
             test-path: "tests/test_litellm/llms --ignore=tests/test_litellm/llms/vertex_ai"
+            unit-flag: llm-other-providers
             workers: 2
             reruns: 2
             timeout-minutes: 20
@@ -104,24 +108,12 @@ jobs:
           - shard: misc
             artifact-name: misc
             test-path: >-
-              tests/test_litellm/batches
-              tests/test_litellm/secret_managers
-              tests/test_litellm/a2a_protocol
-              tests/test_litellm/chat_completions
-              tests/test_litellm/completion_extras
-              tests/test_litellm/containers
-              tests/test_litellm/endpoints
-              tests/test_litellm/images
               tests/test_litellm/interactions
-              tests/test_litellm/messages
               tests/test_litellm/ocr
               tests/test_litellm/passthrough
-              tests/test_litellm/rag
-              tests/test_litellm/rerank_api
               tests/test_litellm/rust_bridge
-              tests/test_litellm/vector_stores
-              tests/test_litellm/videos
               tests/test_litellm/test_*.py
+            unit-flag: misc
             workers: 2
             reruns: 2
             timeout-minutes: 20
@@ -201,7 +193,7 @@ jobs:
               tests/test_litellm/proxy/types_utils
               tests/test_litellm/proxy/logging_endpoints
               tests/test_litellm/proxy/test_*.py
-              tests/test_gateway
+            unit-flag: proxy-infra
             workers: 4
             reruns: 2
             timeout-minutes: 20
@@ -209,11 +201,8 @@ jobs:
 
           - shard: caching-local
             artifact-name: caching-local
-            test-path: >-
-              tests/local_testing/test_cache_preset_key.py
-              tests/local_testing/test_caching_handler.py
-              tests/local_testing/test_responses_stream_cache_keys.py
-              tests/local_testing/test_unit_test_caching.py
+            test-path: ""
+            unit-flag: caching-local
             workers: 2
             reruns: 2
             timeout-minutes: 20
@@ -221,7 +210,8 @@ jobs:
 
           - shard: proxy-extras
             artifact-name: proxy-extras
-            test-path: "tests/litellm-proxy-extras"
+            test-path: ""
+            unit-flag: proxy-extras
             workers: 2
             reruns: 2
             timeout-minutes: 20
@@ -229,7 +219,8 @@ jobs:
 
           - shard: enterprise-package
             artifact-name: enterprise-package
-            test-path: "tests/enterprise"
+            test-path: ""
+            unit-flag: enterprise-package
             workers: 4
             reruns: 2
             timeout-minutes: 20
@@ -240,7 +231,7 @@ jobs:
             test-path: >-
               tests/test_litellm/responses
               tests/test_litellm/caching
-              tests/test_litellm/types
+            unit-flag: responses-caching-types
             workers: 2
             reruns: 2
             timeout-minutes: 20
@@ -248,6 +239,7 @@ jobs:
     uses: ./.github/workflows/_test-unit-base.yml
     with:
       test-path: ${{ matrix.test-path }}
+      unit-flag: ${{ matrix.unit-flag || '' }}
       workers: ${{ matrix.workers }}
       reruns: ${{ matrix.reruns }}
       timeout-minutes: ${{ matrix.timeout-minutes }}

.github/workflows/test-vscode-extension.yml

@@ -6,8 +6,6 @@ on:
   pull_request:
     branches:
       - main
-      - litellm_internal_staging
-      - litellm_oss_staging
       - "litellm_**"
     paths:
       - "vscode-extension/**"

.github/workflows/zizmor.yml

@@ -2,12 +2,10 @@ name: GitHub Actions Security Analysis
 
 on:
   push:
-    branches: [main, litellm_internal_staging]
+    branches: [main]
   pull_request:
     branches:
       - main
-      - litellm_internal_staging
-      - litellm_oss_staging
       - "litellm_**"
 
 concurrency:

AGENTS.md

@@ -33,11 +33,11 @@ End-to-end tests belong in `tests/e2e/` and must follow the harness conventions
 
 When creating PRs, target the repository's current default branch for both internal and external / OSS contributions. Check it with `python3 scripts/default_branch.py --branch` instead of assuming a branch name or relying on cached `origin/HEAD`
 
-When writing a PR body, treat the comments and imperative instructions inside .github/pull_request_template.md as rules to follow, not just layout. Agent harnesses may strip HTML comments from copies of that file injected into context, so read .github/pull_request_template.md from disk before writing a PR body to make sure you see every comment rule
+When writing a PR body, treat the comments and imperative instructions inside .github/pull_request_template.md as rules to follow, not just layout. Agent harnesses may strip HTML comments from copies of that file injected into context, so read .github/pull_request_template.md from disk before writing a PR body to make sure you see every comment rule. A section you have nothing to put in (Relevant issues, Affected release, Linear ticket, Caveats, QA runbook, and so on) is removed entirely, heading included, never left as an empty title
 
 Same applies for filing bug reports and feature requests, with .github/ISSUE_TEMPLATE/bug_report.yml and .github/ISSUE_TEMPLATE/feature_request.yml, respectively
 
-If you're resolving a linear ticket, in the "## Linear ticket" section of the PR, say "Resolves LIT-1234", replacing "LIT-1234" with the actual ticket id that you're resolving. If you don't have the ticket id, don't make one up or search for it. Just leave the section blank
+If you're resolving a linear ticket, in the "## Linear ticket" section of the PR, say "Resolves LIT-1234", replacing "LIT-1234" with the actual ticket id that you're resolving. If you don't have the ticket id, don't make one up or search for it. Just drop the section
 
 Never use `pytest` commands or the like as "Screenshots / Proof of Fix". We prefer curl'ing a live proxy instance running on localhost:4000 (I like to run it with `python litellm/proxy/proxy_cli.py --config litellm/proxy/dev_config.yaml --detailed_debug --reload --use_v2_migration_resolver 2>&1 | tee litellm.log`; the Admin UI dev server is `npm run dev` in `ui/litellm-dashboard`, served on port 3000) and showing both the command run and the output. Also, it should hit real LLM provider APIs, not mocks, and cost real $$$ because that is the most realistic test. The proof of fix should be exactly what the end user / customer would see / do. The run logs in PR #27703 is a prime example of how to do it (not a huge fan of using a python test script that future me and the team will have no visibility into; I prefer just curl commands or a short list of bash commands (e.g., using `for`)). If it's a UI thing, or the main use case runs through a headful agentic coding tool like Claude Code or Codex, drive that surface yourself and embed your own before and after screenshots of it in the PR (the Admin UI page, or what the coding tool shows), next to an ordered list of the URLs to go to (e.g., http://localhost:4000/ui/?page=logs), where to click, and what fields to fill out so a reviewer can reproduce it
 
@@ -96,6 +96,7 @@ Follow these coding conventions for new/updated code (a three-line fix in a lega
 - No mutation; don't reassign variables, global or local. Instead of mutable lists and dicts, prefer tuples, frozen dataclasses (with slots=True), `MappingProxyType`, etc.
   - Annotate every variable with `: Final` (LIT010). Unpacking and walrus targets cannot carry the annotation, so they are implicitly final. Don't rebind them. Never rebind or mutate function parameters (LIT011); `self`/`cls` attribute stores are the exception. If rebinding or in-place mutation is truly unavoidable, suppress with `# rebind-ok: <reason>`
   - Qualify every TypedDict field with `ReadOnly[...]` (LIT012), which nests freely with `Required` / `NotRequired` / `Annotated` in any order. If making the key writable is truly unavoidable, suppress with `# writable-ok: <reason>`
+  - Comprehensions take at most one `for` clause and one `if` clause (LIT014); split stacked clauses into a helper generator, a named intermediate, or a plain loop. Suppress with `# comprehension-ok: <reason>` only when unavoidable
 - Use dependency injection
 - Fully typed; no `Any` or coarse types like `dict[str, Any]` or just `dict`. Every function parameter must be strongly typed
 - Use tagged unions + match

Makefile

@@ -51,8 +51,8 @@ help:
 	@echo "  make test-unit-core-utils - Run core utils tests (~32 files)"
 	@echo "  make test-unit-other    - Run other tests (caching, responses, etc., ~69 files)"
 	@echo "  make test-unit-root     - Run root-level tests (~34 files)"
-	@echo "  make test-proxy-unit-a  - Run proxy_unit_tests (a-o, ~20 files)"
-	@echo "  make test-proxy-unit-b  - Run proxy_unit_tests (p-z, ~28 files)"
+	@echo "  make test-proxy-unit-a  - Run tests/unit/proxy (a-o)"
+	@echo "  make test-proxy-unit-b  - Run tests/unit/proxy (p-z)"
 	@echo "  make test-integration   - Run integration tests"
 	@echo "  make test-unit-helm     - Run helm unit tests"
 	@echo "  make test-rust-extension - Build the Rust extension and run its public Python tests"
@@ -314,7 +314,7 @@ test-unit: install-test-deps
 
 # Matrix test targets (matching CI workflow groups)
 test-unit-llms: install-test-deps
-	$(UV_RUN) pytest tests/test_litellm/llms --tb=short -vv -n 4 --durations=20
+	$(UV_RUN) pytest tests/unit/llms --tb=short -vv -n 4 --durations=20
 
 test-unit-proxy-guardrails: install-test-deps
 	$(UV_RUN) pytest tests/test_litellm/proxy/guardrails tests/test_litellm/proxy/management_endpoints tests/test_litellm/proxy/management_helpers --tb=short -vv -n 4 --durations=20
@@ -326,23 +326,23 @@ test-unit-proxy-misc: install-test-deps
 	$(UV_RUN) pytest tests/test_litellm/proxy/_experimental tests/test_litellm/proxy/agent_endpoints tests/test_litellm/proxy/anthropic_endpoints tests/test_litellm/proxy/common_utils tests/test_litellm/proxy/discovery_endpoints tests/test_litellm/proxy/experimental tests/test_litellm/proxy/google_endpoints tests/test_litellm/proxy/health_endpoints tests/test_litellm/proxy/image_endpoints tests/test_litellm/proxy/middleware tests/test_litellm/proxy/openai_files_endpoint tests/test_litellm/proxy/pass_through_endpoints tests/test_litellm/proxy/prompts tests/test_litellm/proxy/public_endpoints tests/test_litellm/proxy/response_api_endpoints tests/test_litellm/proxy/shutdown tests/test_litellm/proxy/spend_tracking tests/test_litellm/proxy/ui_crud_endpoints tests/test_litellm/proxy/vector_store_endpoints tests/test_litellm/proxy/test_*.py --tb=short -vv -n 4 --durations=20
 
 test-unit-integrations: install-test-deps
-	$(UV_RUN) pytest tests/test_litellm/integrations --tb=short -vv -n 4 --durations=20
+	$(UV_RUN) pytest tests/unit/integrations --tb=short -vv -n 4 --durations=20
 
 test-unit-core-utils: install-test-deps
 	$(UV_RUN) pytest tests/test_litellm/litellm_core_utils --tb=short -vv -n 2 --durations=20
 
 test-unit-other: install-test-deps
-	$(UV_RUN) pytest tests/test_litellm/caching tests/test_litellm/responses tests/test_litellm/secret_managers tests/test_litellm/vector_stores tests/test_litellm/a2a_protocol tests/test_litellm/anthropic_interface tests/test_litellm/completion_extras tests/test_litellm/containers tests/test_litellm/enterprise tests/test_litellm/experimental_mcp_client tests/test_litellm/google_genai tests/test_litellm/images tests/test_litellm/interactions tests/test_litellm/passthrough tests/test_litellm/router_strategy tests/test_litellm/router_utils tests/test_litellm/types --tb=short -vv -n 4 --durations=20
+	$(UV_RUN) pytest tests/test_litellm/caching tests/test_litellm/responses tests/unit/secret_managers tests/unit/vector_stores tests/unit/a2a_protocol tests/test_litellm/anthropic_interface tests/unit/completion_extras tests/unit/containers tests/unit/enterprise tests/unit/experimental_mcp_client tests/unit/google_genai tests/unit/images tests/unit/interactions tests/test_litellm/interactions tests/test_litellm/passthrough tests/test_litellm/router_strategy tests/test_litellm/router_utils tests/unit/types --tb=short -vv -n 4 --durations=20
 
 test-unit-root: install-test-deps
-	$(UV_RUN) pytest tests/test_litellm/test_*.py --tb=short -vv -n 4 --durations=20
+	$(UV_RUN) pytest tests/unit/test_*.py tests/test_litellm/test_*.py --tb=short -vv -n 4 --durations=20
 
-# Proxy unit tests (tests/proxy_unit_tests split alphabetically)
+# Proxy unit tests (tests/unit/proxy split alphabetically)
 test-proxy-unit-a: install-test-deps
-	$(UV_RUN) pytest tests/proxy_unit_tests/test_[a-o]*.py --tb=short -vv -n 2 --durations=20
+	$(UV_RUN) pytest tests/unit/proxy --ignore-glob='tests/unit/proxy/test_[p-z]*.py' --tb=short -vv -n 2 --durations=20
 
 test-proxy-unit-b: install-test-deps
-	$(UV_RUN) pytest tests/proxy_unit_tests/test_[p-z]*.py --tb=short -vv -n 2 --durations=20
+	$(UV_RUN) pytest tests/unit/proxy/test_[p-z]*.py tests/unit/skills --tb=short -vv -n 2 --durations=20
 
 test-integration: install-test-deps
 	$(UV_RUN) pytest tests/ -k "not test_litellm"

cookbook/litellm_proxy_server/grafana_dashboard/dashboard_all_metrics/grafana_dashboard.json

@@ -1697,6 +1697,63 @@
       "title": "litellm_video_duration_seconds_metric rate",
       "type": "timeseries"
     },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "${DS_PROMETHEUS}"
+      },
+      "description": "Share of the provider's bill LiteLLM captured as spend over the scheduled capture-rate check's window (needs general_settings.spend_capture_rate_check); NaN while no rate is available",
+      "fieldConfig": {
+        "defaults": {
+          "color": {
+            "mode": "palette-classic"
+          },
+          "custom": {
+            "drawStyle": "line",
+            "fillOpacity": 10,
+            "lineWidth": 1,
+            "showPoints": "never",
+            "spanNulls": false
+          },
+          "unit": "percentunit"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 12,
+        "y": 107
+      },
+      "id": 111,
+      "options": {
+        "legend": {
+          "calcs": [],
+          "displayMode": "list",
+          "placement": "bottom",
+          "showLegend": true
+        },
+        "tooltip": {
+          "mode": "multi",
+          "sort": "desc"
+        }
+      },
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "${DS_PROMETHEUS}"
+          },
+          "editorMode": "code",
+          "expr": "max by (api_provider) (litellm_spend_capture_rate)",
+          "legendFormat": "{{api_provider}}",
+          "range": true,
+          "refId": "A"
+        }
+      ],
+      "title": "litellm_spend_capture_rate",
+      "type": "timeseries"
+    },
     {
       "collapsed": false,
       "gridPos": {

cookbook/litellm_proxy_server/grafana_dashboard/dashboard_all_metrics/readme.md

@@ -1,8 +1,8 @@
 # LiteLLM All Prometheus Metrics dashboard
 
-Every `litellm_*` metric family the proxy can expose on `/metrics` (134 families across 95 panels), grouped into rows: proxy traffic, latency, spend and tokens, cache, LLM API deployments, key and team rate limits, budgets, guardrails, MCP, managed files and batches, users and teams, the Redis circuit breaker, the spend log cleanup job, and the `prometheus_system` service callback metrics (per-service latency, request and failure rates, spend update queue sizes). Panel titles are the metric names so you can grep the JSON for the metric you care about
+Every `litellm_*` metric family the proxy can expose on `/metrics` (136 families across 97 panels), grouped into rows: proxy traffic, latency, spend and tokens, cache, LLM API deployments, key and team rate limits, budgets, guardrails, MCP, managed files and batches, users and teams, the Redis circuit breaker, the spend log cleanup job, and the `prometheus_system` service callback metrics (per-service latency, request and failure rates, spend update queue sizes). Panel titles are the metric names so you can grep the JSON for the metric you care about
 
-Import `grafana_dashboard.json` from **Dashboards > New > Import** and pick your Prometheus data source when prompted (the `DS_PROMETHEUS` variable). Counters are plotted as `rate()` over `$__rate_interval`, histograms as p50 / p95 / p99, gauges as the raw value grouped by the most useful label. Every query names the metric exactly as the proxy emits it (counters carry the `_total` suffix the Prometheus client adds), and `tests/test_litellm/integrations/test_prometheus_metric_name_consistency.py` fails if a metric is renamed without updating this dashboard
+Import `grafana_dashboard.json` from **Dashboards > New > Import** and pick your Prometheus data source when prompted (the `DS_PROMETHEUS` variable). Counters are plotted as `rate()` over `$__rate_interval`, histograms as p50 / p95 / p99, gauges as the raw value grouped by the most useful label. Every query names the metric exactly as the proxy emits it (counters carry the `_total` suffix the Prometheus client adds), and `tests/unit/integrations/test_prometheus_metric_name_consistency.py` fails if a metric is renamed without updating this dashboard
 
 The first eleven rows need only `callbacks: ["prometheus"]`. The last three rows and the `litellm_admission_*` panels are emitted by other subsystems and stay empty until those are on: the service callback row needs `service_callback: ["prometheus_system"]` in `litellm_settings`, the circuit breaker row needs a Redis cache, the cleanup row needs spend log retention, and admission control needs its middleware enabled. Within the base rows, many panels only fill in once the matching feature is in use: budgets need keys, teams, users or orgs with `max_budget` set, cache panels need caching on, guardrail and MCP panels need those features configured, deployment health needs the router with more than one deployment or a failure to record, and `litellm_in_flight_requests` needs traffic at scrape time. An empty panel for a feature you do not use is expected
 

db_scripts/backfill_key_total_spend.sql

@@ -0,0 +1,43 @@
+-- One-shot backfill of LiteLLM_VerificationToken.total_spend (lifetime spend)
+-- for keys created before the column was introduced in LiteLLM v1.103.0.
+--
+-- The column was added with DEFAULT 0 and no backfill, so keys that predate
+-- the upgrade report lifetime spend below their current period spend. New
+-- deployments do not need this script: total_spend is updated at request
+-- time from the moment the release is deployed. Run it only if you want
+-- pre-upgrade keys to show their historical lifetime spend. It sets lifetime
+-- spend to at least the current spend on every key, active and archived,
+-- because current period spend is a valid lower bound on lifetime spend.
+-- For keys with no budget reset that is already the exact lifetime value;
+-- for resetting keys it only recovers the current period. It is idempotent:
+-- it only touches rows where total_spend is below spend, so re-running is a
+-- no-op. It touches no spend logs and runs in seconds.
+--
+-- IMPORTANT caveats before running:
+--
+-- 1. Take a backup of the affected tables first:
+--      pg_dump "$DATABASE_URL" -t '"LiteLLM_VerificationToken"' -t '"LiteLLM_DeletedVerificationToken"' > key_total_spend_backup.sql
+--
+-- 2. A key "resets" when its own budget_duration IS NOT NULL, or when its
+--    budget_id links to a LiteLLM_BudgetTable row whose budget_duration IS
+--    NOT NULL (a linked budget resets the key's spend each period too). For
+--    those keys this script only recovers the current period;
+--    db_scripts/backfill_key_total_spend_from_spend_logs.sql is an optional
+--    follow-up that rebuilds the earlier periods from LiteLLM_SpendLogs.
+--
+-- 3. No proxy restart is needed. The proxy picks up the corrected values on
+--    its next read of each key.
+--
+-- Usage:
+--   psql "$DATABASE_URL" -f db_scripts/backfill_key_total_spend.sql
+
+UPDATE "LiteLLM_VerificationToken"
+SET total_spend = spend
+WHERE total_spend < spend;
+
+UPDATE "LiteLLM_DeletedVerificationToken"
+SET total_spend = spend
+WHERE total_spend < spend;
+
+-- Verify: this should return 0.
+--   SELECT count(*) FROM "LiteLLM_VerificationToken" WHERE total_spend < spend;

db_scripts/backfill_key_total_spend_from_spend_logs.sql

@@ -0,0 +1,89 @@
+-- Optional follow-up to db_scripts/backfill_key_total_spend.sql. Run that
+-- script first; this one rebuilds earlier budget periods for the keys it
+-- can only partially fix: keys whose spend resets each period, because their own
+-- budget_duration IS NOT NULL or because their budget_id links to a
+-- LiteLLM_BudgetTable row whose budget_duration IS NOT NULL.
+--
+-- For those keys the "spend" column only covers the current period, so
+-- lifetime spend is reconstructed from LiteLLM_SpendLogs. The join matches
+-- l.api_key against both the stored token and its second sha256
+-- (encode(sha256(convert_to(token, 'UTF8')), 'hex')), because spend logs
+-- written by older paths recorded the re-hashed digest instead of the
+-- token. It is idempotent and never lowers a value: every statement only
+-- touches rows where total_spend is below the rebuilt sum, so re-running is
+-- a no-op, and a key whose log history is shorter than its current period
+-- keeps the value backfill_key_total_spend.sql already gave it.
+--
+-- IMPORTANT caveats before running:
+--
+-- 1. Take a backup of the affected tables first:
+--      pg_dump "$DATABASE_URL" -t '"LiteLLM_VerificationToken"' -t '"LiteLLM_DeletedVerificationToken"' > key_total_spend_backup.sql
+--
+-- 2. It requires spend logs to have been enabled, and coverage is bounded
+--    by maximum_spend_logs_retention_period: spend older than the retention
+--    window is already gone and cannot be recovered.
+--
+-- 3. On a large SpendLogs table the join scan is slow, so run it off peak.
+--
+-- 4. Run it while the proxy is idle (or with traffic paused). The proxy
+--    flushes spend logs in batches, so a request that already raised
+--    total_spend but whose log is still queued is missing from the sum, and
+--    the rebuilt value would be short by that in-flight amount.
+--
+-- 5. A custom token can be deleted and recreated, so the archived table can
+--    hold several lifetimes of one token. The update only rewrites archived
+--    rows that reset, and the log sum covers every lifetime of that token.
+--
+-- 6. No proxy restart is needed. The proxy picks up the corrected values on
+--    its next read of each key.
+--
+-- Usage:
+--   psql "$DATABASE_URL" -f db_scripts/backfill_key_total_spend_from_spend_logs.sql
+
+-- Active keys whose spend resets (own budget_duration, or a linked
+-- LiteLLM_BudgetTable row with one). Rebuild from LiteLLM_SpendLogs,
+-- matching api_key against the stored token and its second sha256 digest.
+UPDATE "LiteLLM_VerificationToken" k
+SET total_spend = s.sum_spend
+FROM (
+    SELECT k2.token, SUM(l.spend) AS sum_spend
+    FROM "LiteLLM_VerificationToken" k2
+    JOIN "LiteLLM_SpendLogs" l
+      ON l.api_key IN (k2.token, encode(sha256(convert_to(k2.token, 'UTF8')), 'hex'))
+    WHERE k2.budget_duration IS NOT NULL
+       OR k2.budget_id IN (
+           SELECT budget_id FROM "LiteLLM_BudgetTable" WHERE budget_duration IS NOT NULL
+       )
+    GROUP BY k2.token
+) s
+WHERE k.token = s.token
+  AND k.total_spend < s.sum_spend;
+
+-- Archived tokens are not unique, so collapse them to one row per token
+-- before joining spend logs; the update then hits every resetting archived
+-- row.
+UPDATE "LiteLLM_DeletedVerificationToken" k
+SET total_spend = s.sum_spend
+FROM (
+    SELECT k2.token, SUM(l.spend) AS sum_spend
+    FROM (
+        SELECT DISTINCT token
+        FROM "LiteLLM_DeletedVerificationToken"
+        WHERE budget_duration IS NOT NULL
+           OR budget_id IN (
+               SELECT budget_id FROM "LiteLLM_BudgetTable" WHERE budget_duration IS NOT NULL
+           )
+    ) k2
+    JOIN "LiteLLM_SpendLogs" l
+      ON l.api_key IN (k2.token, encode(sha256(convert_to(k2.token, 'UTF8')), 'hex'))
+    GROUP BY k2.token
+) s
+WHERE k.token = s.token
+  AND k.total_spend < s.sum_spend
+  AND (k.budget_duration IS NOT NULL
+       OR k.budget_id IN (
+           SELECT budget_id FROM "LiteLLM_BudgetTable" WHERE budget_duration IS NOT NULL
+       ));
+
+-- Verify: this should return 0.
+--   SELECT count(*) FROM "LiteLLM_VerificationToken" WHERE total_spend < spend;

enterprise/litellm_enterprise/proxy/common_utils/check_batch_cost.py

@@ -11,6 +11,7 @@ from typing import TYPE_CHECKING, Final, List, Literal, Optional, Protocol, Tupl
 from litellm._logging import verbose_proxy_logger
 from litellm._uuid import uuid
 from litellm.constants import (
+    CLI_SESSION_KEY_PREFIX,
     MANAGED_OBJECT_STALENESS_CUTOFF_DAYS,
     MAX_OBJECTS_PER_POLL_CYCLE,
 )
@@ -147,10 +148,12 @@ class CheckBatchCost:
             verbose_proxy_logger.error(f"CheckBatchCost: could not look up user {user_id} for batch {batch_id}: {e}")
             return {}
 
-    async def _get_key_alias(self, batch_id: str, api_key: str | None) -> str | None:
+    async def _get_key_alias(self, batch_id: str, api_key: str | None, created_by: str | None) -> str | None:
         """Resolve the creating virtual key's alias from its hashed token."""
         if not api_key:
             return None
+        if created_by and api_key == f"{CLI_SESSION_KEY_PREFIX}-{created_by}":
+            return api_key
         try:
             key_row: prisma_models.LiteLLM_VerificationToken | None = await _token_table(
                 self.prisma_client
@@ -231,7 +234,7 @@ class CheckBatchCost:
             **(await self._get_user_info(batch_id, job.created_by)),
         }
 
-        key_alias = await self._get_key_alias(batch_id, api_key)
+        key_alias = await self._get_key_alias(batch_id, api_key, job.created_by)
         if key_alias is not None:
             metadata["user_api_key_alias"] = key_alias
         team_alias = await self._get_team_alias(team_id)

enterprise/litellm_enterprise/proxy/hooks/managed_files.py

@@ -50,6 +50,7 @@ from litellm.proxy._types import (
     ProxyException,
     UserAPIKeyAuth,
 )
+from litellm.proxy.litellm_pre_call_utils import LiteLLMProxyRequestSetup
 from litellm.proxy.openai_files_endpoints.common_utils import (
     BATCH_CREATE_HIDDEN_PARAM,
     FILE_LIST_CONTINUATION_CHUNK_SIZE,
@@ -359,7 +360,7 @@ class _PROXY_LiteLLMManagedFiles(CustomLogger, BaseFileEndpoints):
 
         from prisma import Json
 
-        api_key = user_api_key_dict.api_key or None
+        api_key = LiteLLMProxyRequestSetup.get_logged_api_key(user_api_key_dict) or None
         attribution_columns = (
             {
                 **({"api_key": api_key} if api_key is not None else {}),

enterprise/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "litellm-enterprise"
-version = "0.1.70"
+version = "0.1.71"
 description = "Package for LiteLLM Enterprise features"
 readme = "README.md"
 requires-python = ">=3.9"
@@ -26,7 +26,7 @@ required-version = ">=0.10.9"
 module-root = ""
 
 [tool.commitizen]
-version = "0.1.70"
+version = "0.1.71"
 version_files = [
     "pyproject.toml:^version",
     "../pyproject.toml:litellm-enterprise==",

litellm-proxy-extras/litellm_proxy_extras/migrations/20260923000000_add_agent_kill_switch/migration.sql

@@ -0,0 +1,2 @@
+-- AlterTable
+ALTER TABLE "LiteLLM_AgentsTable" ADD COLUMN IF NOT EXISTS "kill_switch" JSONB;

litellm-proxy-extras/litellm_proxy_extras/schema.prisma

@@ -72,6 +72,7 @@ model LiteLLM_AgentsTable {
   agent_card_params Json
   static_headers       Json?          @default("{}")
   extra_headers        String[]       @default([])
+  kill_switch          Json?
   agent_access_groups String[]       @default([])
   access_group_ids     String[]       @default([])
   object_permission_id String?

litellm-proxy-extras/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "litellm-proxy-extras"
-version = "0.4.101"
+version = "0.4.102"
 description = "Additional files for the LiteLLM Proxy. Reduces the size of the main litellm package."
 readme = "README.md"
 requires-python = ">=3.9"
@@ -26,7 +26,7 @@ required-version = ">=0.10.9"
 module-root = ""
 
 [tool.commitizen]
-version = "0.4.101"
+version = "0.4.102"
 version_files = [
     "pyproject.toml:^version",
     "../pyproject.toml:litellm-proxy-extras==",

litellm-rust/AGENTS.md

@@ -0,0 +1,18 @@
+# Rust workspace rules
+
+## Test placement
+
+- Never create a `tests.rs` (or `test.rs`) file under `src/`, and never `#[path = "tests.rs"] mod tests;`
+- A test that reaches private items lives inline, in a `#[cfg(test)] mod tests { ... }` at the bottom of the file that owns those items
+- A test that only uses the crate's public API lives in `crates/<crate>/tests/<subject>.rs`, next to `src/`
+- Split a mixed test file along that line instead of widening visibility to move it
+- A test for another crate's item belongs in that crate, not in a downstream one
+- Never set `autotests = false` or hand-list `[[test]]` targets; every file directly under `tests/` is discovered by cargo, and a shared helper goes in `tests/<name>/mod.rs` or `tests/<subject>/support.rs` so it is not picked up as a test crate of its own
+
+## Error definitions
+
+- A crate's errors live in `src/error.rs`, defined with `thiserror`, and re-exported from `lib.rs`
+- Default to one top-level `Error` enum per crate, with one variant per failure mode and a `#[error(...)]` message on each
+- Wrap a lower-level error as a variant with `#[from]` or `#[source]` instead of flattening it to a string
+- Exception: split into separate types when different functions fail in disjoint ways, especially when different callers see them. A shared enum would force every caller to match variants its function can never return
+- Name a split type after what went wrong (a unit struct is fine for a single failure mode), not after the function that returns it

litellm-rust/Cargo.lock

@@ -73,6 +73,15 @@ version = "1.0.104"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "330a5ed07fa54e4702c9d6c4174f74427fc0ef6e214bbd677ae50a5099946470"
 
+[[package]]
+name = "arbitrary"
+version = "1.4.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c3d036a3c4ab069c7b410a2ce876bd74808d2d0888a82667669f8e783a898bf1"
+dependencies = [
+ "derive_arbitrary",
+]
+
 [[package]]
 name = "arc-swap"
 version = "1.9.2"
@@ -88,6 +97,45 @@ version = "1.2.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "03918c3dbd7701a85c6b9887732e2921175f26c350b4563841d0958c21d57e6d"
 
+[[package]]
+name = "asn1-rs"
+version = "0.7.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b7f43a50ac4fdca5df8e885c21b835997f0a1cdee65494a6847694a98652d9d8"
+dependencies = [
+ "asn1-rs-derive",
+ "asn1-rs-impl",
+ "displaydoc",
+ "nom",
+ "num-traits",
+ "rusticata-macros",
+ "thiserror 2.0.19",
+ "time",
+]
+
+[[package]]
+name = "asn1-rs-derive"
+version = "0.6.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "3109e49b1e4909e9db6515a30c633684d68cdeaa252f215214cb4fa1a5bfee2c"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 2.0.119",
+ "synstructure 0.13.2",
+]
+
+[[package]]
+name = "asn1-rs-impl"
+version = "0.2.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7b18050c2cd6fe86c3a76584ef5e0baf286d038cda203eb6223df2cc413565f7"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 2.0.119",
+]
+
 [[package]]
 name = "assert-json-diff"
 version = "2.0.2"
@@ -810,7 +858,7 @@ version = "0.8.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "08807e080ed7f9d5433fa9b275196cfc35414f66a0c79d864dc51a0d825231a3"
 dependencies = [
- "bit-vec",
+ "bit-vec 0.8.0",
 ]
 
 [[package]]
@@ -819,6 +867,15 @@ version = "0.8.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "5e764a1d40d510daf35e07be9eb06e75770908c27d411ee6c92109c9840eaaf7"
 
+[[package]]
+name = "bit-vec"
+version = "0.9.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b71798fca2c1fe1086445a7258a4bc81e6e49dcd24c8d0dd9a1e57395b603f51"
+dependencies = [
+ "serde",
+]
+
 [[package]]
 name = "bitflags"
 version = "1.3.2"
@@ -849,6 +906,12 @@ dependencies = [
  "hybrid-array",
 ]
 
+[[package]]
+name = "borrow-or-share"
+version = "0.2.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "dc0b364ead1874514c8c2855ab558056ebfeb775653e7ae45ff72f28f8f3166c"
+
 [[package]]
 name = "bstr"
 version = "1.13.1"
@@ -866,6 +929,12 @@ version = "3.20.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "72f5acc6cb2ba439de613abc23857ec3d78374d8ed5ac84e9d11336e87da8649"
 
+[[package]]
+name = "bytecount"
+version = "0.6.9"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "175812e0be2bccb6abe50bb8d566126198344f707e304f45c648fd8f2cc0365e"
+
 [[package]]
 name = "byteorder"
 version = "1.5.0"
@@ -1120,6 +1189,15 @@ version = "0.4.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "338089f42c427b86394a5ee60ff321da23a5c89c9d89514c829687b26359fcff"
 
+[[package]]
+name = "crc32c"
+version = "0.6.8"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "3a47af21622d091a8f0fb295b88bc886ac74efcc613efc19f5d0b21de5c89e47"
+dependencies = [
+ "rustc_version",
+]
+
 [[package]]
 name = "crc32fast"
 version = "1.5.1"
@@ -1378,6 +1456,20 @@ dependencies = [
  "thiserror 2.0.19",
 ]
 
+[[package]]
+name = "der-parser"
+version = "10.0.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "07da5016415d5a3c4dd39b11ed26f915f52fc4e0dc197d87908bc916e51bc1a6"
+dependencies = [
+ "asn1-rs",
+ "displaydoc",
+ "nom",
+ "num-bigint 0.4.8",
+ "num-traits",
+ "rusticata-macros",
+]
+
 [[package]]
 name = "deranged"
 version = "0.5.8"
@@ -1387,6 +1479,17 @@ dependencies = [
  "serde_core",
 ]
 
+[[package]]
+name = "derive_arbitrary"
+version = "1.4.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1e567bd82dcff979e4b03460c307b3cdc9e96fde3d73bed1496d2bc75d9dd62a"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 2.0.119",
+]
+
 [[package]]
 name = "derive_builder"
 version = "0.20.2"
@@ -1469,6 +1572,15 @@ version = "1.16.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "91622ff5e7162018101f2fea40d6ebf4a78bbe5a49736a2020649edf9693679e"
 
+[[package]]
+name = "email_address"
+version = "0.2.9"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e079f19b08ca6239f47f8ba8509c11cf3ea30095831f7fed61441475edd8c449"
+dependencies = [
+ "serde",
+]
+
 [[package]]
 name = "equivalent"
 version = "1.0.2"
@@ -1551,6 +1663,16 @@ version = "2.5.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "da7c62ceae207dd37ea5b845da6a0696c799f85e97da1ab5b7910be3c1c80223"
 
+[[package]]
+name = "filetime"
+version = "0.2.29"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "5c287a33c7f0a620c38e641e7f60827713987b3c0f26e8ddc9462cc69cf75759"
+dependencies = [
+ "cfg-if",
+ "libc",
+]
+
 [[package]]
 name = "find-msvc-tools"
 version = "0.1.9"
@@ -1568,6 +1690,17 @@ dependencies = [
  "zlib-rs",
 ]
 
+[[package]]
+name = "fluent-uri"
+version = "0.4.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "bc74ac4d8359ae70623506d512209619e5cf8f347124910440dbc221714b328e"
+dependencies = [
+ "borrow-or-share",
+ "ref-cast",
+ "serde",
+]
+
 [[package]]
 name = "fnv"
 version = "1.0.7"
@@ -1589,6 +1722,16 @@ dependencies = [
  "percent-encoding",
 ]
 
+[[package]]
+name = "fraction"
+version = "0.17.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e246562084dde8ebbcc943b261c406ce4f68e5032ec28029a251a47d6a295500"
+dependencies = [
+ "num",
+ "num-bigint 0.4.8",
+]
+
 [[package]]
 name = "fs_extra"
 version = "1.3.0"
@@ -1746,9 +1889,11 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "899def5c37c4fd7b2664648c28120ecec138e4d395b459e5ca34f9cce2dd77fd"
 dependencies = [
  "cfg-if",
+ "js-sys",
  "libc",
  "r-efi 5.3.0",
  "wasip2",
+ "wasm-bindgen",
 ]
 
 [[package]]
@@ -2590,18 +2735,56 @@ dependencies = [
 ]
 
 [[package]]
-name = "jsonwebtoken"
-version = "11.1.0"
+name = "jsonschema"
+version = "0.55.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "e75fe14a82d81e5f5af639997db37d8b96045938a7ac6ab18cdbe1c7467e05e1"
+checksum = "b68339c3d874e48151d74ffe256d93a58cffa240983cb0967d3cbaea083a44fe"
 dependencies = [
- "base64 0.22.1",
- "getrandom 0.2.17",
- "js-sys",
+ "ahash",
+ "bytecount",
+ "data-encoding",
+ "email_address",
+ "fancy-regex 0.19.2",
+ "fraction",
+ "getrandom 0.3.4",
+ "itoa",
+ "jsonschema-regex",
+ "jsonschema-value",
+ "num-cmp",
+ "num-traits",
+ "percent-encoding",
+ "referencing",
+ "regex",
  "serde",
  "serde_json",
- "signature",
- "zeroize",
+ "strum",
+ "unicode-general-category",
+ "uuid-simd",
+]
+
+[[package]]
+name = "jsonschema-regex"
+version = "0.55.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "6307b5b51216ec9b941b52244c74043fa0b1d6b657b56199f57cb1416d3641c5"
+dependencies = [
+ "regex-syntax",
+]
+
+[[package]]
+name = "jsonschema-value"
+version = "0.55.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0230ac05e09c6111e96c147b75c390579f5cbd45654b980c68ac60fe17b3f129"
+dependencies = [
+ "ahash",
+ "bytecount",
+ "fraction",
+ "getrandom 0.3.4",
+ "num-cmp",
+ "num-traits",
+ "serde_json",
+ "zmij",
 ]
 
 [[package]]
@@ -2939,6 +3122,7 @@ dependencies = [
  "tokio-tungstenite",
  "url",
  "veil",
+ "wiremock",
 ]
 
 [[package]]
@@ -2957,6 +3141,15 @@ dependencies = [
  "url",
 ]
 
+[[package]]
+name = "litellm-coroutine"
+version = "0.1.0"
+dependencies = [
+ "rstest",
+ "thiserror 2.0.19",
+ "tokio",
+]
+
 [[package]]
 name = "litellm-cost"
 version = "0.1.0"
@@ -2973,10 +3166,11 @@ dependencies = [
  "aws-smithy-types",
  "bytes",
  "futures-util",
+ "proptest",
  "rstest",
- "sse-stream",
  "thiserror 2.0.19",
  "tokio",
+ "tokio-util",
 ]
 
 [[package]]
@@ -2984,6 +3178,7 @@ name = "litellm-host"
 version = "0.1.0"
 dependencies = [
  "litellm-auth",
+ "litellm-coroutine",
  "rstest",
  "serde_json",
  "tokio",
@@ -2993,6 +3188,7 @@ dependencies = [
 name = "litellm-host-python"
 version = "0.1.0"
 dependencies = [
+ "bytes",
  "futures-util",
  "litellm-host",
  "pyo3",
@@ -3059,14 +3255,14 @@ dependencies = [
 name = "litellm-model-catalog"
 version = "0.1.0"
 dependencies = [
- "criterion",
  "indexmap 2.14.0",
- "litellm-model-catalog",
+ "jsonschema",
  "rstest",
  "schemars 1.2.2",
  "serde",
  "serde_json",
  "thiserror 2.0.19",
+ "time",
 ]
 
 [[package]]
@@ -3119,6 +3315,7 @@ dependencies = [
  "tokio",
  "tokio-tungstenite",
  "url",
+ "veil",
  "wiremock",
 ]
 
@@ -3143,10 +3340,11 @@ version = "0.1.0"
 dependencies = [
  "aws-sdk-kms",
  "base64 0.22.1",
+ "futures-util",
  "google-cloud-auth",
  "google-cloud-kms-v1",
- "jsonwebtoken",
  "litellm-core-utils",
+ "litellm-python-compat",
  "litellm-secrets-aws",
  "litellm-secrets-azure",
  "litellm-secrets-cyberark",
@@ -3179,6 +3377,7 @@ dependencies = [
  "litellm-tracing",
  "rstest",
  "serde_json",
+ "tempfile",
  "thiserror 2.0.19",
  "tokio",
  "veil",
@@ -3215,10 +3414,12 @@ dependencies = [
  "litellm-tracing",
  "moka",
  "percent-encoding",
+ "rcgen",
  "reqwest 0.12.28",
  "rstest",
  "serde",
  "serde_json",
+ "tempfile",
  "thiserror 2.0.19",
  "tokio",
  "veil",
@@ -3230,6 +3431,7 @@ name = "litellm-secrets-google"
 version = "0.1.0"
 dependencies = [
  "base64 0.22.1",
+ "crc32c",
  "google-cloud-auth",
  "google-cloud-gax",
  "google-cloud-kms-v1",
@@ -3255,7 +3457,6 @@ version = "0.1.0"
 dependencies = [
  "litellm-core-utils",
  "litellm-secrets-types",
- "moka",
  "rstest",
  "rustify",
  "rustify_derive",
@@ -3274,6 +3475,7 @@ name = "litellm-secrets-types"
 version = "0.1.0"
 dependencies = [
  "litellm-auth-types",
+ "moka",
  "rstest",
  "serde",
  "serde_json",
@@ -3282,6 +3484,27 @@ dependencies = [
  "veil",
 ]
 
+[[package]]
+name = "litellm-testkit"
+version = "0.1.0"
+dependencies = [
+ "flate2",
+ "futures-util",
+ "reqwest 0.12.28",
+ "rstest",
+ "semver",
+ "serde",
+ "serde_json",
+ "sha2 0.10.9",
+ "tar",
+ "target-lexicon",
+ "tempfile",
+ "thiserror 2.0.19",
+ "tokio",
+ "toml",
+ "zip",
+]
+
 [[package]]
 name = "litellm-token-counter"
 version = "0.1.0"
@@ -3352,6 +3575,7 @@ dependencies = [
 name = "litellm-types"
 version = "0.1.0"
 dependencies = [
+ "rstest",
  "serde",
  "serde_json",
 ]
@@ -3430,6 +3654,12 @@ version = "2.8.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "cf8baf1c55e62ffcace7a9f06f4bd9cd3f0c4beb022d3b367256b91b87513d98"
 
+[[package]]
+name = "micromap"
+version = "0.3.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c2a86d3146ed3995b5913c414f6664344b9617457320782e64f0bb44afd49d74"
+
 [[package]]
 name = "mime"
 version = "0.3.17"
@@ -3525,6 +3755,20 @@ dependencies = [
  "minimal-lexical",
 ]
 
+[[package]]
+name = "num"
+version = "0.4.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "35bd024e8b2ff75562e5f34e7f4905839deb4b22955ef5e73d2fea1b9813cb23"
+dependencies = [
+ "num-bigint 0.4.8",
+ "num-complex",
+ "num-integer",
+ "num-iter",
+ "num-rational",
+ "num-traits",
+]
+
 [[package]]
 name = "num-bigint"
 version = "0.4.8"
@@ -3545,6 +3789,12 @@ dependencies = [
  "num-traits",
 ]
 
+[[package]]
+name = "num-cmp"
+version = "0.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "63335b2e2c34fae2fb0aa2cecfd9f0832a1e24b3b32ecec612c3426d46dc8aaa"
+
 [[package]]
 name = "num-complex"
 version = "0.4.6"
@@ -3569,6 +3819,27 @@ dependencies = [
  "num-traits",
 ]
 
+[[package]]
+name = "num-iter"
+version = "0.1.46"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c92800bd69a1eac91786bcfe9da64a897eb72911b8dc3095decbd07429e8048b"
+dependencies = [
+ "num-integer",
+ "num-traits",
+]
+
+[[package]]
+name = "num-rational"
+version = "0.4.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f83d14da390562dca69fc84082e73e548e1ad308d24accdedd2720017cb37824"
+dependencies = [
+ "num-bigint 0.4.8",
+ "num-integer",
+ "num-traits",
+]
+
 [[package]]
 name = "num-traits"
 version = "0.2.19"
@@ -3588,6 +3859,15 @@ dependencies = [
  "libc",
 ]
 
+[[package]]
+name = "oid-registry"
+version = "0.8.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "12f40cff3dde1b6087cc5d5f5d4d65712f34016a03ed60e9c08dcc392736b5b7"
+dependencies = [
+ "asn1-rs",
+]
+
 [[package]]
 name = "once_cell"
 version = "1.21.4"
@@ -3721,6 +4001,16 @@ version = "0.2.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "2ee67f1008b1ba2321834326597b8e186293b049a023cdef258527550b9935b4"
 
+[[package]]
+name = "pem"
+version = "4.0.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d354a98a3d1251555de99e8fdd8afda05573c31b82f59063a7b0a29b5527f120"
+dependencies = [
+ "base64 0.23.1",
+ "serde_core",
+]
+
 [[package]]
 name = "percent-encoding"
 version = "2.3.2"
@@ -3899,7 +4189,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "4b45fcc2344c680f5025fe57779faef368840d0bd1f42f216291f0dc4ace4744"
 dependencies = [
  "bit-set",
- "bit-vec",
+ "bit-vec 0.8.0",
  "bitflags 2.13.1",
  "num-traits",
  "rand 0.9.5",
@@ -4288,6 +4578,20 @@ dependencies = [
  "crossbeam-utils",
 ]
 
+[[package]]
+name = "rcgen"
+version = "0.14.10"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8774e05a7d0de114588e6a28fe7e71694b82614ed569d86d8b389dfbc98b8ad8"
+dependencies = [
+ "pem",
+ "ring",
+ "rustls-pki-types",
+ "time",
+ "x509-parser",
+ "yasna",
+]
+
 [[package]]
 name = "redis"
 version = "1.7.0"
@@ -4351,6 +4655,23 @@ dependencies = [
  "syn 3.0.0",
 ]
 
+[[package]]
+name = "referencing"
+version = "0.55.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a196a5b4a8a12f46b6353174df865a05d41a6055aff212ec30877492788618b6"
+dependencies = [
+ "ahash",
+ "fluent-uri",
+ "getrandom 0.3.4",
+ "hashbrown 0.17.1",
+ "itoa",
+ "micromap",
+ "parking_lot",
+ "percent-encoding",
+ "serde_json",
+]
+
 [[package]]
 name = "regex"
 version = "1.13.1"
@@ -4570,6 +4891,15 @@ dependencies = [
  "semver",
 ]
 
+[[package]]
+name = "rusticata-macros"
+version = "4.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "faf0c4a6ece9950b9abdb62b1cfcf2a68b3b67a10ba445b3bb85be2a293d0632"
+dependencies = [
+ "nom",
+]
+
 [[package]]
 name = "rustify"
 version = "0.7.0"
@@ -4928,6 +5258,15 @@ dependencies = [
  "serde_core",
 ]
 
+[[package]]
+name = "serde_spanned"
+version = "1.1.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "6662b5879511e06e8999a8a235d848113e942c9124f211511b16466ee2995f26"
+dependencies = [
+ "serde_core",
+]
+
 [[package]]
 name = "serde_urlencoded"
 version = "0.7.1"
@@ -5046,15 +5385,6 @@ dependencies = [
  "libc",
 ]
 
-[[package]]
-name = "signature"
-version = "2.2.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "77549399552de45a898a580c1b41d445bf730df867cc44e6c0233bbc4b8329de"
-dependencies = [
- "rand_core 0.6.4",
-]
-
 [[package]]
 name = "simd-adler32"
 version = "0.3.10"
@@ -5139,19 +5469,6 @@ dependencies = [
  "wasm-bindgen",
 ]
 
-[[package]]
-name = "sse-stream"
-version = "0.2.6"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "c25ac7aff0abd1dbc474536e40416e1102c7dd9bfba0b9861c6d357f835dcfb4"
-dependencies = [
- "bytes",
- "futures-util",
- "http-body 1.1.0",
- "http-body-util",
- "pin-project-lite",
-]
-
 [[package]]
 name = "stable_deref_trait"
 version = "1.2.1"
@@ -5268,6 +5585,17 @@ version = "0.2.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "7b2093cf4c8eb1e67749a6762251bc9cd836b6fc171623bd0a9d324d37af2417"
 
+[[package]]
+name = "tar"
+version = "0.4.46"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "3f6221d9a6003c78398e3b239969f352578258df48c8eb051caadae0015bc840"
+dependencies = [
+ "filetime",
+ "libc",
+ "xattr",
+]
+
 [[package]]
 name = "target-lexicon"
 version = "0.13.5"
@@ -5537,6 +5865,30 @@ dependencies = [
  "tokio",
 ]
 
+[[package]]
+name = "toml"
+version = "0.9.12+spec-1.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "cf92845e79fc2e2def6a5d828f0801e29a2f8acc037becc5ab08595c7d5e9863"
+dependencies = [
+ "indexmap 2.14.0",
+ "serde_core",
+ "serde_spanned",
+ "toml_datetime 0.7.5+spec-1.1.0",
+ "toml_parser",
+ "toml_writer",
+ "winnow 0.7.15",
+]
+
+[[package]]
+name = "toml_datetime"
+version = "0.7.5+spec-1.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "92e1cfed4a3038bc5a127e35a2d360f145e1f4b971b551a2ba5fd7aedf7e1347"
+dependencies = [
+ "serde_core",
+]
+
 [[package]]
 name = "toml_datetime"
 version = "1.1.1+spec-1.1.0"
@@ -5553,9 +5905,9 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "6975367e4d2ef766d86af01ffad14b622fecc8d4357a998fbc4deb6e9bacaf9b"
 dependencies = [
  "indexmap 2.14.0",
- "toml_datetime",
+ "toml_datetime 1.1.1+spec-1.1.0",
  "toml_parser",
- "winnow",
+ "winnow 1.0.4",
 ]
 
 [[package]]
@@ -5564,9 +5916,15 @@ version = "1.1.3+spec-1.1.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "1d38ac1cf9b95face32296c0a3ede1fdc270627c9d9c02a7274dd6d960dc4d56"
 dependencies = [
- "winnow",
+ "winnow 1.0.4",
 ]
 
+[[package]]
+name = "toml_writer"
+version = "1.1.2+spec-1.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7d56353a2a665ad0f41a421187180aab746c8c325620617ad883a99a1cbe66d2"
+
 [[package]]
 name = "tonic"
 version = "0.14.6"
@@ -5834,6 +6192,12 @@ version = "2.9.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "dbc4bc3a9f746d862c45cb89d705aa10f187bb96c76001afab07a0d35ce60142"
 
+[[package]]
+name = "unicode-general-category"
+version = "1.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0b993bddc193ae5bd0d623b49ec06ac3e9312875fdae725a975c51db1cc1677f"
+
 [[package]]
 name = "unicode-ident"
 version = "1.0.24"
@@ -5914,6 +6278,16 @@ dependencies = [
  "wasm-bindgen",
 ]
 
+[[package]]
+name = "uuid-simd"
+version = "0.8.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "23b082222b4f6619906941c17eb2297fff4c2fb96cb60164170522942a200bd8"
+dependencies = [
+ "outref",
+ "vsimd",
+]
+
 [[package]]
 name = "valuable"
 version = "0.1.1"
@@ -6312,6 +6686,12 @@ version = "0.52.6"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "589f6da84c646204747d1270a2a5661ea66ed1cced2631d546fdfb155959f9ec"
 
+[[package]]
+name = "winnow"
+version = "0.7.15"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "df79d97927682d2fd8adb29682d1140b343be4ac0f08fd68b7765d9c059d3945"
+
 [[package]]
 name = "winnow"
 version = "1.0.4"
@@ -6356,6 +6736,34 @@ version = "0.6.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "1ffae5123b2d3fc086436f8834ae3ab053a283cfac8fe0a0b8eaae044768a4c4"
 
+[[package]]
+name = "x509-parser"
+version = "0.18.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d43b0f71ce057da06bc0851b23ee24f3f86190b07203dd8f567d0b706a185202"
+dependencies = [
+ "asn1-rs",
+ "data-encoding",
+ "der-parser",
+ "lazy_static",
+ "nom",
+ "oid-registry",
+ "ring",
+ "rusticata-macros",
+ "thiserror 2.0.19",
+ "time",
+]
+
+[[package]]
+name = "xattr"
+version = "1.6.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "32e45ad4206f6d2479085147f02bc2ef834ac85886624a23575ae137c8aa8156"
+dependencies = [
+ "libc",
+ "rustix",
+]
+
 [[package]]
 name = "xmlparser"
 version = "0.13.6"
@@ -6368,6 +6776,16 @@ version = "0.8.18"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "aee1b19627c7c60102ab80d3a9cbe18de90bfe03bfa6c3715447681f0e8c8af6"
 
+[[package]]
+name = "yasna"
+version = "0.6.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b5f6765e852b9b4dc8e2a76843e4d64d1cea8e79bcde0b6901aea8e7c7f08282"
+dependencies = [
+ "bit-vec 0.9.1",
+ "time",
+]
+
 [[package]]
 name = "yoke"
 version = "0.8.3"
@@ -6437,20 +6855,6 @@ name = "zeroize"
 version = "1.9.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "e13c156562582aa81c60cb29407084cdb54c4164760106ab78e6c5b0858cf64e"
-dependencies = [
- "zeroize_derive",
-]
-
-[[package]]
-name = "zeroize_derive"
-version = "1.5.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "3c50655cbb0fe3fc43170059e702f1ce5e19b84cec58dc87b037a09935c2f328"
-dependencies = [
- "proc-macro2",
- "quote",
- "syn 2.0.119",
-]
 
 [[package]]
 name = "zerotrie"
@@ -6485,6 +6889,23 @@ dependencies = [
  "syn 2.0.119",
 ]
 
+[[package]]
+name = "zip"
+version = "2.4.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "fabe6324e908f85a1c52063ce7aa26b68dcb7eb6dbc83a2d148403c9bc3eba50"
+dependencies = [
+ "arbitrary",
+ "crc32fast",
+ "crossbeam-utils",
+ "displaydoc",
+ "flate2",
+ "indexmap 2.14.0",
+ "memchr",
+ "thiserror 2.0.19",
+ "zopfli",
+]
+
 [[package]]
 name = "zlib-rs"
 version = "0.6.7"
@@ -6496,3 +6917,15 @@ name = "zmij"
 version = "1.0.23"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "29666d0abbfad1e3dc4dcf6144730dd3a3ab225bbbdac83319345b1b44ccfc1b"
+
+[[package]]
+name = "zopfli"
+version = "0.8.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f05cd8797d63865425ff89b5c4a48804f35ba0ce8d125800027ad6017d2b5249"
+dependencies = [
+ "bumpalo",
+ "crc32fast",
+ "log",
+ "simd-adler32",
+]

litellm-rust/Cargo.toml

@@ -12,6 +12,7 @@ repository = "https://github.com/BerriAI/litellm"
 litellm-tracing = { path = "crates/tracing" }
 tracing = "0.1"
 litellm-core = { path = "crates/core" }
+litellm-coroutine = { path = "crates/coroutine" }
 litellm-host = { path = "crates/host" }
 litellm-callbacks-legacy-python = { path = "crates/callbacks-legacy-python" }
 litellm-framing = { path = "crates/framer" }
@@ -80,6 +81,12 @@ tokio = { version = "1", features = ["rt-multi-thread", "macros", "time", "net"]
 tokio-tungstenite = { version = "0.24", default-features = false, features = ["connect", "rustls-tls-native-roots"] }
 futures-util = { version = "0.3", default-features = false, features = ["sink", "std"] }
 base64 = "0.22"
+flate2 = "1"
+semver = "1"
+tar = "0.4"
+target-lexicon = "0.13.5"
+tempfile = "3"
+zip = { version = "2", default-features = false, features = ["deflate"] }
 moka = { version = "0.12.16", features = ["future"] }
 strum = { version = "0.28.0", features = ["derive"] }
 url = "2.5.8"

litellm-rust/crates/callbacks-legacy-python/Cargo.toml

@@ -4,7 +4,6 @@ version = "0.1.0"
 edition.workspace = true
 license.workspace = true
 repository.workspace = true
-autotests = false
 
 [dependencies]
 litellm-host.workspace = true

litellm-rust/crates/callbacks-legacy-python/src/call.rs

@@ -3,8 +3,8 @@
 //! lifetime. No other callback host has that obligation, which is why nothing outside
 //! this crate holds them.
 
-use litellm_host::{machine::Machine, route::Route};
-use litellm_host_python::{RouteHost, lookup, run_call};
+use litellm_host::{machine::Machine, protocol::Protocol};
+use litellm_host_python::{ProtocolHost, lookup, run_call};
 use pyo3::{
     gc::{PyTraverseError, PyVisit},
     prelude::*,
@@ -63,25 +63,25 @@ impl PublicCall {
     }
 }
 
-/// Runs one native call under the legacy `Logging` contract: the route host projects from
+/// Runs one native call under the legacy `Logging` contract: the protocol host projects from
 /// the keyword view the contract prepares, and the contract observes the call.
 pub fn run_legacy_call<H, M>(
     py: Python<'_>,
     surface: LegacySurface,
     call: PublicCall,
     machine: M,
-    route: H,
+    host: H,
     asynchronous: bool,
 ) -> PyResult<Py<PyAny>>
 where
-    H: RouteHost + 'static,
-    M: Machine<Route = H::Route, Complete = <H::Route as Route>::Response> + 'static,
+    H: ProtocolHost + 'static,
+    M: Machine<Protocol = H::Protocol, Complete = <H::Protocol as Protocol>::Response> + 'static,
 {
     let arguments = call.kwargs.clone_ref(py);
     run_call(
         py,
         machine,
-        route,
+        host,
         Box::new(LegacyLogging::new(py, surface, call, asynchronous)),
         arguments,
         asynchronous,

litellm-rust/crates/callbacks-legacy-python/src/deferred.rs

@@ -63,5 +63,151 @@ impl PendingLogging {
 }
 
 #[cfg(test)]
-#[path = "../tests/deferred.rs"]
-mod tests;
+mod tests {
+    use std::ffi::CStr;
+
+    use pyo3::prelude::*;
+    use pyo3::types::PyDict;
+    use rstest::rstest;
+
+    use super::{PendingLogging, PendingSuccess};
+    use crate::PythonLogger;
+    use crate::test_support::{local, namespace, run};
+
+    /// A deferred success for the namespace's `logger` and `response`, bound as `pending`.
+    fn defer<'py>(py: Python<'py>, script: &CStr) -> Bound<'py, PyDict> {
+        let locals = namespace(py, c"response = object()");
+        run(py, &locals, script);
+        let pending = Py::new(
+            py,
+            PendingLogging {
+                pending: Some(PendingSuccess {
+                    logger: PythonLogger::new(local(&locals, "logger").unbind()),
+                    response: Some(local(&locals, "response").unbind()),
+                    start: py.None(),
+                    end: Some(py.None()),
+                }),
+            },
+        )
+        .unwrap();
+        locals.set_item("pending", pending).unwrap();
+        locals
+    }
+
+    #[test]
+    fn release_enqueues_the_success_once_in_the_releasing_context() {
+        Python::initialize();
+        Python::attach(|py| {
+            let locals = defer(
+                py,
+                c"
+from contextvars import ContextVar
+
+marker = ContextVar('marker', default='unset')
+observed = []
+
+def on_enqueue(coroutine):
+    observed.append(marker.get())
+    pending.release(True)
+
+logger.on_enqueue = on_enqueue
+",
+            );
+            run(
+                py,
+                &locals,
+                c"
+marker.set('release')
+pending.release(True)
+pending.release(True)
+assert observed == ['release'], observed
+assert logger.names() == ['async_success_handler', 'enqueued'], logger.calls
+assert logger.calls[0][1] is response
+",
+            );
+        });
+    }
+
+    #[test]
+    fn a_blocked_release_drops_the_success_for_good() {
+        Python::initialize();
+        Python::attach(|py| {
+            let locals = defer(py, c"");
+            run(
+                py,
+                &locals,
+                c"
+pending.release(False)
+pending.release(True)
+assert logger.calls == [], logger.calls
+",
+            );
+        });
+    }
+
+    #[rstest]
+    #[case::ordinary_error(c"RuntimeError('queue full')", false)]
+    #[case::cancellation(c"asyncio.CancelledError()", true)]
+    fn a_failed_enqueue_closes_the_coroutine_and_is_never_replayed(
+        #[case] failure: &CStr,
+        #[case] propagates: bool,
+    ) {
+        Python::initialize();
+        Python::attach(|py| {
+            let locals = defer(
+                py,
+                c"
+import asyncio
+
+def on_enqueue(coroutine):
+    raise failure
+
+logger.on_enqueue = on_enqueue
+",
+            );
+            locals
+                .set_item("failure", py.eval(failure, None, Some(&locals)).unwrap())
+                .unwrap();
+            let released = local(&locals, "pending").call_method1("release", (true,));
+            match released {
+                Ok(_) => assert!(!propagates),
+                Err(error) => {
+                    assert!(propagates);
+                    assert!(error.value(py).is(local(&locals, "failure")));
+                }
+            }
+            locals.set_item("propagates", propagates).unwrap();
+            run(
+                py,
+                &locals,
+                c"
+pending.release(True)
+assert logger.names() == ['async_success_handler', 'enqueued', 'closed'], logger.calls
+assert unraisable_from(logger) == ([] if propagates else [failure])
+",
+            );
+        });
+    }
+
+    #[test]
+    fn an_unreleased_success_does_not_keep_its_logger_alive() {
+        Python::initialize();
+        Python::attach(|py| {
+            let locals = defer(py, c"");
+            run(
+                py,
+                &locals,
+                c"
+import gc
+import weakref
+
+logger.pending = pending
+reference = weakref.ref(logger)
+del logger, pending
+gc.collect()
+assert reference() is None
+",
+            );
+        });
+    }
+}

litellm-rust/crates/callbacks-legacy-python/src/lib.rs

@@ -16,13 +16,218 @@ mod deferred;
 mod logger;
 mod preparation;
 mod python;
-#[cfg(test)]
-#[path = "../tests/support.rs"]
-mod test_support;
-
 pub(crate) use adapter::LegacyLogging;
 pub use adapter::{LegacySurface, PassThroughStream};
 pub use call::{PublicCall, run_legacy_call};
 pub(crate) use callbacks::{LegacyCallbacks, is_internal_call};
 pub(crate) use logger::{DeploymentHooks, PythonLogger, finalize, setup};
 pub(crate) use preparation::prepare;
+
+#[cfg(test)]
+mod test_support {
+    use std::ffi::CStr;
+
+    use pyo3::prelude::*;
+    use pyo3::types::{PyDict, PyTuple};
+
+    use crate::{LegacyLogging, LegacySurface, PublicCall};
+
+    /// The parameters of every `callbacks_legacy_python` function, as the real module declares them.
+    /// `tests/test_litellm/rust_bridge/test_callbacks_legacy_python.py` pins this file to the Python
+    /// signatures, and [`namespace`] binds every fake call against it.
+    pub(crate) const PYTHON_CONTRACT: &str = include_str!("../python_contract.json");
+
+    /// Stand-ins for `callbacks_legacy_python`, the only Python module the crate calls. Tests
+    /// share one interpreter and run concurrently, so each fake is installed idempotently and
+    /// forwards to the per-test `StubLogger` it is handed (directly, or as `kwargs['logger']`).
+    /// Every fake is bound against the contract first, so a call the real module would reject
+    /// fails here too.
+    const STUBS: &CStr = c"
+import contextvars
+import inspect
+import json
+import sys
+import traceback
+import types
+
+for name in ('litellm', 'litellm.rust_bridge', 'litellm.rust_bridge.callbacks_legacy_python'):
+    sys.modules.setdefault(name, types.ModuleType(name))
+
+legacy = sys.modules['litellm.rust_bridge.callbacks_legacy_python']
+CONTRACT = json.loads(python_contract)
+
+
+def contracted(name, fake):
+    signature = inspect.Signature(
+        [inspect.Parameter(parameter, inspect.Parameter.POSITIONAL_OR_KEYWORD) for parameter in CONTRACT[name]]
+    )
+
+    def checked(*args, **kwargs):
+        signature.bind(*args, **kwargs)
+        return fake(*args, **kwargs)
+
+    return checked
+
+
+if not hasattr(legacy, 'is_internal'):
+    legacy.is_internal = contextvars.ContextVar('is_internal_call', default=False)
+
+FAKES = {
+    'setup': lambda call_type, args, kwargs, start, asynchronous: types.SimpleNamespace(
+        logger=kwargs['logger_factory'](kwargs) if 'logger_factory' in kwargs else kwargs['logger'],
+        kwargs=kwargs,
+    ),
+    'check_limits': lambda arguments: arguments['logger'].check_limits(arguments),
+    'finalize': lambda response, logger, kwargs, start, end: logger.record('finalize', response),
+    'update_logging': lambda logger, kwargs, model, optional_params, litellm_params, provider: logger.update_from_kwargs(
+        kwargs=kwargs,
+        model=model,
+        optional_params=optional_params,
+        litellm_params=litellm_params,
+        custom_llm_provider=provider,
+    ),
+    'pre_call': lambda logger, input, api_key, additional_args: logger.pre_call(input, api_key, additional_args),
+    'post_call': lambda logger, original_response, api_key, additional_args: logger.post_call(
+        original_response, api_key, additional_args
+    ),
+    'defers_async_logging': lambda logger: bool(getattr(logger, '_defer_async_logging', False)),
+    'defer_success': lambda logger, pending: setattr(logger, '_native_pending_logging', pending),
+    'sync_success_for_async_call': lambda logger, response, start, end: logger.handle_sync_success_callbacks_for_async_calls(
+        response, start, end
+    ),
+    'failure_handler': lambda logger, error, start, end, asynchronous: (
+        logger.async_failure_handler if asynchronous else logger.failure_handler
+    )(error, ''.join(traceback.format_exception(error)), start, end),
+    'submit_success': lambda logger, response, start, end: logger.record('submit', (response, start, end)),
+    'async_success_handler': lambda logger, response, start, end: logger.async_success_handler(response, start, end),
+    'enqueue_logging': lambda coroutine: coroutine.enqueue(),
+    'restore_context': lambda logger: logger.record('restore', None),
+    'custom_pricing_fields': lambda: ('ocr_cost_per_page',),
+    'is_internal_call': lambda: legacy.is_internal.get(),
+    'credential_list': lambda: [],
+    'warn_unknown_credential': lambda name, loaded: None,
+    'before_deployment_call': lambda kwargs, call_type: kwargs['logger'].hook('pre', kwargs, call_type),
+    'after_deployment_success': lambda kwargs, response, call_type: kwargs['logger'].hook(
+        'success', response, call_type
+    ),
+    'after_deployment_failure': lambda kwargs, error, call_type: kwargs['logger'].hook('failure', error, call_type),
+    'stream_opened': lambda logger: logger.record('stream_opened', None),
+    'stream_success': lambda logger, request_body, chunks, start, end, first_chunk: logger.record(
+        'stream_success', list(chunks)
+    ),
+    'stream_failure': lambda logger, request_body, chunks, error: logger.record('stream_failure', error),
+}
+assert FAKES.keys() == CONTRACT.keys(), sorted(FAKES.keys() ^ CONTRACT.keys())
+for name, fake in FAKES.items():
+    setattr(legacy, name, contracted(name, fake))
+
+
+unraisable = sys.modules.setdefault(
+    'litellm_test_unraisable', types.ModuleType('litellm_test_unraisable')
+)
+if not hasattr(unraisable, 'events'):
+    unraisable.events = []
+    sys.unraisablehook = lambda event: unraisable.events.append((event.object, event.exc_value))
+
+
+def unraisable_from(owner):
+    return [error for source, error in unraisable.events if source is owner]
+
+
+class StubCoroutine:
+    def __init__(self, logger):
+        self.logger = logger
+
+    def enqueue(self):
+        self.logger.record('enqueued', None)
+        self.logger.on_enqueue(self)
+
+    def close(self):
+        self.logger.record('closed', None)
+
+
+class StubLogger:
+    def __init__(self):
+        self.calls = []
+        self.hooks = {}
+        self.on_enqueue = lambda coroutine: None
+
+    def record(self, name, value):
+        self.calls.append((name, value))
+
+    def names(self):
+        return [name for name, _ in self.calls]
+
+    def hook(self, phase, value, call_type):
+        self.record(phase + '_hook', call_type)
+        return self.hooks.get(phase, lambda value: 'awaitable')(value)
+
+    def check_limits(self, arguments):
+        self.record('check_limits', arguments)
+
+    def failure_handler(self, error, trace, start, end):
+        self.record('failure_handler', error)
+
+    def async_failure_handler(self, error, trace, start, end):
+        self.record('async_failure_handler', error)
+        return 'awaitable'
+
+    def success_handler(self, response, start, end):
+        self.record('success_handler', response)
+
+    def async_success_handler(self, response, start, end):
+        self.record('async_success_handler', response)
+        return StubCoroutine(self)
+
+    def handle_sync_success_callbacks_for_async_calls(self, response, start, end):
+        self.record('sync_success_for_async_call', response)
+
+
+logger = StubLogger()
+";
+
+    /// A namespace with the stubs, `StubLogger` and a fresh `logger`, after `script` ran in it.
+    pub(crate) fn namespace<'py>(py: Python<'py>, script: &CStr) -> Bound<'py, PyDict> {
+        let locals = PyDict::new(py);
+        locals.set_item("python_contract", PYTHON_CONTRACT).unwrap();
+        py.run(STUBS, Some(&locals), Some(&locals)).unwrap();
+        py.run(script, Some(&locals), Some(&locals)).unwrap();
+        locals
+    }
+
+    pub(crate) fn run(py: Python<'_>, locals: &Bound<'_, PyDict>, code: &CStr) {
+        py.run(code, Some(locals), Some(locals)).unwrap();
+    }
+
+    pub(crate) fn local<'py>(locals: &Bound<'py, PyDict>, name: &str) -> Bound<'py, PyAny> {
+        locals.get_item(name).unwrap().unwrap()
+    }
+
+    /// A legacy call over the namespace's `kwargs` (or none) and `request` (or `None`).
+    pub(crate) fn legacy_call(
+        py: Python<'_>,
+        locals: &Bound<'_, PyDict>,
+        asynchronous: bool,
+    ) -> LegacyLogging {
+        let request = locals
+            .get_item("request")
+            .unwrap()
+            .unwrap_or_else(|| py.None().into_bound(py));
+        let kwargs = locals
+            .get_item("kwargs")
+            .unwrap()
+            .map(|kwargs| kwargs.cast_into::<PyDict>().unwrap())
+            .unwrap_or_else(|| PyDict::new(py));
+        let call = PublicCall::capture(&request, &PyTuple::empty(py), &kwargs).unwrap();
+        LegacyLogging::new(
+            py,
+            LegacySurface {
+                call_type: "test",
+                input_description: "test input",
+                stream: None,
+            },
+            call,
+            asynchronous,
+        )
+    }
+}

litellm-rust/crates/callbacks-legacy-python/tests/deferred.rs

@@ -1,146 +0,0 @@
-use std::ffi::CStr;
-
-use pyo3::prelude::*;
-use pyo3::types::PyDict;
-use rstest::rstest;
-
-use super::{PendingLogging, PendingSuccess};
-use crate::PythonLogger;
-use crate::test_support::{local, namespace, run};
-
-/// A deferred success for the namespace's `logger` and `response`, bound as `pending`.
-fn defer<'py>(py: Python<'py>, script: &CStr) -> Bound<'py, PyDict> {
-    let locals = namespace(py, c"response = object()");
-    run(py, &locals, script);
-    let pending = Py::new(
-        py,
-        PendingLogging {
-            pending: Some(PendingSuccess {
-                logger: PythonLogger::new(local(&locals, "logger").unbind()),
-                response: Some(local(&locals, "response").unbind()),
-                start: py.None(),
-                end: Some(py.None()),
-            }),
-        },
-    )
-    .unwrap();
-    locals.set_item("pending", pending).unwrap();
-    locals
-}
-
-#[test]
-fn release_enqueues_the_success_once_in_the_releasing_context() {
-    Python::initialize();
-    Python::attach(|py| {
-        let locals = defer(
-            py,
-            c"
-from contextvars import ContextVar
-
-marker = ContextVar('marker', default='unset')
-observed = []
-
-def on_enqueue(coroutine):
-    observed.append(marker.get())
-    pending.release(True)
-
-logger.on_enqueue = on_enqueue
-",
-        );
-        run(
-            py,
-            &locals,
-            c"
-marker.set('release')
-pending.release(True)
-pending.release(True)
-assert observed == ['release'], observed
-assert logger.names() == ['async_success_handler', 'enqueued'], logger.calls
-assert logger.calls[0][1] is response
-",
-        );
-    });
-}
-
-#[test]
-fn a_blocked_release_drops_the_success_for_good() {
-    Python::initialize();
-    Python::attach(|py| {
-        let locals = defer(py, c"");
-        run(
-            py,
-            &locals,
-            c"
-pending.release(False)
-pending.release(True)
-assert logger.calls == [], logger.calls
-",
-        );
-    });
-}
-
-#[rstest]
-#[case::ordinary_error(c"RuntimeError('queue full')", false)]
-#[case::cancellation(c"asyncio.CancelledError()", true)]
-fn a_failed_enqueue_closes_the_coroutine_and_is_never_replayed(
-    #[case] failure: &CStr,
-    #[case] propagates: bool,
-) {
-    Python::initialize();
-    Python::attach(|py| {
-        let locals = defer(
-            py,
-            c"
-import asyncio
-
-def on_enqueue(coroutine):
-    raise failure
-
-logger.on_enqueue = on_enqueue
-",
-        );
-        locals
-            .set_item("failure", py.eval(failure, None, Some(&locals)).unwrap())
-            .unwrap();
-        let released = local(&locals, "pending").call_method1("release", (true,));
-        match released {
-            Ok(_) => assert!(!propagates),
-            Err(error) => {
-                assert!(propagates);
-                assert!(error.value(py).is(local(&locals, "failure")));
-            }
-        }
-        locals.set_item("propagates", propagates).unwrap();
-        run(
-            py,
-            &locals,
-            c"
-pending.release(True)
-assert logger.names() == ['async_success_handler', 'enqueued', 'closed'], logger.calls
-assert unraisable_from(logger) == ([] if propagates else [failure])
-",
-        );
-    });
-}
-
-#[test]
-fn an_unreleased_success_does_not_keep_its_logger_alive() {
-    Python::initialize();
-    Python::attach(|py| {
-        let locals = defer(py, c"");
-        run(
-            py,
-            &locals,
-            c"
-import gc
-import weakref
-
-logger.pending = pending
-reference = weakref.ref(logger)
-del logger, pending
-gc.collect()
-assert reference() is None
-",
-        );
-    });
-}

litellm-rust/crates/callbacks-legacy-python/tests/deployment_hooks.rs

@@ -1,282 +0,0 @@
-use std::ffi::CStr;
-
-use litellm_host::event::{FailureOrigin, Timing};
-use litellm_host_python::{LifecycleEvent, LifecycleStep, PythonLifecycle};
-use pyo3::exceptions::asyncio::CancelledError;
-use pyo3::prelude::*;
-use pyo3::types::PyDict;
-use rstest::rstest;
-
-use super::LegacyLogging;
-use crate::test_support::{legacy_call, local, namespace, run};
-
-const CALL: &CStr = c"
-document = {'type': 'document_url', 'document_url': 'data:application/pdf;base64,YWJj'}
-kwargs = {'logger': logger, 'document': document}
-";
-
-const TIMING: Timing = Timing {
-    start_time: 0.0,
-    end_time: 1.0,
-};
-
-fn begin<'py>(
-    py: Python<'py>,
-    locals: &Bound<'py, PyDict>,
-    asynchronous: bool,
-) -> (LegacyLogging, LifecycleStep) {
-    let mut logging = legacy_call(py, locals, asynchronous);
-    let kwargs = local(locals, "kwargs")
-        .cast_into::<PyDict>()
-        .unwrap()
-        .unbind();
-    let step = logging.begin(py, kwargs, 0.0).unwrap();
-    (logging, step)
-}
-
-fn arguments<'py>(py: Python<'py>, step: LifecycleStep) -> Bound<'py, PyDict> {
-    let LifecycleStep::Arguments(arguments) = step else {
-        panic!("expected the prepared arguments");
-    };
-    arguments.into_bound(py)
-}
-
-fn awaits_deployment_hook(step: &LifecycleStep) -> bool {
-    matches!(step, LifecycleStep::Await(_))
-}
-
-#[rstest]
-#[case::synchronous(false)]
-#[case::asynchronous(true)]
-fn deployment_pre_call_hook_runs_only_for_asynchronous_calls(#[case] asynchronous: bool) {
-    Python::initialize();
-    Python::attach(|py| {
-        let locals = namespace(py, CALL);
-        let (_, step) = begin(py, &locals, asynchronous);
-        assert_eq!(awaits_deployment_hook(&step), asynchronous);
-        let names: Vec<String> = local(&locals, "logger")
-            .call_method0("names")
-            .unwrap()
-            .extract()
-            .unwrap();
-        assert_eq!(names.contains(&"pre_hook".to_string()), asynchronous);
-    });
-}
-
-#[test]
-fn kwargs_returned_by_the_pre_call_hook_are_what_the_call_prepares() {
-    Python::initialize();
-    Python::attach(|py| {
-        let locals = namespace(
-            py,
-            c"
-document = {'type': 'document_url', 'document_url': 'data:application/pdf;base64,YWJj'}
-replacement = {'type': 'document_url', 'document_url': 'data:application/pdf;base64,ZWRpdGVk'}
-kwargs = {'logger': logger, 'document': document}
-replaced_kwargs = {'logger': logger, 'document': replacement, 'pages': [0]}
-",
-        );
-        let (mut logging, step) = begin(py, &locals, true);
-        assert!(awaits_deployment_hook(&step));
-        let step = logging
-            .resume(py, Ok(local(&locals, "replaced_kwargs").unbind()))
-            .unwrap();
-        locals.set_item("prepared", arguments(py, step)).unwrap();
-        run(
-            py,
-            &locals,
-            c"
-assert prepared['document'] is replacement
-assert prepared['pages'] is replaced_kwargs['pages']
-assert prepared['litellm_logging_obj'] is logger
-assert 'litellm_logging_obj' not in replaced_kwargs
-[checked] = [value for name, value in logger.calls if name == 'check_limits']
-assert checked is prepared
-",
-        );
-    });
-}
-
-#[rstest]
-#[case::synchronous(false)]
-#[case::asynchronous(true)]
-fn a_keyword_the_bridge_never_reads_reaches_every_reader_as_the_callers_object(
-    #[case] asynchronous: bool,
-) {
-    Python::initialize();
-    Python::attach(|py| {
-        let locals = namespace(
-            py,
-            c"
-opaque = object()
-hooked = []
-logger.hooks = {'pre': lambda kwargs: hooked.append(kwargs['vendor_extension']) or kwargs}
-kwargs = {'logger': logger, 'vendor_extension': opaque}
-",
-        );
-        let (mut logging, step) = begin(py, &locals, asynchronous);
-        let step = match step {
-            LifecycleStep::Await(hook_result) => logging.resume(py, Ok(hook_result)).unwrap(),
-            step => step,
-        };
-        locals.set_item("prepared", arguments(py, step)).unwrap();
-        locals.set_item("asynchronous", asynchronous).unwrap();
-        run(
-            py,
-            &locals,
-            c"
-assert prepared['vendor_extension'] is opaque
-[checked] = [value for name, value in logger.calls if name == 'check_limits']
-assert checked['vendor_extension'] is opaque
-assert hooked == ([opaque] if asynchronous else []), hooked
-",
-        );
-    });
-}
-
-#[test]
-fn response_returned_by_the_post_call_hook_is_finalized_and_returned() {
-    Python::initialize();
-    Python::attach(|py| {
-        let locals = namespace(
-            py,
-            c"
-kwargs = {'logger': logger}
-response = object()
-replacement = object()
-logger.hooks = {'pre': lambda kwargs: kwargs}
-",
-        );
-        let (mut logging, _) = begin(py, &locals, true);
-        logging
-            .resume(py, Ok(local(&locals, "kwargs").unbind()))
-            .unwrap();
-        let step = logging
-            .after_success(py, local(&locals, "response").unbind(), TIMING)
-            .unwrap();
-        assert!(awaits_deployment_hook(&step));
-        let step = logging
-            .resume(py, Ok(local(&locals, "replacement").unbind()))
-            .unwrap();
-        let LifecycleStep::Response(returned) = step else {
-            panic!("expected the finalized response");
-        };
-        assert!(returned.bind(py).is(local(&locals, "replacement")));
-        run(
-            py,
-            &locals,
-            c"
-[finalized] = [value for name, value in logger.calls if name == 'finalize']
-assert finalized is replacement
-",
-        );
-    });
-}
-
-#[rstest]
-#[case::pre_call(false)]
-#[case::post_call(true)]
-fn cancelling_a_deployment_hook_ends_the_call_with_that_cancellation(#[case] post_call: bool) {
-    Python::initialize();
-    Python::attach(|py| {
-        let locals = namespace(py, c"kwargs = {'logger': logger}\nresponse = object()");
-        let (mut logging, _) = begin(py, &locals, true);
-        if post_call {
-            logging
-                .resume(py, Ok(local(&locals, "kwargs").unbind()))
-                .unwrap();
-            logging
-                .after_success(py, local(&locals, "response").unbind(), TIMING)
-                .unwrap();
-        }
-        let cancellation = CancelledError::new_err("cancelled");
-        let cancelled = cancellation.value(py).clone();
-        let error = logging.resume(py, Err(cancellation)).err().unwrap();
-        assert!(error.value(py).is(&cancelled));
-        let names: Vec<String> = local(&locals, "logger")
-            .call_method0("names")
-            .unwrap()
-            .extract()
-            .unwrap();
-        assert!(!names.iter().any(|name| name.contains("handler")));
-    });
-}
-
-#[rstest]
-#[case::hook_completed(false)]
-#[case::hook_cancelled(true)]
-fn failure_callbacks_run_after_the_failure_hook_however_it_ends(#[case] cancelled: bool) {
-    Python::initialize();
-    Python::attach(|py| {
-        let locals = namespace(
-            py,
-            c"kwargs = {'logger': logger}\nfailure = ValueError('provider')",
-        );
-        let (mut logging, _) = begin(py, &locals, true);
-        logging
-            .resume(py, Ok(local(&locals, "kwargs").unbind()))
-            .unwrap();
-        let failure = PyErr::from_value(local(&locals, "failure"));
-        let failed = LifecycleEvent::Failed {
-            timing: TIMING,
-            origin: FailureOrigin::Call,
-            error: &failure,
-        };
-        let step = logging.emit(py, failed).unwrap();
-        assert!(awaits_deployment_hook(&step));
-        let hook_result = if cancelled {
-            Err(CancelledError::new_err("cancelled"))
-        } else {
-            Ok(py.None())
-        };
-        assert!(matches!(
-            logging.resume(py, hook_result).unwrap(),
-            LifecycleStep::Await(_)
-        ));
-        run(
-            py,
-            &locals,
-            c"
-assert logger.names()[-3:] == ['failure_hook', 'failure_handler', 'async_failure_handler'], logger.calls
-assert all(value is failure for name, value in logger.calls if name.endswith('_handler'))
-",
-        );
-    });
-}
-
-#[rstest]
-#[case::synchronous(false)]
-#[case::asynchronous(true)]
-fn a_limit_rejected_before_the_call_surfaces_as_the_callers_error(#[case] asynchronous: bool) {
-    Python::initialize();
-    Python::attach(|py| {
-        let locals = namespace(
-            py,
-            c"
-class BudgetExceeded(Exception):
-    pass
-
-rejection = BudgetExceeded('over budget')
-
-class LimitedLogger(StubLogger):
-    def check_limits(self, arguments):
-        raise rejection
-
-logger = LimitedLogger()
-logger.hooks = {'pre': lambda kwargs: kwargs}
-kwargs = {'logger': logger}
-",
-        );
-        let mut logging = legacy_call(py, &locals, asynchronous);
-        let kwargs = local(&locals, "kwargs")
-            .cast_into::<PyDict>()
-            .unwrap()
-            .unbind();
-        let result = logging.begin(py, kwargs, 0.0).and_then(|step| match step {
-            LifecycleStep::Await(_) => logging.resume(py, Ok(local(&locals, "kwargs").unbind())),
-            step => Ok(step),
-        });
-        let error = result.err().unwrap();
-        assert!(error.value(py).is(local(&locals, "rejection")));
-    });
-}

litellm-rust/crates/callbacks-legacy-python/tests/payload.rs

@@ -1,523 +0,0 @@
-use std::ffi::CStr;
-
-use litellm_auth::SecretValue;
-use litellm_host::event::{MachineEvent, RawResponse, RequestContext, WireRequest};
-use litellm_host_python::{LifecycleEvent, LifecycleStep, PythonLifecycle, to_py};
-use proptest::prelude::*;
-use pyo3::prelude::*;
-use rstest::rstest;
-use serde_json::{Map, Value, json};
-
-use super::LegacyLogging;
-use crate::PythonLogger;
-use crate::test_support::{legacy_call, local, namespace, run};
-
-/// The payload phases of `Logging` on top of `StubLogger`, with `pre_call` handing the
-/// payload to the case's `on_pre_call`.
-const PAYLOAD_LOGGER: &CStr = c"
-class Request:
-    pass
-
-class PayloadLogger(StubLogger):
-    def update_from_kwargs(self, **update):
-        self.update = update
-
-    def pre_call(self, input, api_key, additional_args):
-        self.record('pre_call', None)
-        self.pre = additional_args
-        self.pre_api_key = api_key
-        on_pre_call(additional_args)
-
-    def post_call(self, original_response, api_key, additional_args):
-        self.record('post_call', None)
-        self.post = (original_response, api_key, additional_args)
-
-request = Request()
-kwargs = {}
-logger = PayloadLogger()
-on_pre_call = lambda additional_args: None
-check = lambda: None
-";
-
-const DOCUMENT: &str = "data:application/pdf;base64,YWJj";
-const EDITED: &str = "data:application/pdf;base64,ZWRpdGVk";
-
-fn document(source: &str) -> Value {
-    json!({"type": "document_url", "document_url": source})
-}
-
-fn before_send(script: &CStr, body: Value) -> WireRequest {
-    before_send_with_secrets(script, json!({}), body, &[])
-}
-
-/// Runs `before_send` over `body` for a route whose parameters are `optional_params`, with
-/// the Python objects `script` binds, then delivers the provider's raw response the way the
-/// driver does and runs the script's `check()`.
-fn before_send_with_secrets(
-    script: &CStr,
-    optional_params: Value,
-    body: Value,
-    secret_fields: &[&str],
-) -> WireRequest {
-    before_send_bound(&[], script, optional_params, body, secret_fields)
-}
-
-/// [`before_send_with_secrets`] with `bindings` placed in the namespace before `script` runs.
-fn before_send_bound(
-    bindings: &[(&str, &Value)],
-    script: &CStr,
-    optional_params: Value,
-    body: Value,
-    secret_fields: &[&str],
-) -> WireRequest {
-    Python::initialize();
-    Python::attach(|py| {
-        let locals = namespace(py, PAYLOAD_LOGGER);
-        for &(name, value) in bindings {
-            locals.set_item(name, to_py(py, value).unwrap()).unwrap();
-        }
-        run(py, &locals, script);
-        let mut logging = LegacyLogging {
-            logger: Some(PythonLogger::new(local(&locals, "logger").unbind())),
-            ..legacy_call(py, &locals, false)
-        };
-        let context = RequestContext {
-            model: "model".into(),
-            custom_llm_provider: "provider".into(),
-            optional_params,
-            secret_fields: secret_fields.iter().map(|name| name.to_string()).collect(),
-            api_key: Some(SecretValue::new("route-key")),
-        };
-        let wire = WireRequest {
-            url: "https://provider.invalid/ocr".into(),
-            headers: vec![("x-route".into(), "route".into())],
-            body,
-        };
-        let step = logging.before_send(py, Box::new(wire), &context).unwrap();
-        let raw = MachineEvent::ResponseReceived {
-            raw: RawResponse {
-                body: "raw response".into(),
-            },
-        };
-        assert!(matches!(
-            logging.emit(py, LifecycleEvent::Machine(&raw)).unwrap(),
-            LifecycleStep::Done
-        ));
-        run(py, &locals, c"check()");
-        let LifecycleStep::Wire(wire) = step else {
-            panic!("before_send did not hand back the wire request");
-        };
-        *wire
-    })
-}
-
-#[rstest]
-#[case::caller_keyword(c"
-document = {'type': 'document_url', 'document_url': 'data:application/pdf;base64,YWJj'}
-pages = [0]
-kwargs = {'document': document, 'pages': pages}
-observed = []
-on_pre_call = lambda args: observed.append(
-    (args['complete_input_dict']['document'] is document, args['complete_input_dict']['pages'] is pages)
-)
-def check():
-    assert observed == [(True, True)], observed
-")]
-#[case::request_attribute_behind_an_omitted_keyword(c"
-document = {'type': 'document_url', 'document_url': 'data:application/pdf;base64,YWJj'}
-pages = [0]
-request.document = document
-kwargs = {'pages': pages}
-observed = []
-on_pre_call = lambda args: observed.append(
-    (args['complete_input_dict']['document'] is document, args['complete_input_dict']['pages'] is pages)
-)
-def check():
-    assert observed == [(True, True)], observed
-")]
-fn passthrough_keys_reach_pre_call_as_the_callers_own_objects(#[case] script: &CStr) {
-    let body = json!({"model": "model", "document": document(DOCUMENT), "pages": [0]});
-    let wire = before_send(script, body.clone());
-    assert_eq!(wire.body, body);
-}
-
-#[test]
-fn pre_call_edit_of_a_passthrough_object_reaches_the_caller_and_the_wire() {
-    let wire = before_send(
-        c"
-document = {'type': 'document_url', 'document_url': 'data:application/pdf;base64,YWJj'}
-kwargs = {'document': document}
-def on_pre_call(args):
-    args['complete_input_dict']['document']['document_url'] = 'data:application/pdf;base64,ZWRpdGVk'
-def check():
-    assert document['document_url'] == 'data:application/pdf;base64,ZWRpdGVk'
-",
-        json!({"document": document(DOCUMENT)}),
-    );
-    assert_eq!(wire.body["document"], document(EDITED));
-}
-
-#[test]
-fn a_body_key_the_route_rewrote_is_not_the_callers_object() {
-    let wire = before_send(
-        c"
-document = {'type': 'document_url', 'document_url': 'https://example.invalid/scan.pdf'}
-kwargs = {'document': document}
-observed = []
-def on_pre_call(args):
-    observed.append(args['complete_input_dict']['document'] is document)
-    args['complete_input_dict']['document']['document_name'] = 'edited.pdf'
-def check():
-    assert observed == [False], observed
-    assert document == {'type': 'document_url', 'document_url': 'https://example.invalid/scan.pdf'}
-",
-        json!({"document": document(DOCUMENT)}),
-    );
-    assert_eq!(
-        wire.body["document"],
-        json!({"type": "document_url", "document_url": DOCUMENT, "document_name": "edited.pdf"})
-    );
-}
-
-#[test]
-fn a_caller_value_with_no_json_form_is_left_out_of_realiasing() {
-    let body = json!({"pages": [0]});
-    let wire = before_send(
-        c"
-opaque = object()
-kwargs = {'pages': opaque}
-observed = []
-on_pre_call = lambda args: observed.append(args['complete_input_dict']['pages'])
-def check():
-    assert observed == [[0]], observed
-",
-        body.clone(),
-    );
-    assert_eq!(wire.body, body);
-}
-
-#[rstest]
-#[case::body(
-    c"
-def on_pre_call(args):
-    args['complete_input_dict'] = {'replacement': True}
-"
-)]
-#[case::headers(
-    c"
-def on_pre_call(args):
-    args['headers'] = {'x-replacement': 'yes'}
-"
-)]
-fn rebinding_the_payload_envelope_does_not_reach_the_wire(#[case] script: &CStr) {
-    let body = json!({"document": document(DOCUMENT)});
-    let wire = before_send(script, body.clone());
-    assert_eq!(wire.body, body);
-    assert_eq!(wire.headers, [("x-route".to_string(), "route".to_string())]);
-}
-
-#[test]
-fn pre_call_header_edit_reaches_the_wire() {
-    let wire = before_send(
-        c"
-def on_pre_call(args):
-    args['headers']['x-callback'] = 'edited'
-",
-        json!({}),
-    );
-    assert_eq!(
-        wire.headers,
-        [
-            ("x-route".to_string(), "route".to_string()),
-            ("x-callback".to_string(), "edited".to_string()),
-        ]
-    );
-}
-
-#[test]
-fn pre_call_receives_the_wire_request_and_the_logger_its_redacted_request() {
-    let body = json!({"model": "model", "document": document(DOCUMENT)});
-    before_send_with_secrets(
-        c"
-logger_fn = lambda *args: None
-kwargs = {
-    'litellm_call_id': 'call-1',
-    'client_secret': 'shh',
-    'proxy_server_request': {'body': {}},
-    'logger_fn': logger_fn,
-    'litellm_request_debug': True,
-    'ocr_cost_per_page': 0.05,
-}
-observed = []
-on_pre_call = observed.append
-def check():
-    [args] = observed
-    assert args['api_base'] == 'https://provider.invalid/ocr', args
-    assert args['complete_input_dict'] == {
-        'model': 'model',
-        'document': {'type': 'document_url', 'document_url': 'data:application/pdf;base64,YWJj'},
-    }, args
-    update = logger.update
-    assert update['model'] == 'model' and update['custom_llm_provider'] == 'provider', update
-    assert update['litellm_params']['litellm_call_id'] == 'call-1', update
-    assert update['litellm_params']['api_base'] == 'https://provider.invalid/ocr', update
-    assert update['litellm_params']['logger_fn'] is logger_fn, update
-    assert update['litellm_params']['litellm_request_debug'] is True, update
-    assert update['litellm_params']['ocr_cost_per_page'] == 0.05, update
-    assert update['kwargs']['client_secret'] == '****', update
-    assert 'proxy_server_request' not in update['kwargs'], update
-    assert update['optional_params']['client_secret'] == '****', update
-",
-        json!({"client_secret": "shh"}),
-        body,
-        &["client_secret"],
-    );
-}
-
-#[rstest]
-#[case::added_key(
-    c"
-def on_pre_call(args):
-    args['complete_input_dict']['include_image_base64'] = True
-",
-    json!({"document": document(DOCUMENT), "include_image_base64": true})
-)]
-#[case::replaced_document(
-    c"
-document = {'type': 'document_url', 'document_url': 'data:application/pdf;base64,YWJj'}
-kwargs = {'document': document}
-def on_pre_call(args):
-    args['complete_input_dict']['document'] = {
-        'type': 'document_url', 'document_url': 'data:application/pdf;base64,ZWRpdGVk'
-    }
-def check():
-    assert document['document_url'] == 'data:application/pdf;base64,YWJj', document
-",
-    json!({"document": document(EDITED)})
-)]
-#[case::retained_body_edited_after_rebinding(
-    c"
-def on_pre_call(args):
-    retained = args['complete_input_dict']
-    args['complete_input_dict'] = {'rebound': True}
-    retained['include_image_base64'] = True
-",
-    json!({"document": document(DOCUMENT), "include_image_base64": true})
-)]
-fn pre_call_body_edits_reach_the_wire(#[case] script: &CStr, #[case] expected: Value) {
-    let body = json!({"document": document(DOCUMENT)});
-    let wire = before_send(script, body);
-    assert_eq!(wire.body, expected);
-}
-
-#[test]
-fn retained_headers_edited_after_rebinding_reach_the_wire() {
-    let wire = before_send(
-        c"
-def on_pre_call(args):
-    retained = args['headers']
-    args['headers'] = {'x-rebound': 'rebound'}
-    retained['x-retained'] = 'sent'
-",
-        json!({}),
-    );
-    assert_eq!(
-        wire.headers,
-        [
-            ("x-route".to_string(), "route".to_string()),
-            ("x-retained".to_string(), "sent".to_string()),
-        ]
-    );
-}
-
-#[test]
-fn post_call_receives_the_raw_response_the_route_key_and_the_body_and_headers_pre_call_saw() {
-    before_send(
-        c"
-def check():
-    original_response, api_key, additional_args = logger.post
-    assert original_response == 'raw response', original_response
-    assert api_key == logger.pre_api_key == 'route-key', (api_key, logger.pre_api_key)
-    assert additional_args == {
-        'complete_input_dict': logger.pre['complete_input_dict'],
-        'headers': logger.pre['headers'],
-    }, additional_args
-    assert additional_args['complete_input_dict'] is logger.pre['complete_input_dict']
-    assert additional_args['headers'] is logger.pre['headers']
-",
-        json!({"document": document(DOCUMENT)}),
-    );
-}
-
-#[test]
-fn every_request_runs_the_full_pre_call_and_post_call() {
-    let wire = before_send(
-        c"
-def on_pre_call(args):
-    args['complete_input_dict']['include_image_base64'] = True
-def check():
-    assert logger.names() == ['pre_call', 'post_call'], logger.calls
-",
-        json!({"document": document(DOCUMENT)}),
-    );
-    assert_eq!(
-        wire.body,
-        json!({"document": document(DOCUMENT), "include_image_base64": true})
-    );
-}
-
-/// What one pre-call callback does to the payload it is handed.
-#[derive(Clone, Debug)]
-enum Edit {
-    Nothing,
-    Set(String, Value),
-    Remove(String),
-    Rebind(Value),
-    RebindThenSetRetained(String, Value),
-}
-
-impl Edit {
-    fn script(&self) -> Value {
-        match self {
-            Self::Nothing => json!({"kind": "nothing"}),
-            Self::Set(key, value) => json!({"kind": "set", "key": key, "value": value}),
-            Self::Remove(key) => json!({"kind": "remove", "key": key}),
-            Self::Rebind(value) => json!({"kind": "rebind", "value": value}),
-            Self::RebindThenSetRetained(key, value) => {
-                json!({"kind": "rebind_then_set_retained", "key": key, "value": value})
-            }
-        }
-    }
-
-    /// The legacy contract: the provider is sent the body object `pre_call` received, as
-    /// the callback left it. Rebinding the envelope's key points the envelope elsewhere and
-    /// leaves that object alone.
-    fn sent(&self, body: &Map<String, Value>) -> Value {
-        let mut sent = body.clone();
-        match self {
-            Self::Nothing | Self::Rebind(_) => {}
-            Self::Set(key, value) | Self::RebindThenSetRetained(key, value) => {
-                sent.insert(key.clone(), value.clone());
-            }
-            Self::Remove(key) => {
-                sent.remove(key);
-            }
-        }
-        Value::Object(sent)
-    }
-}
-
-/// How the caller's keyword for a body key relates to what the route sends under it.
-#[derive(Clone, Copy, Debug, PartialEq, Eq)]
-enum Caller {
-    PassedUnchanged,
-    RewrittenByTheRoute,
-    NotPassed,
-}
-
-const MODEL: &CStr = c"
-aliased = {}
-def on_pre_call(args):
-    body = args['complete_input_dict']
-    aliased.update({name: body[name] is kwargs[name] for name in unchanged})
-    kind = edit['kind']
-    if kind == 'set':
-        body[edit['key']] = edit['value']
-    elif kind == 'remove':
-        body.pop(edit['key'], None)
-    elif kind == 'rebind':
-        args['complete_input_dict'] = edit['value']
-    elif kind == 'rebind_then_set_retained':
-        args['complete_input_dict'] = {}
-        body[edit['key']] = edit['value']
-def check():
-    assert aliased == {name: True for name in unchanged}, aliased
-    assert logger.names() == ['pre_call', 'post_call'], logger.calls
-";
-
-fn json_value() -> impl Strategy<Value = Value> {
-    let leaf = prop_oneof![
-        Just(Value::Null),
-        any::<bool>().prop_map(Value::from),
-        any::<i64>().prop_map(Value::from),
-        any::<f64>()
-            .prop_filter("JSON has no NaN or infinity", |number| number.is_finite())
-            .prop_map(Value::from),
-        ".{0,8}".prop_map(Value::from),
-    ];
-    leaf.prop_recursive(3, 24, 4, |inner| {
-        prop_oneof![
-            prop::collection::vec(inner.clone(), 0..4).prop_map(Value::from),
-            prop::collection::btree_map(key(), inner, 0..4)
-                .prop_map(|fields| Value::Object(fields.into_iter().collect())),
-        ]
-    })
-}
-
-fn key() -> impl Strategy<Value = String> {
-    "[a-z]{1,6}"
-}
-
-fn caller() -> impl Strategy<Value = Caller> {
-    prop_oneof![
-        Just(Caller::PassedUnchanged),
-        Just(Caller::RewrittenByTheRoute),
-        Just(Caller::NotPassed),
-    ]
-}
-
-fn edit() -> impl Strategy<Value = Edit> {
-    prop_oneof![
-        Just(Edit::Nothing),
-        (key(), json_value()).prop_map(|(key, value)| Edit::Set(key, value)),
-        key().prop_map(Edit::Remove),
-        json_value().prop_map(Edit::Rebind),
-        (key(), json_value()).prop_map(|(key, value)| Edit::RebindThenSetRetained(key, value)),
-    ]
-}
-
-proptest! {
-    #![proptest_config(ProptestConfig::with_cases(128))]
-
-    /// For any body, any caller keywords and any callback edit: every keyword the route
-    /// sends unchanged reaches `pre_call` as the caller's own object, and the provider is
-    /// sent exactly what the model says, so a callback that edits nothing changes nothing.
-    #[test]
-    fn the_wire_is_the_body_pre_call_received_as_the_callback_left_it(
-        fields in prop::collection::btree_map(key(), (json_value(), caller()), 0..5),
-        edit in edit(),
-    ) {
-        let body: Map<String, Value> = fields
-            .iter()
-            .map(|(name, (value, _))| (name.clone(), value.clone()))
-            .collect();
-        let kwargs: Map<String, Value> = fields
-            .iter()
-            .filter_map(|(name, (value, caller))| match caller {
-                Caller::PassedUnchanged => Some((name.clone(), value.clone())),
-                Caller::RewrittenByTheRoute => Some((name.clone(), json!([value]))),
-                Caller::NotPassed => None,
-            })
-            .collect();
-        let unchanged: Value = fields
-            .iter()
-            .filter(|(_, (_, caller))| *caller == Caller::PassedUnchanged)
-            .map(|(name, _)| Value::from(name.clone()))
-            .collect();
-
-        let wire = before_send_bound(
-            &[
-                ("kwargs", &Value::Object(kwargs)),
-                ("unchanged", &unchanged),
-                ("edit", &edit.script()),
-            ],
-            MODEL,
-            json!({}),
-            Value::Object(body.clone()),
-            &[],
-        );
-
-        prop_assert_eq!(wire.body, edit.sent(&body));
-        prop_assert_eq!(wire.headers, [("x-route".to_string(), "route".to_string())]);
-    }
-}

litellm-rust/crates/callbacks-legacy-python/tests/support.rs

@@ -1,205 +0,0 @@
-use std::ffi::CStr;
-
-use pyo3::prelude::*;
-use pyo3::types::{PyDict, PyTuple};
-
-use crate::{LegacyLogging, LegacySurface, PublicCall};
-
-/// The parameters of every `callbacks_legacy_python` function, as the real module declares them.
-/// `tests/test_litellm/rust_bridge/test_callbacks_legacy_python.py` pins this file to the Python
-/// signatures, and [`namespace`] binds every fake call against it.
-pub(crate) const PYTHON_CONTRACT: &str = include_str!("../python_contract.json");
-
-/// Stand-ins for `callbacks_legacy_python`, the only Python module the crate calls. Tests
-/// share one interpreter and run concurrently, so each fake is installed idempotently and
-/// forwards to the per-test `StubLogger` it is handed (directly, or as `kwargs['logger']`).
-/// Every fake is bound against the contract first, so a call the real module would reject
-/// fails here too.
-const STUBS: &CStr = c"
-import contextvars
-import inspect
-import json
-import sys
-import traceback
-import types
-
-for name in ('litellm', 'litellm.rust_bridge', 'litellm.rust_bridge.callbacks_legacy_python'):
-    sys.modules.setdefault(name, types.ModuleType(name))
-
-legacy = sys.modules['litellm.rust_bridge.callbacks_legacy_python']
-CONTRACT = json.loads(python_contract)
-
-
-def contracted(name, fake):
-    signature = inspect.Signature(
-        [inspect.Parameter(parameter, inspect.Parameter.POSITIONAL_OR_KEYWORD) for parameter in CONTRACT[name]]
-    )
-
-    def checked(*args, **kwargs):
-        signature.bind(*args, **kwargs)
-        return fake(*args, **kwargs)
-
-    return checked
-
-
-if not hasattr(legacy, 'is_internal'):
-    legacy.is_internal = contextvars.ContextVar('is_internal_call', default=False)
-
-FAKES = {
-    'setup': lambda call_type, args, kwargs, start, asynchronous: types.SimpleNamespace(
-        logger=kwargs['logger_factory'](kwargs) if 'logger_factory' in kwargs else kwargs['logger'],
-        kwargs=kwargs,
-    ),
-    'check_limits': lambda arguments: arguments['logger'].check_limits(arguments),
-    'finalize': lambda response, logger, kwargs, start, end: logger.record('finalize', response),
-    'update_logging': lambda logger, kwargs, model, optional_params, litellm_params, provider: logger.update_from_kwargs(
-        kwargs=kwargs,
-        model=model,
-        optional_params=optional_params,
-        litellm_params=litellm_params,
-        custom_llm_provider=provider,
-    ),
-    'pre_call': lambda logger, input, api_key, additional_args: logger.pre_call(input, api_key, additional_args),
-    'post_call': lambda logger, original_response, api_key, additional_args: logger.post_call(
-        original_response, api_key, additional_args
-    ),
-    'defers_async_logging': lambda logger: bool(getattr(logger, '_defer_async_logging', False)),
-    'defer_success': lambda logger, pending: setattr(logger, '_native_pending_logging', pending),
-    'sync_success_for_async_call': lambda logger, response, start, end: logger.handle_sync_success_callbacks_for_async_calls(
-        response, start, end
-    ),
-    'failure_handler': lambda logger, error, start, end, asynchronous: (
-        logger.async_failure_handler if asynchronous else logger.failure_handler
-    )(error, ''.join(traceback.format_exception(error)), start, end),
-    'submit_success': lambda logger, response, start, end: logger.record('submit', (response, start, end)),
-    'async_success_handler': lambda logger, response, start, end: logger.async_success_handler(response, start, end),
-    'enqueue_logging': lambda coroutine: coroutine.enqueue(),
-    'restore_context': lambda logger: logger.record('restore', None),
-    'custom_pricing_fields': lambda: ('ocr_cost_per_page',),
-    'is_internal_call': lambda: legacy.is_internal.get(),
-    'credential_list': lambda: [],
-    'warn_unknown_credential': lambda name, loaded: None,
-    'before_deployment_call': lambda kwargs, call_type: kwargs['logger'].hook('pre', kwargs, call_type),
-    'after_deployment_success': lambda kwargs, response, call_type: kwargs['logger'].hook(
-        'success', response, call_type
-    ),
-    'after_deployment_failure': lambda kwargs, error, call_type: kwargs['logger'].hook('failure', error, call_type),
-    'stream_opened': lambda logger: logger.record('stream_opened', None),
-    'stream_success': lambda logger, request_body, chunks, start, end, first_chunk: logger.record(
-        'stream_success', list(chunks)
-    ),
-    'stream_failure': lambda logger, request_body, chunks, error: logger.record('stream_failure', error),
-}
-assert FAKES.keys() == CONTRACT.keys(), sorted(FAKES.keys() ^ CONTRACT.keys())
-for name, fake in FAKES.items():
-    setattr(legacy, name, contracted(name, fake))
-
-
-unraisable = sys.modules.setdefault(
-    'litellm_test_unraisable', types.ModuleType('litellm_test_unraisable')
-)
-if not hasattr(unraisable, 'events'):
-    unraisable.events = []
-    sys.unraisablehook = lambda event: unraisable.events.append((event.object, event.exc_value))
-
-
-def unraisable_from(owner):
-    return [error for source, error in unraisable.events if source is owner]
-
-
-class StubCoroutine:
-    def __init__(self, logger):
-        self.logger = logger
-
-    def enqueue(self):
-        self.logger.record('enqueued', None)
-        self.logger.on_enqueue(self)
-
-    def close(self):
-        self.logger.record('closed', None)
-
-
-class StubLogger:
-    def __init__(self):
-        self.calls = []
-        self.hooks = {}
-        self.on_enqueue = lambda coroutine: None
-
-    def record(self, name, value):
-        self.calls.append((name, value))
-
-    def names(self):
-        return [name for name, _ in self.calls]
-
-    def hook(self, phase, value, call_type):
-        self.record(phase + '_hook', call_type)
-        return self.hooks.get(phase, lambda value: 'awaitable')(value)
-
-    def check_limits(self, arguments):
-        self.record('check_limits', arguments)
-
-    def failure_handler(self, error, trace, start, end):
-        self.record('failure_handler', error)
-
-    def async_failure_handler(self, error, trace, start, end):
-        self.record('async_failure_handler', error)
-        return 'awaitable'
-
-    def success_handler(self, response, start, end):
-        self.record('success_handler', response)
-
-    def async_success_handler(self, response, start, end):
-        self.record('async_success_handler', response)
-        return StubCoroutine(self)
-
-    def handle_sync_success_callbacks_for_async_calls(self, response, start, end):
-        self.record('sync_success_for_async_call', response)
-
-
-logger = StubLogger()
-";
-
-/// A namespace with the stubs, `StubLogger` and a fresh `logger`, after `script` ran in it.
-pub(crate) fn namespace<'py>(py: Python<'py>, script: &CStr) -> Bound<'py, PyDict> {
-    let locals = PyDict::new(py);
-    locals.set_item("python_contract", PYTHON_CONTRACT).unwrap();
-    py.run(STUBS, Some(&locals), Some(&locals)).unwrap();
-    py.run(script, Some(&locals), Some(&locals)).unwrap();
-    locals
-}
-
-pub(crate) fn run(py: Python<'_>, locals: &Bound<'_, PyDict>, code: &CStr) {
-    py.run(code, Some(locals), Some(locals)).unwrap();
-}
-
-pub(crate) fn local<'py>(locals: &Bound<'py, PyDict>, name: &str) -> Bound<'py, PyAny> {
-    locals.get_item(name).unwrap().unwrap()
-}
-
-/// A legacy call over the namespace's `kwargs` (or none) and `request` (or `None`).
-pub(crate) fn legacy_call(
-    py: Python<'_>,
-    locals: &Bound<'_, PyDict>,
-    asynchronous: bool,
-) -> LegacyLogging {
-    let request = locals
-        .get_item("request")
-        .unwrap()
-        .unwrap_or_else(|| py.None().into_bound(py));
-    let kwargs = locals
-        .get_item("kwargs")
-        .unwrap()
-        .map(|kwargs| kwargs.cast_into::<PyDict>().unwrap())
-        .unwrap_or_else(|| PyDict::new(py));
-    let call = PublicCall::capture(&request, &PyTuple::empty(py), &kwargs).unwrap();
-    LegacyLogging::new(
-        py,
-        LegacySurface {
-            call_type: "test",
-            input_description: "test input",
-            stream: None,
-        },
-        call,
-        asynchronous,
-    )
-}

litellm-rust/crates/callbacks-legacy-python/tests/terminal.rs

@@ -1,291 +0,0 @@
-use std::ffi::CStr;
-
-use litellm_host::event::{FailureOrigin, Timing};
-use litellm_host_python::{LifecycleEvent, LifecycleStep, PythonLifecycle};
-use pyo3::exceptions::PyRuntimeError;
-use pyo3::exceptions::asyncio::CancelledError;
-use pyo3::prelude::*;
-use pyo3::types::PyDict;
-use rstest::rstest;
-
-use super::LegacyLogging;
-use crate::PythonLogger;
-use crate::test_support::{legacy_call, local, namespace, run};
-
-const TIMING: Timing = Timing {
-    start_time: 0.0,
-    end_time: 1.0,
-};
-
-fn logged(py: Python<'_>, locals: &Bound<'_, PyDict>, asynchronous: bool) -> LegacyLogging {
-    LegacyLogging {
-        logger: Some(PythonLogger::new(local(locals, "logger").unbind())),
-        ..legacy_call(py, locals, asynchronous)
-    }
-}
-
-fn succeed(
-    py: Python<'_>,
-    locals: &Bound<'_, PyDict>,
-    logging: &mut LegacyLogging,
-) -> LifecycleStep {
-    let response = local(locals, "response").unbind();
-    logging
-        .emit(
-            py,
-            LifecycleEvent::Succeeded {
-                timing: TIMING,
-                response: &response,
-            },
-        )
-        .unwrap()
-}
-
-fn fail(py: Python<'_>, locals: &Bound<'_, PyDict>, logging: &mut LegacyLogging) -> LifecycleStep {
-    let failure = PyErr::from_value(local(locals, "failure"));
-    logging
-        .emit(
-            py,
-            LifecycleEvent::Failed {
-                timing: TIMING,
-                origin: FailureOrigin::Host,
-                error: &failure,
-            },
-        )
-        .unwrap()
-}
-
-#[rstest]
-#[case::sync_listened(false, c"", &["submit"])]
-#[case::async_listened(
-    true,
-    c"",
-    &["async_success_handler", "enqueued", "sync_success_for_async_call"]
-)]
-#[case::async_deferred(true, c"logger._defer_async_logging = True", &["sync_success_for_async_call"])]
-#[case::async_with_fallbacks(true, c"kwargs = {'fallbacks': ['other']}", &["sync_success_for_async_call"])]
-fn success_reaches_the_logging_handlers(
-    #[case] asynchronous: bool,
-    #[case] script: &CStr,
-    #[case] expected: &[&str],
-) {
-    Python::initialize();
-    Python::attach(|py| {
-        let locals = namespace(py, c"response = object()");
-        run(py, &locals, script);
-        let mut logging = logged(py, &locals, asynchronous);
-        assert!(matches!(
-            succeed(py, &locals, &mut logging),
-            LifecycleStep::Done
-        ));
-        let names: Vec<String> = local(&locals, "logger")
-            .call_method0("names")
-            .unwrap()
-            .extract()
-            .unwrap();
-        assert_eq!(names, expected);
-        run(
-            py,
-            &locals,
-            c"
-assert all(value is response for name, value in logger.calls if name.endswith('_handler'))
-assert hasattr(logger, '_native_pending_logging') == getattr(logger, '_defer_async_logging', False)
-",
-        );
-    });
-}
-
-#[rstest]
-#[case::synchronous(false, &["failure_handler"])]
-#[case::asynchronous(true, &[])]
-fn internal_calls_skip_failure_callbacks_only_when_asynchronous(
-    #[case] asynchronous: bool,
-    #[case] expected: &[&str],
-) {
-    Python::initialize();
-    Python::attach(|py| {
-        let locals = namespace(py, c"failure = ValueError('provider')");
-        let mut logging = LegacyLogging {
-            internal: true,
-            ..logged(py, &locals, asynchronous)
-        };
-        assert!(matches!(
-            fail(py, &locals, &mut logging),
-            LifecycleStep::Done
-        ));
-        let names: Vec<String> = local(&locals, "logger")
-            .call_method0("names")
-            .unwrap()
-            .extract()
-            .unwrap();
-        assert_eq!(names, expected);
-    });
-}
-
-#[test]
-fn internal_async_calls_skip_the_async_success_fan_out() {
-    Python::initialize();
-    Python::attach(|py| {
-        let locals = namespace(py, c"response = object()");
-        let mut logging = LegacyLogging {
-            internal: true,
-            ..logged(py, &locals, true)
-        };
-        succeed(py, &locals, &mut logging);
-        run(
-            py,
-            &locals,
-            c"assert logger.names() == ['sync_success_for_async_call'], logger.calls",
-        );
-    });
-}
-
-#[test]
-fn a_failing_success_callback_is_reported_without_replacing_the_response() {
-    Python::initialize();
-    Python::attach(|py| {
-        let locals = namespace(
-            py,
-            c"
-response = object()
-failure = ValueError('terminal diagnostic')
-
-class FailingLogger(StubLogger):
-    def handle_sync_success_callbacks_for_async_calls(self, *args):
-        raise failure
-
-logger = FailingLogger()
-",
-        );
-        let mut logging = logged(py, &locals, true);
-        assert!(matches!(
-            succeed(py, &locals, &mut logging),
-            LifecycleStep::Done
-        ));
-        assert!(
-            logging
-                .response
-                .as_ref()
-                .unwrap()
-                .bind(py)
-                .is(local(&locals, "response"))
-        );
-        run(py, &locals, c"assert unraisable_from(logger) == [failure]");
-    });
-}
-
-#[rstest]
-#[case::sync_listened(false, c"", &["failure_handler"])]
-#[case::async_listened(true, c"", &["failure_handler", "async_failure_handler"])]
-fn failure_reaches_the_logging_handlers(
-    #[case] asynchronous: bool,
-    #[case] script: &CStr,
-    #[case] expected: &[&str],
-) {
-    Python::initialize();
-    Python::attach(|py| {
-        let locals = namespace(py, c"failure = ValueError('provider')");
-        run(py, &locals, script);
-        let mut logging = logged(py, &locals, asynchronous);
-        let step = fail(py, &locals, &mut logging);
-        let awaits_async_handler = expected.contains(&"async_failure_handler");
-        assert_eq!(
-            matches!(step, LifecycleStep::Await(_)),
-            awaits_async_handler
-        );
-        let names: Vec<String> = local(&locals, "logger")
-            .call_method0("names")
-            .unwrap()
-            .extract()
-            .unwrap();
-        assert_eq!(names, expected);
-        run(
-            py,
-            &locals,
-            c"assert all(value is failure for name, value in logger.calls if name.endswith('_handler'))",
-        );
-    });
-}
-
-#[test]
-fn a_failing_sync_failure_callback_keeps_the_error_and_still_runs_the_async_family() {
-    Python::initialize();
-    Python::attach(|py| {
-        let locals = namespace(
-            py,
-            c"
-failure = ValueError('selected')
-
-class FailingLogger(StubLogger):
-    def failure_handler(self, error, trace, start, end):
-        self.record('failure_handler', error)
-        raise RuntimeError('handler failed')
-
-logger = FailingLogger()
-",
-        );
-        let mut logging = logged(py, &locals, true);
-        assert!(matches!(
-            fail(py, &locals, &mut logging),
-            LifecycleStep::Await(_)
-        ));
-        assert!(
-            logging
-                .error
-                .as_ref()
-                .unwrap()
-                .bind(py)
-                .is(local(&locals, "failure"))
-        );
-        run(
-            py,
-            &locals,
-            c"assert logger.names() == ['failure_handler', 'async_failure_handler'], logger.calls",
-        );
-    });
-}
-
-#[rstest]
-#[case::completed(None, true)]
-#[case::handler_error(Some(false), true)]
-#[case::cancelled(Some(true), false)]
-fn the_async_failure_handler_ends_the_call_unless_it_was_cancelled(
-    #[case] error: Option<bool>,
-    #[case] done: bool,
-) {
-    Python::initialize();
-    Python::attach(|py| {
-        let locals = namespace(py, c"failure = ValueError('provider')");
-        let mut logging = logged(py, &locals, true);
-        fail(py, &locals, &mut logging);
-        let result = match error {
-            None => Ok(py.None()),
-            Some(false) => Err(PyRuntimeError::new_err("handler failed")),
-            Some(true) => Err(CancelledError::new_err("cancelled")),
-        };
-        let expected = result.as_ref().err().map(|error| error.value(py).clone());
-        match logging.resume(py, result) {
-            Ok(step) => assert!(done && matches!(step, LifecycleStep::Done)),
-            Err(propagated) => {
-                assert!(!done);
-                assert!(propagated.value(py).is(expected.unwrap()));
-            }
-        }
-    });
-}
-
-#[test]
-fn closing_restores_the_correlation_context_once() {
-    Python::initialize();
-    Python::attach(|py| {
-        let locals = namespace(py, c"");
-        let mut logging = logged(py, &locals, true);
-        logging.close(py);
-        logging.close(py);
-        run(
-            py,
-            &locals,
-            c"assert logger.names() == ['restore'], logger.calls",
-        );
-    });
-}

litellm-rust/crates/core-utils/src/dot_notation_indexing.rs

@@ -0,0 +1,274 @@
+use serde_json::Value;
+
+#[derive(Clone, Debug, PartialEq, Eq)]
+enum Segment {
+    Field(String),
+    Every,
+    Index(usize),
+}
+
+fn parse_segments(path: &str) -> Option<Vec<Segment>> {
+    let mut segments = Vec::new();
+    let mut rest = path;
+    while !rest.is_empty() {
+        if let Some(after_open) = rest.strip_prefix('[') {
+            let (inside, after) = after_open.split_once(']')?;
+            segments.push(match inside {
+                "*" => Segment::Every,
+                index => Segment::Index(index.trim().parse().ok()?),
+            });
+            rest = after.strip_prefix('.').unwrap_or(after);
+            continue;
+        }
+        let end = rest.find(['.', '[']).unwrap_or(rest.len());
+        let (field, after) = rest.split_at(end);
+        if !field.is_empty() {
+            segments.push(Segment::Field(field.to_string()));
+        }
+        rest = after.strip_prefix('.').unwrap_or(after);
+    }
+    Some(segments)
+}
+
+fn without_path(value: Value, segments: &[Segment]) -> Value {
+    let Some((segment, tail)) = segments.split_first() else {
+        return value;
+    };
+    match (segment, value) {
+        (Segment::Field(name), Value::Object(object)) => Value::Object(
+            object
+                .into_iter()
+                .filter_map(|(key, item)| {
+                    if key != *name {
+                        return Some((key, item));
+                    }
+                    (!tail.is_empty()).then(|| (key, without_path(item, tail)))
+                })
+                .collect(),
+        ),
+        (Segment::Every, Value::Array(items)) => Value::Array(
+            items
+                .into_iter()
+                .map(|item| without_path(item, tail))
+                .collect(),
+        ),
+        (Segment::Index(index), Value::Array(items)) => Value::Array(
+            items
+                .into_iter()
+                .enumerate()
+                .map(|(position, item)| {
+                    if position == *index {
+                        without_path(item, tail)
+                    } else {
+                        item
+                    }
+                })
+                .collect(),
+        ),
+        (_, value) => value,
+    }
+}
+
+pub fn delete_nested_value(value: Value, path: &str) -> Value {
+    match parse_segments(path) {
+        Some(segments) => without_path(value, &segments),
+        None => value,
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use rstest::{fixture, rstest};
+    use serde_json::json;
+
+    use super::*;
+
+    #[fixture]
+    fn body() -> Value {
+        json!({
+            "tools": [
+                {"name": "t0", "examples": ["a"], "arr": [{"f": 1, "k": 1}, {"f": 2, "k": 2}]},
+                {"name": "t1", "examples": ["b"], "arr": [{"f": 3, "k": 3}]}
+            ],
+            "meta": {"user": "u", "inner": {"drop": 1, "keep": 2}},
+            "top": 0.7
+        })
+    }
+
+    #[rstest]
+    #[case::top_level_field("top", json!({
+        "tools": [
+            {"name": "t0", "examples": ["a"], "arr": [{"f": 1, "k": 1}, {"f": 2, "k": 2}]},
+            {"name": "t1", "examples": ["b"], "arr": [{"f": 3, "k": 3}]}
+        ],
+        "meta": {"user": "u", "inner": {"drop": 1, "keep": 2}}
+    }))]
+    #[case::whole_object("meta", json!({
+        "tools": [
+            {"name": "t0", "examples": ["a"], "arr": [{"f": 1, "k": 1}, {"f": 2, "k": 2}]},
+            {"name": "t1", "examples": ["b"], "arr": [{"f": 3, "k": 3}]}
+        ],
+        "top": 0.7
+    }))]
+    #[case::nested_field("meta.inner.drop", json!({
+        "tools": [
+            {"name": "t0", "examples": ["a"], "arr": [{"f": 1, "k": 1}, {"f": 2, "k": 2}]},
+            {"name": "t1", "examples": ["b"], "arr": [{"f": 3, "k": 3}]}
+        ],
+        "meta": {"user": "u", "inner": {"keep": 2}},
+        "top": 0.7
+    }))]
+    #[case::trailing_dot("meta.inner.drop.", json!({
+        "tools": [
+            {"name": "t0", "examples": ["a"], "arr": [{"f": 1, "k": 1}, {"f": 2, "k": 2}]},
+            {"name": "t1", "examples": ["b"], "arr": [{"f": 3, "k": 3}]}
+        ],
+        "meta": {"user": "u", "inner": {"keep": 2}},
+        "top": 0.7
+    }))]
+    #[case::leading_and_doubled_dots(".meta..inner.drop", json!({
+        "tools": [
+            {"name": "t0", "examples": ["a"], "arr": [{"f": 1, "k": 1}, {"f": 2, "k": 2}]},
+            {"name": "t1", "examples": ["b"], "arr": [{"f": 3, "k": 3}]}
+        ],
+        "meta": {"user": "u", "inner": {"keep": 2}},
+        "top": 0.7
+    }))]
+    #[case::field_in_every_element("tools[*].examples", json!({
+        "tools": [
+            {"name": "t0", "arr": [{"f": 1, "k": 1}, {"f": 2, "k": 2}]},
+            {"name": "t1", "arr": [{"f": 3, "k": 3}]}
+        ],
+        "meta": {"user": "u", "inner": {"drop": 1, "keep": 2}},
+        "top": 0.7
+    }))]
+    #[case::whole_array_field_in_every_element("tools[*].arr", json!({
+        "tools": [
+            {"name": "t0", "examples": ["a"]},
+            {"name": "t1", "examples": ["b"]}
+        ],
+        "meta": {"user": "u", "inner": {"drop": 1, "keep": 2}},
+        "top": 0.7
+    }))]
+    #[case::field_in_indexed_element("tools[1].examples", json!({
+        "tools": [
+            {"name": "t0", "examples": ["a"], "arr": [{"f": 1, "k": 1}, {"f": 2, "k": 2}]},
+            {"name": "t1", "arr": [{"f": 3, "k": 3}]}
+        ],
+        "meta": {"user": "u", "inner": {"drop": 1, "keep": 2}},
+        "top": 0.7
+    }))]
+    #[case::padded_index("tools[ 1 ].examples", json!({
+        "tools": [
+            {"name": "t0", "examples": ["a"], "arr": [{"f": 1, "k": 1}, {"f": 2, "k": 2}]},
+            {"name": "t1", "arr": [{"f": 3, "k": 3}]}
+        ],
+        "meta": {"user": "u", "inner": {"drop": 1, "keep": 2}},
+        "top": 0.7
+    }))]
+    #[case::field_right_after_bracket("tools[0]examples", json!({
+        "tools": [
+            {"name": "t0", "arr": [{"f": 1, "k": 1}, {"f": 2, "k": 2}]},
+            {"name": "t1", "examples": ["b"], "arr": [{"f": 3, "k": 3}]}
+        ],
+        "meta": {"user": "u", "inner": {"drop": 1, "keep": 2}},
+        "top": 0.7
+    }))]
+    #[case::index_then_wildcard("tools[0].arr[*].f", json!({
+        "tools": [
+            {"name": "t0", "examples": ["a"], "arr": [{"k": 1}, {"k": 2}]},
+            {"name": "t1", "examples": ["b"], "arr": [{"f": 3, "k": 3}]}
+        ],
+        "meta": {"user": "u", "inner": {"drop": 1, "keep": 2}},
+        "top": 0.7
+    }))]
+    #[case::wildcard_then_index_only_where_it_exists("tools[*].arr[1].f", json!({
+        "tools": [
+            {"name": "t0", "examples": ["a"], "arr": [{"f": 1, "k": 1}, {"k": 2}]},
+            {"name": "t1", "examples": ["b"], "arr": [{"f": 3, "k": 3}]}
+        ],
+        "meta": {"user": "u", "inner": {"drop": 1, "keep": 2}},
+        "top": 0.7
+    }))]
+    #[case::nested_wildcards("tools[*].arr[*].f", json!({
+        "tools": [
+            {"name": "t0", "examples": ["a"], "arr": [{"k": 1}, {"k": 2}]},
+            {"name": "t1", "examples": ["b"], "arr": [{"k": 3}]}
+        ],
+        "meta": {"user": "u", "inner": {"drop": 1, "keep": 2}},
+        "top": 0.7
+    }))]
+    fn deletes_the_addressed_field(body: Value, #[case] path: &str, #[case] expected: Value) {
+        assert_eq!(delete_nested_value(body, path), expected);
+    }
+
+    #[rstest]
+    #[case::empty_path("")]
+    #[case::missing_field("missing")]
+    #[case::missing_parent("missing.field")]
+    #[case::field_through_a_scalar("top.value")]
+    #[case::field_on_an_array("tools.name")]
+    #[case::index_on_an_object("meta[0].user")]
+    #[case::wildcard_on_an_object("meta[*].user")]
+    #[case::wildcard_over_scalars("tools[*].examples[*].name")]
+    #[case::index_out_of_range("tools[5].name")]
+    #[case::every_element_itself("tools[*]")]
+    #[case::indexed_element_itself("tools[0]")]
+    #[case::nested_element_itself("tools[*].arr[0]")]
+    #[case::negative_index("tools[-1].name")]
+    #[case::non_numeric_index("tools[x].name")]
+    #[case::empty_index("tools[].name")]
+    #[case::unclosed_bracket("top[0")]
+    fn leaves_the_value_untouched(body: Value, #[case] path: &str) {
+        assert_eq!(delete_nested_value(body.clone(), path), body);
+    }
+
+    #[rstest]
+    #[case::wildcards_indices_and_nesting(
+        json!({"tools": [
+            {"name": "t0", "configs": [{"id": "c0", "remove_me": 1, "keep": 1}, {"id": "c1", "remove_me": 2, "keep": 2}], "metadata": {"drop_this": 1, "preserve": 1}},
+            {"name": "t1", "configs": [{"id": "c0", "remove_me": 3, "keep": 3}, {"id": "c1", "remove_me": 4, "keep": 4}], "metadata": {"drop_this": 2, "preserve": 2}},
+            {"name": "t2", "configs": [{"id": "c0", "remove_me": 5, "keep": 5}], "metadata": {"drop_this": 3, "preserve": 3}}
+        ]}),
+        &["tools[*].configs[1].remove_me", "tools[1].metadata.drop_this", "tools[*].configs[*].id"],
+        json!({"tools": [
+            {"name": "t0", "configs": [{"remove_me": 1, "keep": 1}, {"keep": 2}], "metadata": {"drop_this": 1, "preserve": 1}},
+            {"name": "t1", "configs": [{"remove_me": 3, "keep": 3}, {"keep": 4}], "metadata": {"preserve": 2}},
+            {"name": "t2", "configs": [{"remove_me": 5, "keep": 5}], "metadata": {"drop_this": 3, "preserve": 3}}
+        ]}),
+    )]
+    #[case::simple_and_wildcard_nesting(
+        json!({
+            "tools": [{"name": "t1", "simple_nested": {"remove": 1, "keep": 2}, "complex": [{"nested": {"remove": 3, "keep": 4}}]}],
+            "top_level_remove": "should_go",
+            "top_level_keep": "should_stay"
+        }),
+        &["tools[*].simple_nested.remove", "tools[*].complex[*].nested.remove"],
+        json!({
+            "tools": [{"name": "t1", "simple_nested": {"keep": 2}, "complex": [{"nested": {"keep": 4}}]}],
+            "top_level_remove": "should_go",
+            "top_level_keep": "should_stay"
+        }),
+    )]
+    #[case::triple_nested_wildcards(
+        json!({"tools": [{"name": "t1", "arr1": [
+            {"arr2": [{"field": 1, "keep": 1}, {"field": 2, "keep": 2}]},
+            {"arr2": [{"field": 3, "keep": 3}]}
+        ]}]}),
+        &["tools[*].arr1[*].arr2[*].field"],
+        json!({"tools": [{"name": "t1", "arr1": [
+            {"arr2": [{"keep": 1}, {"keep": 2}]},
+            {"arr2": [{"keep": 3}]}
+        ]}]}),
+    )]
+    fn applies_paths_in_sequence(
+        #[case] value: Value,
+        #[case] paths: &[&str],
+        #[case] expected: Value,
+    ) {
+        let deleted = paths
+            .iter()
+            .fold(value, |value, path| delete_nested_value(value, path));
+        assert_eq!(deleted, expected);
+    }
+}

litellm-rust/crates/core-utils/src/get_provider_specific_headers.rs

@@ -0,0 +1,93 @@
+use litellm_types::utils::{ProviderSpecificHeader, ProviderSpecificHeaders};
+use serde_json::{Map, Value};
+
+pub fn get_provider_specific_headers(
+    provider_specific_header: Option<&ProviderSpecificHeaders>,
+    custom_llm_provider: &str,
+) -> Map<String, Value> {
+    let entries: &[ProviderSpecificHeader] = match provider_specific_header {
+        None => &[],
+        Some(ProviderSpecificHeaders::One(entry)) => std::slice::from_ref(entry),
+        Some(ProviderSpecificHeaders::Many(entries)) => entries,
+    };
+    entries
+        .iter()
+        .filter(|entry| {
+            entry
+                .custom_llm_provider
+                .split(',')
+                .any(|scoped| scoped.trim() == custom_llm_provider)
+        })
+        .flat_map(|entry| entry.extra_headers.clone())
+        .collect()
+}
+
+#[cfg(test)]
+mod tests {
+    use rstest::rstest;
+    use serde_json::json;
+
+    use super::*;
+
+    #[rstest]
+    #[case::single_entry_for_the_provider(
+        json!({"custom_llm_provider": "anthropic", "extra_headers": {"Authorization": "Bearer t", "Custom-Header": "v"}}),
+        json!({"Authorization": "Bearer t", "Custom-Header": "v"}),
+    )]
+    #[case::single_entry_for_another_provider(
+        json!({"custom_llm_provider": "openai", "extra_headers": {"Authorization": "Bearer t"}}),
+        json!({}),
+    )]
+    #[case::provider_in_a_comma_separated_scope(
+        json!({"custom_llm_provider": "bedrock,anthropic,vertex_ai", "extra_headers": {"anthropic-beta": "context-1m-2025-08-07"}}),
+        json!({"anthropic-beta": "context-1m-2025-08-07"}),
+    )]
+    #[case::provider_missing_from_a_comma_separated_scope(
+        json!({"custom_llm_provider": "bedrock,vertex_ai", "extra_headers": {"anthropic-beta": "test"}}),
+        json!({}),
+    )]
+    #[case::scope_with_spaces(
+        json!({"custom_llm_provider": "bedrock, anthropic , vertex_ai", "extra_headers": {"anthropic-beta": "test"}}),
+        json!({"anthropic-beta": "test"}),
+    )]
+    #[case::scope_names_must_match_exactly(
+        json!({"custom_llm_provider": "anthropic_text", "extra_headers": {"anthropic-beta": "test"}}),
+        json!({}),
+    )]
+    #[case::entries_scope_independently(
+        json!([
+            {"custom_llm_provider": "anthropic,bedrock,vertex_ai", "extra_headers": {"anthropic-beta": "context-1m-2025-08-07"}},
+            {"custom_llm_provider": "bedrock", "extra_headers": {"x-bedrock-only": "no"}},
+            {"custom_llm_provider": "anthropic", "extra_headers": {"authorization": "Bearer sk-ant-oat01-fake-token"}}
+        ]),
+        json!({"anthropic-beta": "context-1m-2025-08-07", "authorization": "Bearer sk-ant-oat01-fake-token"}),
+    )]
+    #[case::later_entries_win(
+        json!([
+            {"custom_llm_provider": "anthropic", "extra_headers": {"x-scoped": "first"}},
+            {"custom_llm_provider": "anthropic", "extra_headers": {"x-scoped": "second"}}
+        ]),
+        json!({"x-scoped": "second"}),
+    )]
+    #[case::empty_list(json!([]), json!({}))]
+    #[case::entry_without_scope(json!({"extra_headers": {"x-scoped": "yes"}}), json!({}))]
+    #[case::entry_without_headers(json!({"custom_llm_provider": "anthropic"}), json!({}))]
+    fn provider_specific_headers_match_the_scoped_provider(
+        #[case] configured: Value,
+        #[case] expected: Value,
+    ) {
+        let configured: ProviderSpecificHeaders = serde_json::from_value(configured).unwrap();
+        assert_eq!(
+            Value::Object(get_provider_specific_headers(
+                Some(&configured),
+                "anthropic"
+            )),
+            expected
+        );
+    }
+
+    #[test]
+    fn no_configured_headers_match_nothing() {
+        assert_eq!(get_provider_specific_headers(None, "anthropic"), Map::new());
+    }
+}

litellm-rust/crates/core-utils/src/lib.rs

@@ -1,7 +1,9 @@
 pub mod call_arguments;
 pub mod core_helpers;
+pub mod dot_notation_indexing;
 pub mod exception_mapping_utils;
 pub mod get_llm_provider_logic;
+pub mod get_provider_specific_headers;
 pub mod params;
 pub mod prompt_templates;
 pub mod secret_redaction;

litellm-rust/crates/core/Cargo.toml

@@ -4,9 +4,9 @@ version = "0.1.0"
 edition.workspace = true
 license.workspace = true
 repository.workspace = true
-autotests = false
 
 [dependencies]
+litellm-secrets.workspace = true
 litellm-types.workspace = true
 litellm-core-utils.workspace = true
 litellm-host.workspace = true
@@ -36,8 +36,8 @@ url.workspace = true
 veil.workspace = true
 
 [dev-dependencies]
-litellm-secrets.workspace = true
 litellm-auth-gcp.workspace = true
 litellm-llms = { workspace = true, features = ["test-support"] }
 rstest.workspace = true
 rstest_reuse.workspace = true
+wiremock = "0.6.5"

litellm-rust/crates/core/src/audio_transcription/mod.rs

@@ -14,6 +14,3 @@ pub async fn audio_transcription(request: AudioTranscriptionRequest<'_>) -> Resu
     execute_audio_transcription_provider_call(prepare_audio_transcription_provider_call(request)?)
         .await
 }
-
-#[cfg(test)]
-mod tests;

litellm-rust/crates/core/src/audio_transcription/tests.rs

@@ -1,52 +0,0 @@
-use std::{
-    io::{Read, Write},
-    net::TcpListener,
-    thread,
-};
-
-use serde_json::{Map, json};
-
-use super::audio_transcription;
-use crate::audio_transcription::types::AudioTranscriptionRequest;
-
-#[tokio::test]
-async fn bedrock_request_is_signed_and_contains_audio() {
-    let listener = TcpListener::bind("127.0.0.1:0").expect("listener");
-    let address = listener.local_addr().expect("address");
-    let server = thread::spawn(move || {
-        let (mut stream, _) = listener.accept().expect("connection");
-        let mut request = Vec::new();
-        let mut buffer = [0_u8; 16_384];
-        let count = stream.read(&mut buffer).expect("request");
-        request.extend_from_slice(&buffer[..count]);
-        let request = String::from_utf8_lossy(&request);
-        assert!(request.contains("POST /model/mistral.voxtral-mini-3b-2507/converse"));
-        assert!(request.contains("authorization: AWS4-HMAC-SHA256"));
-        assert!(request.contains("x-amz-date:"));
-        assert!(request.contains("\"bytes\":\"AQI=\""));
-        assert!(request.contains("Transcribe the audio. Respond with only the transcript."));
-        let response = b"HTTP/1.1 200 OK\r\nContent-Type: application/json\r\nContent-Length: 53\r\nConnection: close\r\n\r\n{\"output\":{\"message\":{\"content\":[{\"text\":\"hello\"}]}}}";
-        stream.write_all(response).expect("response");
-    });
-
-    let optional_params = Map::from_iter([
-        ("aws_access_key_id".to_string(), json!("access-key")),
-        ("aws_secret_access_key".to_string(), json!("secret-key")),
-        ("aws_region_name".to_string(), json!("us-east-1")),
-    ]);
-    let api_base = format!("http://{address}");
-    let response = audio_transcription(AudioTranscriptionRequest {
-        model: "mistral.voxtral-mini-3b-2507",
-        audio: json!({"data": "AQI=", "format": "wav", "filename": "audio.wav"}),
-        api_key: None,
-        api_base: Some(&api_base),
-        custom_llm_provider: Some("bedrock"),
-        extra_headers: None,
-        optional_params,
-        timeout: None,
-    })
-    .await
-    .expect("transcription");
-    assert_eq!(response, json!({"text": "hello"}));
-    server.join().expect("server");
-}

litellm-rust/crates/core/src/chat_completions/handler.rs

@@ -85,3 +85,29 @@ pub(super) async fn outbound_request(
         other => other,
     })
 }
+
+#[cfg(test)]
+mod tests {
+    use super::{Error, as_response_error};
+
+    #[test]
+    fn response_errors_collapse_to_one_variant_that_can_only_mean_already_sent() {
+        for original in [
+            Error::MissingField("usage"),
+            Error::Unsupported("non-text response content block"),
+            Error::InvalidRequest("whatever".to_string()),
+            Error::Auth(litellm_auth::Error::InvalidHeader),
+        ] {
+            let label = format!("{original:?}");
+            assert!(
+                matches!(as_response_error(original), Error::InvalidResponse(_)),
+                "{label} must not stay retryable once the provider has answered"
+            );
+        }
+        let upstream = Error::Transport(litellm_http::transport::Error::Http {
+            status: 500,
+            body: "boom".to_string(),
+        });
+        assert_eq!(as_response_error(upstream.clone()), upstream);
+    }
+}

litellm-rust/crates/core/src/chat_completions/mod.rs

@@ -51,6 +51,3 @@ pub fn chat_completions_decline_reason(
         .unsupported_reason(&messages, optional_params)
         .map(|reason| reason.0)
 }
-
-#[cfg(test)]
-mod tests;

litellm-rust/crates/core/src/chat_completions/prepare.rs

@@ -143,3 +143,597 @@ pub(super) fn prepare_provider_request(
         timeout: request.timeout,
     })
 }
+
+#[cfg(test)]
+mod tests {
+    use litellm_llms::base_llm::chat::transformation::RequestAuth;
+    use serde_json::{Map, Value, json};
+
+    use super::{prepare_provider_request, resolve_request};
+    use crate::chat_completions::{
+        Error,
+        types::{ChatCompletionsRequest, ProviderChatCompletionsRequest},
+    };
+
+    fn prepare_chat_completions_call(
+        request: ChatCompletionsRequest<'_>,
+    ) -> Result<ProviderChatCompletionsRequest, Error> {
+        prepare_provider_request(resolve_request(request)?)
+    }
+
+    fn request<'a>(
+        model: &'a str,
+        provider: Option<&'a str>,
+        messages: Value,
+        optional_params: Value,
+    ) -> ChatCompletionsRequest<'a> {
+        ChatCompletionsRequest {
+            model,
+            messages,
+            optional_params: match optional_params {
+                Value::Object(map) => map,
+                other => panic!("params must be an object, got {other}"),
+            },
+            api_key: Some("sk-test"),
+            api_base: None,
+            custom_llm_provider: provider,
+            extra_headers: None,
+            timeout: None,
+        }
+    }
+
+    /// `ProviderChatCompletionsRequest` deliberately has no `Debug` (its headers
+    /// carry resolved credentials), so unwrap the failure case by hand.
+    fn decline(request: ChatCompletionsRequest<'_>) -> Error {
+        match prepare_chat_completions_call(request) {
+            Err(error) => error,
+            Ok(prepared) => panic!("expected a decline, prepared a call to {}", prepared.url),
+        }
+    }
+
+    #[test]
+    fn resolves_the_provider_from_the_model_prefix() {
+        let prepared = prepare_chat_completions_call(request(
+            "anthropic/claude-sonnet-4-5",
+            None,
+            json!([{"role": "user", "content": "hi"}]),
+            json!({"max_tokens": 16}),
+        ))
+        .expect("prepares");
+        assert_eq!(prepared.model, "claude-sonnet-4-5");
+        assert_eq!(prepared.url, "https://api.anthropic.com/v1/messages");
+        assert_eq!(prepared.body["model"], json!("claude-sonnet-4-5"));
+    }
+
+    #[test]
+    fn strips_an_explicit_provider_prefix_from_the_model() {
+        let prepared = prepare_chat_completions_call(request(
+            "anthropic/claude-sonnet-4-5",
+            Some("anthropic"),
+            json!([{"role": "user", "content": "hi"}]),
+            json!({}),
+        ))
+        .expect("prepares");
+        assert_eq!(prepared.model, "claude-sonnet-4-5");
+    }
+
+    #[test]
+    fn adds_the_auth_and_default_headers() {
+        let prepared = prepare_chat_completions_call(request(
+            "claude-sonnet-4-5",
+            Some("anthropic"),
+            json!([{"role": "user", "content": "hi"}]),
+            json!({}),
+        ))
+        .expect("prepares");
+        assert!(
+            prepared
+                .upstream_headers
+                .contains(&("x-api-key".to_string(), "sk-test".to_string()))
+        );
+        assert!(
+            prepared
+                .upstream_headers
+                .contains(&("anthropic-version".to_string(), "2023-06-01".to_string()))
+        );
+        assert!(matches!(
+            prepared.auth,
+            RequestAuth::Header {
+                name: "x-api-key",
+                ..
+            }
+        ));
+    }
+
+    #[test]
+    fn the_deployment_credential_replaces_a_caller_supplied_auth_header() {
+        // Python builds `{**headers, **anthropic_headers}`, so the deployment's key
+        // overwrites a forwarded one. Honouring the caller's would let whoever sends
+        // the request choose the Anthropic principal it bills to.
+        let mut call = request(
+            "claude-sonnet-4-5",
+            Some("anthropic"),
+            json!([{"role": "user", "content": "hi"}]),
+            json!({}),
+        );
+        call.extra_headers = Some(Map::from_iter([(
+            "X-Api-Key".to_string(),
+            json!("sk-caller"),
+        )]));
+        let prepared = prepare_chat_completions_call(call).expect("prepares");
+        let keys: Vec<_> = prepared
+            .upstream_headers
+            .iter()
+            .filter(|(name, _)| name.eq_ignore_ascii_case("x-api-key"))
+            .collect();
+        assert_eq!(keys.len(), 1, "got {:?}", prepared.upstream_headers);
+        assert_eq!(keys[0].1, "sk-test");
+    }
+
+    #[test]
+    fn a_forwarded_authorization_header_suppresses_the_resolved_api_key_header() {
+        // Anthropic's `validate_environment` pops `x-api-key` and sets `authorization`
+        // for an OAuth token, so re-adding the key here would put the credential into
+        // a header the host removed on purpose.
+        let mut call = request(
+            "claude-sonnet-4-5",
+            Some("anthropic"),
+            json!([{"role": "user", "content": "hi"}]),
+            json!({}),
+        );
+        call.extra_headers = Some(Map::from_iter([
+            (
+                "Authorization".to_string(),
+                json!("Bearer sk-ant-oat01-token"),
+            ),
+            ("X-Api-Key".to_string(), json!("sk-caller")),
+        ]));
+        let prepared = prepare_chat_completions_call(call).expect("prepares");
+        assert!(
+            !prepared
+                .upstream_headers
+                .iter()
+                .any(|(name, value)| name.eq_ignore_ascii_case("x-api-key") && value == "sk-test"),
+            "the resolved key must not be applied over an OAuth bearer, got {:?}",
+            prepared.upstream_headers
+        );
+        assert!(
+            prepared
+                .upstream_headers
+                .iter()
+                .any(|(name, value)| name.eq_ignore_ascii_case("authorization")
+                    && value == "Bearer sk-ant-oat01-token")
+        );
+    }
+
+    #[test]
+    fn an_unrelated_forwarded_authorization_does_not_defer_the_resolved_key() {
+        // Only an OAuth bearer replaces the credential. Python sends the deployment's
+        // `x-api-key` alongside any other forwarded `authorization`, so deferring on
+        // the mere presence of that header would drop the deployment's auth.
+        let mut call = request(
+            "claude-sonnet-4-5",
+            Some("anthropic"),
+            json!([{"role": "user", "content": "hi"}]),
+            json!({}),
+        );
+        call.extra_headers = Some(Map::from_iter([
+            ("Authorization".to_string(), json!("Bearer unrelated")),
+            ("X-Api-Key".to_string(), json!("sk-caller")),
+        ]));
+        let prepared = prepare_chat_completions_call(call).expect("prepares");
+        let keys: Vec<_> = prepared
+            .upstream_headers
+            .iter()
+            .filter(|(name, _)| name.eq_ignore_ascii_case("x-api-key"))
+            .collect();
+        assert_eq!(keys.len(), 1, "got {:?}", prepared.upstream_headers);
+        assert_eq!(keys[0].1, "sk-test");
+        assert!(
+            prepared
+                .upstream_headers
+                .iter()
+                .any(|(name, value)| name.eq_ignore_ascii_case("authorization")
+                    && value == "Bearer unrelated"),
+            "the unrelated authorization must survive, got {:?}",
+            prepared.upstream_headers
+        );
+    }
+
+    #[test]
+    fn declines_an_unsupported_request_before_resolving_credentials() {
+        let mut call = request(
+            "claude-sonnet-4-5",
+            Some("anthropic"),
+            json!([{"role": "user", "content": "hi"}]),
+            json!({"stream": true}),
+        );
+        call.api_key = None;
+        // No api_key is set and no env is consulted: the gate must run first, so the
+        // error is the decline rather than a missing-credential error.
+        assert_eq!(decline(call), Error::Unsupported("streaming"));
+    }
+
+    #[test]
+    fn rejects_an_unknown_provider() {
+        assert_eq!(
+            decline(request(
+                "openai/gpt-4o",
+                None,
+                json!([{"role": "user", "content": "hi"}]),
+                json!({}),
+            )),
+            Error::InvalidProvider("openai".to_string())
+        );
+    }
+
+    #[test]
+    fn rejects_a_model_with_no_resolvable_provider() {
+        assert!(matches!(
+            decline(request(
+                "claude-sonnet-4-5",
+                None,
+                json!([{"role": "user", "content": "hi"}]),
+                json!({}),
+            )),
+            Error::InvalidProvider(_)
+        ));
+    }
+
+    #[test]
+    fn rejects_an_empty_or_malformed_message_list() {
+        assert_eq!(
+            decline(request(
+                "anthropic/claude-sonnet-4-5",
+                None,
+                json!([]),
+                json!({}),
+            )),
+            Error::InvalidRequest("chat completions requires at least one message".to_string())
+        );
+        assert!(matches!(
+            decline(request(
+                "anthropic/claude-sonnet-4-5",
+                None,
+                json!("not a list"),
+                json!({}),
+            )),
+            Error::InvalidRequest(_)
+        ));
+    }
+
+    #[test]
+    fn rejects_non_string_extra_headers() {
+        let mut call = request(
+            "anthropic/claude-sonnet-4-5",
+            None,
+            json!([{"role": "user", "content": "hi"}]),
+            json!({}),
+        );
+        call.extra_headers = Some(Map::from_iter([("x-trace".to_string(), json!(7))]));
+        assert_eq!(
+            decline(call),
+            Error::Headers(litellm_http::request::HeaderError {
+                context: "chat completions",
+                name: "x-trace".to_string(),
+                actual: "number",
+            })
+        );
+    }
+
+    #[test]
+    fn prepares_a_bedrock_call_without_resolving_credentials() {
+        let mut call = request(
+            "bedrock/us-east-1/anthropic.claude-v2",
+            None,
+            json!([{"role": "user", "content": "hi"}]),
+            json!({"maxTokens": 16}),
+        );
+        call.api_key = None;
+        let prepared = prepare_chat_completions_call(call).expect("prepares");
+        assert_eq!(
+            prepared.url,
+            "https://bedrock-runtime.us-east-1.amazonaws.com/model/anthropic.claude-v2/converse"
+        );
+        assert_eq!(
+            prepared.auth,
+            RequestAuth::AwsSigV4 {
+                region: "us-east-1".to_string(),
+                service: "bedrock",
+            }
+        );
+        // SigV4 signs the serialized body, so prepare must not have added an
+        // Authorization header; the handler does it.
+        assert!(
+            !prepared
+                .upstream_headers
+                .iter()
+                .any(|(name, _)| name.eq_ignore_ascii_case("authorization"))
+        );
+        assert_eq!(prepared.body["inferenceConfig"], json!({"maxTokens": 16}));
+    }
+
+    #[tokio::test]
+    async fn a_forwarded_client_header_does_not_enter_the_bedrock_signature() {
+        // Python signs only the AWS header set and reattaches the rest, so a header
+        // the caller forwarded rides along without joining the canonical request.
+        // Signing it makes Converse 403 on a deployment that works on Python.
+        let mut call = request(
+            "bedrock/us-east-1/anthropic.claude-v2",
+            None,
+            json!([{"role": "user", "content": "hi"}]),
+            json!({
+                "maxTokens": 16,
+                "aws_access_key_id": "AKIDEXAMPLE",
+                "aws_secret_access_key": "wJalrXUtnFEMI/K7MDENG+bPxRfiCYEXAMPLEKEY"
+            }),
+        );
+        // A key would resolve to a bearer token and never reach the signer.
+        call.api_key = None;
+        call.extra_headers = Some(Map::from_iter([(
+            "x-request-id".to_string(),
+            json!("abc-123"),
+        )]));
+        let prepared = prepare_chat_completions_call(call).expect("prepares");
+        let signed = crate::chat_completions::handler::outbound_request(&prepared)
+            .await
+            .expect("signs");
+
+        let authorization = signed
+            .header("authorization")
+            .expect("carries an authorization header")
+            .to_string();
+        assert!(
+            authorization.starts_with("AWS4-HMAC-SHA256"),
+            "expected a SigV4 signature, got {authorization}"
+        );
+        assert!(
+            !authorization.contains("x-request-id"),
+            "forwarded header reached SignedHeaders: {authorization}"
+        );
+        // It still goes on the wire, it is just not part of the signature.
+        assert!(
+            signed
+                .headers()
+                .iter()
+                .any(|(name, value)| name == "x-request-id" && value == "abc-123"),
+            "forwarded header was dropped instead of reattached"
+        );
+    }
+
+    #[tokio::test]
+    async fn a_forwarded_header_the_signer_computes_declines_to_python() {
+        // Reattaching the caller's copy next to the computed one puts the name on
+        // the wire twice and Bedrock rejects the pair, so a request carrying one
+        // has to go to Python instead of being signed here.
+        for forwarded in [
+            "Authorization",
+            "x-amz-date",
+            "x-amz-security-token",
+            "Date",
+        ] {
+            let mut call = request(
+                "bedrock/us-east-1/anthropic.claude-v2",
+                None,
+                json!([{"role": "user", "content": "hi"}]),
+                json!({
+                    "maxTokens": 16,
+                    "aws_access_key_id": "AKIDEXAMPLE",
+                    "aws_secret_access_key": "wJalrXUtnFEMI/K7MDENG+bPxRfiCYEXAMPLEKEY"
+                }),
+            );
+            call.api_key = None;
+            call.extra_headers = Some(Map::from_iter([(forwarded.to_string(), json!("forged"))]));
+            let prepared = prepare_chat_completions_call(call).expect("prepares");
+            let error = crate::chat_completions::handler::outbound_request(&prepared)
+                .await
+                .expect_err("{forwarded} should decline instead of being signed");
+            assert!(
+                matches!(error, Error::Unsupported(_)),
+                "{forwarded} declined as {error:?}, which the host would not fall back on"
+            );
+        }
+    }
+
+    #[test]
+    fn a_bedrock_deployment_bearer_outranks_a_forwarded_authorization() {
+        // `get_request_headers` assigns `headers["Authorization"]` unconditionally
+        // once a bearer token resolves, so the deployment's identity wins on
+        // Python. Keeping the caller's would authorize and bill the call as a
+        // different principal, and only when the deployment carries `rust: true`.
+        let mut call = request(
+            "bedrock/us-east-1/anthropic.claude-v2",
+            None,
+            json!([{"role": "user", "content": "hi"}]),
+            json!({"maxTokens": 16}),
+        );
+        call.extra_headers = Some(Map::from_iter([(
+            "Authorization".to_string(),
+            json!("Bearer caller-supplied"),
+        )]));
+        let prepared = prepare_chat_completions_call(call).expect("prepares");
+        let authorizations: Vec<_> = prepared
+            .upstream_headers
+            .iter()
+            .filter(|(name, _)| name.eq_ignore_ascii_case("authorization"))
+            .map(|(_, value)| value.as_str())
+            .collect();
+        assert_eq!(
+            authorizations,
+            vec!["Bearer sk-test"],
+            "the deployment token must be the only authorization on the wire"
+        );
+    }
+
+    #[test]
+    fn an_anthropic_forwarded_oauth_bearer_still_outranks_the_resolved_key() {
+        // The opposite precedence, and deliberate: Anthropic's own transform
+        // honours a forwarded OAuth bearer, so the Bedrock fix above must not be
+        // generalized into a rule that the configured key always wins.
+        //
+        // An OAuth bearer is the whole of that exception. This forwarded a plain
+        // `x-api-key` until round 17, which read as the same claim and was not:
+        // Python overwrites a forwarded `x-api-key` with the deployment's.
+        let mut call = request(
+            "claude-sonnet-4-5",
+            Some("anthropic"),
+            json!([{"role": "user", "content": "hi"}]),
+            json!({}),
+        );
+        call.extra_headers = Some(Map::from_iter([(
+            "authorization".to_string(),
+            json!("Bearer sk-ant-oat01-forwarded"),
+        )]));
+        let prepared = prepare_chat_completions_call(call).expect("prepares");
+        let keys: Vec<_> = prepared
+            .upstream_headers
+            .iter()
+            .filter(|(name, _)| name.eq_ignore_ascii_case("x-api-key"))
+            .map(|(_, value)| value.as_str())
+            .collect();
+        assert!(keys.is_empty(), "got {:?}", prepared.upstream_headers);
+        assert!(
+            prepared
+                .upstream_headers
+                .iter()
+                .any(|(name, value)| name.eq_ignore_ascii_case("authorization")
+                    && value == "Bearer sk-ant-oat01-forwarded")
+        );
+    }
+
+    #[test]
+    fn a_bedrock_api_key_is_sent_as_a_bearer_token_instead_of_being_signed() {
+        // The configured bearer identity has its own account and quota boundary,
+        // so a request carrying one must not be signed as whatever principal the
+        // host's AWS credentials resolve to.
+        let prepared = prepare_chat_completions_call(request(
+            "bedrock/us-east-1/anthropic.claude-v2",
+            None,
+            json!([{"role": "user", "content": "hi"}]),
+            json!({"maxTokens": 16}),
+        ))
+        .expect("prepares");
+        assert_eq!(
+            prepared.auth,
+            RequestAuth::Bearer {
+                token: "sk-test".to_string()
+            }
+        );
+        assert!(
+            prepared
+                .upstream_headers
+                .iter()
+                .any(|(name, value)| name.eq_ignore_ascii_case("authorization")
+                    && value == "Bearer sk-test"),
+            "prepare did not carry the bearer token"
+        );
+    }
+
+    fn decline_reason(
+        model: &str,
+        provider: Option<&str>,
+        messages: Value,
+        params: Value,
+    ) -> Option<&'static str> {
+        let params = match params {
+            Value::Object(map) => map,
+            other => panic!("params must be an object, got {other}"),
+        };
+        crate::chat_completions::chat_completions_decline_reason(model, provider, messages, &params)
+    }
+
+    #[test]
+    fn the_gate_accepts_what_prepare_accepts() {
+        assert_eq!(
+            decline_reason(
+                "anthropic/claude-sonnet-4-5",
+                None,
+                json!([{"role": "user", "content": "hi"}]),
+                json!({"max_tokens": 16}),
+            ),
+            None
+        );
+    }
+
+    #[test]
+    fn the_gate_declines_without_resolving_credentials_or_calling_out() {
+        assert_eq!(
+            decline_reason(
+                "anthropic/claude-sonnet-4-5",
+                None,
+                json!([{"role": "user", "content": "hi"}]),
+                json!({"stream": true}),
+            ),
+            Some("streaming")
+        );
+        assert_eq!(
+            decline_reason(
+                "openai/gpt-4o",
+                None,
+                json!([{"role": "user", "content": "hi"}]),
+                json!({}),
+            ),
+            Some("provider is not on the rust chat completions path")
+        );
+        assert_eq!(
+            decline_reason(
+                "claude-sonnet-4-5",
+                None,
+                json!([{"role": "user", "content": "hi"}]),
+                json!({}),
+            ),
+            Some("provider is not on the rust chat completions path")
+        );
+        assert_eq!(
+            decline_reason(
+                "anthropic/claude-sonnet-4-5",
+                None,
+                json!("nope"),
+                json!({})
+            ),
+            Some("unreadable message list")
+        );
+        assert_eq!(
+            decline_reason("anthropic/claude-sonnet-4-5", None, json!([]), json!({})),
+            Some("empty message list")
+        );
+    }
+
+    #[test]
+    fn the_gate_agrees_with_prepare_on_every_case_it_accepts() {
+        // A gate that accepts what prepare then declines would make the host emit
+        // its pre-call logging on a path that falls back, so pin the agreement.
+        for (messages, params) in [
+            (
+                json!([{"role": "user", "content": "hi"}]),
+                json!({"max_tokens": 8}),
+            ),
+            (
+                json!([{"role": "system", "content": "s"}, {"role": "user", "content": "hi"}]),
+                json!({"temperature": 0.1}),
+            ),
+            (
+                json!([{"role": "user", "content": "hi"}, {"role": "assistant", "content": "yo"}]),
+                json!({}),
+            ),
+        ] {
+            assert_eq!(
+                decline_reason(
+                    "anthropic/claude-sonnet-4-5",
+                    None,
+                    messages.clone(),
+                    params.clone()
+                ),
+                None,
+                "gate declined {messages}"
+            );
+            prepare_chat_completions_call(request(
+                "anthropic/claude-sonnet-4-5",
+                None,
+                messages.clone(),
+                params,
+            ))
+            .unwrap_or_else(|error| panic!("prepare declined {messages}: {error}"));
+        }
+    }
+}

litellm-rust/crates/core/src/chat_completions/tests.rs

@@ -1,833 +0,0 @@
-use litellm_llms::base_llm::chat::transformation::RequestAuth;
-use serde_json::{Map, Value, json};
-
-use super::{
-    Error,
-    prepare::{prepare_provider_request, resolve_request},
-};
-use crate::chat_completions::types::{ChatCompletionsRequest, ProviderChatCompletionsRequest};
-
-fn prepare_chat_completions_call(
-    request: ChatCompletionsRequest<'_>,
-) -> Result<ProviderChatCompletionsRequest, Error> {
-    prepare_provider_request(resolve_request(request)?)
-}
-
-fn request<'a>(
-    model: &'a str,
-    provider: Option<&'a str>,
-    messages: Value,
-    optional_params: Value,
-) -> ChatCompletionsRequest<'a> {
-    ChatCompletionsRequest {
-        model,
-        messages,
-        optional_params: match optional_params {
-            Value::Object(map) => map,
-            other => panic!("params must be an object, got {other}"),
-        },
-        api_key: Some("sk-test"),
-        api_base: None,
-        custom_llm_provider: provider,
-        extra_headers: None,
-        timeout: None,
-    }
-}
-
-/// `ProviderChatCompletionsRequest` deliberately has no `Debug` (its headers
-/// carry resolved credentials), so unwrap the failure case by hand.
-fn decline(request: ChatCompletionsRequest<'_>) -> Error {
-    match prepare_chat_completions_call(request) {
-        Err(error) => error,
-        Ok(prepared) => panic!("expected a decline, prepared a call to {}", prepared.url),
-    }
-}
-
-#[test]
-fn resolves_the_provider_from_the_model_prefix() {
-    let prepared = prepare_chat_completions_call(request(
-        "anthropic/claude-sonnet-4-5",
-        None,
-        json!([{"role": "user", "content": "hi"}]),
-        json!({"max_tokens": 16}),
-    ))
-    .expect("prepares");
-    assert_eq!(prepared.model, "claude-sonnet-4-5");
-    assert_eq!(prepared.url, "https://api.anthropic.com/v1/messages");
-    assert_eq!(prepared.body["model"], json!("claude-sonnet-4-5"));
-}
-
-#[test]
-fn strips_an_explicit_provider_prefix_from_the_model() {
-    let prepared = prepare_chat_completions_call(request(
-        "anthropic/claude-sonnet-4-5",
-        Some("anthropic"),
-        json!([{"role": "user", "content": "hi"}]),
-        json!({}),
-    ))
-    .expect("prepares");
-    assert_eq!(prepared.model, "claude-sonnet-4-5");
-}
-
-#[test]
-fn adds_the_auth_and_default_headers() {
-    let prepared = prepare_chat_completions_call(request(
-        "claude-sonnet-4-5",
-        Some("anthropic"),
-        json!([{"role": "user", "content": "hi"}]),
-        json!({}),
-    ))
-    .expect("prepares");
-    assert!(
-        prepared
-            .upstream_headers
-            .contains(&("x-api-key".to_string(), "sk-test".to_string()))
-    );
-    assert!(
-        prepared
-            .upstream_headers
-            .contains(&("anthropic-version".to_string(), "2023-06-01".to_string()))
-    );
-    assert!(matches!(
-        prepared.auth,
-        RequestAuth::Header {
-            name: "x-api-key",
-            ..
-        }
-    ));
-}
-
-#[test]
-fn the_deployment_credential_replaces_a_caller_supplied_auth_header() {
-    // Python builds `{**headers, **anthropic_headers}`, so the deployment's key
-    // overwrites a forwarded one. Honouring the caller's would let whoever sends
-    // the request choose the Anthropic principal it bills to.
-    let mut call = request(
-        "claude-sonnet-4-5",
-        Some("anthropic"),
-        json!([{"role": "user", "content": "hi"}]),
-        json!({}),
-    );
-    call.extra_headers = Some(Map::from_iter([(
-        "X-Api-Key".to_string(),
-        json!("sk-caller"),
-    )]));
-    let prepared = prepare_chat_completions_call(call).expect("prepares");
-    let keys: Vec<_> = prepared
-        .upstream_headers
-        .iter()
-        .filter(|(name, _)| name.eq_ignore_ascii_case("x-api-key"))
-        .collect();
-    assert_eq!(keys.len(), 1, "got {:?}", prepared.upstream_headers);
-    assert_eq!(keys[0].1, "sk-test");
-}
-
-#[test]
-fn a_forwarded_authorization_header_suppresses_the_resolved_api_key_header() {
-    // Anthropic's `validate_environment` pops `x-api-key` and sets `authorization`
-    // for an OAuth token, so re-adding the key here would put the credential into
-    // a header the host removed on purpose.
-    let mut call = request(
-        "claude-sonnet-4-5",
-        Some("anthropic"),
-        json!([{"role": "user", "content": "hi"}]),
-        json!({}),
-    );
-    call.extra_headers = Some(Map::from_iter([
-        (
-            "Authorization".to_string(),
-            json!("Bearer sk-ant-oat01-token"),
-        ),
-        ("X-Api-Key".to_string(), json!("sk-caller")),
-    ]));
-    let prepared = prepare_chat_completions_call(call).expect("prepares");
-    assert!(
-        !prepared
-            .upstream_headers
-            .iter()
-            .any(|(name, value)| name.eq_ignore_ascii_case("x-api-key") && value == "sk-test"),
-        "the resolved key must not be applied over an OAuth bearer, got {:?}",
-        prepared.upstream_headers
-    );
-    assert!(
-        prepared
-            .upstream_headers
-            .iter()
-            .any(|(name, value)| name.eq_ignore_ascii_case("authorization")
-                && value == "Bearer sk-ant-oat01-token")
-    );
-}
-
-#[test]
-fn an_unrelated_forwarded_authorization_does_not_defer_the_resolved_key() {
-    // Only an OAuth bearer replaces the credential. Python sends the deployment's
-    // `x-api-key` alongside any other forwarded `authorization`, so deferring on
-    // the mere presence of that header would drop the deployment's auth.
-    let mut call = request(
-        "claude-sonnet-4-5",
-        Some("anthropic"),
-        json!([{"role": "user", "content": "hi"}]),
-        json!({}),
-    );
-    call.extra_headers = Some(Map::from_iter([
-        ("Authorization".to_string(), json!("Bearer unrelated")),
-        ("X-Api-Key".to_string(), json!("sk-caller")),
-    ]));
-    let prepared = prepare_chat_completions_call(call).expect("prepares");
-    let keys: Vec<_> = prepared
-        .upstream_headers
-        .iter()
-        .filter(|(name, _)| name.eq_ignore_ascii_case("x-api-key"))
-        .collect();
-    assert_eq!(keys.len(), 1, "got {:?}", prepared.upstream_headers);
-    assert_eq!(keys[0].1, "sk-test");
-    assert!(
-        prepared
-            .upstream_headers
-            .iter()
-            .any(|(name, value)| name.eq_ignore_ascii_case("authorization")
-                && value == "Bearer unrelated"),
-        "the unrelated authorization must survive, got {:?}",
-        prepared.upstream_headers
-    );
-}
-
-#[test]
-fn declines_an_unsupported_request_before_resolving_credentials() {
-    let mut call = request(
-        "claude-sonnet-4-5",
-        Some("anthropic"),
-        json!([{"role": "user", "content": "hi"}]),
-        json!({"stream": true}),
-    );
-    call.api_key = None;
-    // No api_key is set and no env is consulted: the gate must run first, so the
-    // error is the decline rather than a missing-credential error.
-    assert_eq!(decline(call), Error::Unsupported("streaming"));
-}
-
-#[test]
-fn rejects_an_unknown_provider() {
-    assert_eq!(
-        decline(request(
-            "openai/gpt-4o",
-            None,
-            json!([{"role": "user", "content": "hi"}]),
-            json!({}),
-        )),
-        Error::InvalidProvider("openai".to_string())
-    );
-}
-
-#[test]
-fn rejects_a_model_with_no_resolvable_provider() {
-    assert!(matches!(
-        decline(request(
-            "claude-sonnet-4-5",
-            None,
-            json!([{"role": "user", "content": "hi"}]),
-            json!({}),
-        )),
-        Error::InvalidProvider(_)
-    ));
-}
-
-#[test]
-fn rejects_an_empty_or_malformed_message_list() {
-    assert_eq!(
-        decline(request(
-            "anthropic/claude-sonnet-4-5",
-            None,
-            json!([]),
-            json!({}),
-        )),
-        Error::InvalidRequest("chat completions requires at least one message".to_string())
-    );
-    assert!(matches!(
-        decline(request(
-            "anthropic/claude-sonnet-4-5",
-            None,
-            json!("not a list"),
-            json!({}),
-        )),
-        Error::InvalidRequest(_)
-    ));
-}
-
-#[test]
-fn rejects_non_string_extra_headers() {
-    let mut call = request(
-        "anthropic/claude-sonnet-4-5",
-        None,
-        json!([{"role": "user", "content": "hi"}]),
-        json!({}),
-    );
-    call.extra_headers = Some(Map::from_iter([("x-trace".to_string(), json!(7))]));
-    assert_eq!(
-        decline(call),
-        Error::Headers(litellm_http::request::HeaderError {
-            context: "chat completions",
-            name: "x-trace".to_string(),
-            actual: "number",
-        })
-    );
-}
-
-#[test]
-fn prepares_a_bedrock_call_without_resolving_credentials() {
-    let mut call = request(
-        "bedrock/us-east-1/anthropic.claude-v2",
-        None,
-        json!([{"role": "user", "content": "hi"}]),
-        json!({"maxTokens": 16}),
-    );
-    call.api_key = None;
-    let prepared = prepare_chat_completions_call(call).expect("prepares");
-    assert_eq!(
-        prepared.url,
-        "https://bedrock-runtime.us-east-1.amazonaws.com/model/anthropic.claude-v2/converse"
-    );
-    assert_eq!(
-        prepared.auth,
-        RequestAuth::AwsSigV4 {
-            region: "us-east-1".to_string(),
-            service: "bedrock",
-        }
-    );
-    // SigV4 signs the serialized body, so prepare must not have added an
-    // Authorization header; the handler does it.
-    assert!(
-        !prepared
-            .upstream_headers
-            .iter()
-            .any(|(name, _)| name.eq_ignore_ascii_case("authorization"))
-    );
-    assert_eq!(prepared.body["inferenceConfig"], json!({"maxTokens": 16}));
-}
-
-#[tokio::test]
-async fn a_forwarded_client_header_does_not_enter_the_bedrock_signature() {
-    // Python signs only the AWS header set and reattaches the rest, so a header
-    // the caller forwarded rides along without joining the canonical request.
-    // Signing it makes Converse 403 on a deployment that works on Python.
-    let mut call = request(
-        "bedrock/us-east-1/anthropic.claude-v2",
-        None,
-        json!([{"role": "user", "content": "hi"}]),
-        json!({
-            "maxTokens": 16,
-            "aws_access_key_id": "AKIDEXAMPLE",
-            "aws_secret_access_key": "wJalrXUtnFEMI/K7MDENG+bPxRfiCYEXAMPLEKEY"
-        }),
-    );
-    // A key would resolve to a bearer token and never reach the signer.
-    call.api_key = None;
-    call.extra_headers = Some(Map::from_iter([(
-        "x-request-id".to_string(),
-        json!("abc-123"),
-    )]));
-    let prepared = prepare_chat_completions_call(call).expect("prepares");
-    let signed = super::handler::outbound_request(&prepared)
-        .await
-        .expect("signs");
-
-    let authorization = signed
-        .header("authorization")
-        .expect("carries an authorization header")
-        .to_string();
-    assert!(
-        authorization.starts_with("AWS4-HMAC-SHA256"),
-        "expected a SigV4 signature, got {authorization}"
-    );
-    assert!(
-        !authorization.contains("x-request-id"),
-        "forwarded header reached SignedHeaders: {authorization}"
-    );
-    // It still goes on the wire, it is just not part of the signature.
-    assert!(
-        signed
-            .headers()
-            .iter()
-            .any(|(name, value)| name == "x-request-id" && value == "abc-123"),
-        "forwarded header was dropped instead of reattached"
-    );
-}
-
-#[tokio::test]
-async fn a_forwarded_header_the_signer_computes_declines_to_python() {
-    // Reattaching the caller's copy next to the computed one puts the name on
-    // the wire twice and Bedrock rejects the pair, so a request carrying one
-    // has to go to Python instead of being signed here.
-    for forwarded in [
-        "Authorization",
-        "x-amz-date",
-        "x-amz-security-token",
-        "Date",
-    ] {
-        let mut call = request(
-            "bedrock/us-east-1/anthropic.claude-v2",
-            None,
-            json!([{"role": "user", "content": "hi"}]),
-            json!({
-                "maxTokens": 16,
-                "aws_access_key_id": "AKIDEXAMPLE",
-                "aws_secret_access_key": "wJalrXUtnFEMI/K7MDENG+bPxRfiCYEXAMPLEKEY"
-            }),
-        );
-        call.api_key = None;
-        call.extra_headers = Some(Map::from_iter([(forwarded.to_string(), json!("forged"))]));
-        let prepared = prepare_chat_completions_call(call).expect("prepares");
-        let error = super::handler::outbound_request(&prepared)
-            .await
-            .expect_err("{forwarded} should decline instead of being signed");
-        assert!(
-            matches!(error, Error::Unsupported(_)),
-            "{forwarded} declined as {error:?}, which the host would not fall back on"
-        );
-    }
-}
-
-#[test]
-fn a_bedrock_deployment_bearer_outranks_a_forwarded_authorization() {
-    // `get_request_headers` assigns `headers["Authorization"]` unconditionally
-    // once a bearer token resolves, so the deployment's identity wins on
-    // Python. Keeping the caller's would authorize and bill the call as a
-    // different principal, and only when the deployment carries `rust: true`.
-    let mut call = request(
-        "bedrock/us-east-1/anthropic.claude-v2",
-        None,
-        json!([{"role": "user", "content": "hi"}]),
-        json!({"maxTokens": 16}),
-    );
-    call.extra_headers = Some(Map::from_iter([(
-        "Authorization".to_string(),
-        json!("Bearer caller-supplied"),
-    )]));
-    let prepared = prepare_chat_completions_call(call).expect("prepares");
-    let authorizations: Vec<_> = prepared
-        .upstream_headers
-        .iter()
-        .filter(|(name, _)| name.eq_ignore_ascii_case("authorization"))
-        .map(|(_, value)| value.as_str())
-        .collect();
-    assert_eq!(
-        authorizations,
-        vec!["Bearer sk-test"],
-        "the deployment token must be the only authorization on the wire"
-    );
-}
-
-#[test]
-fn an_anthropic_forwarded_oauth_bearer_still_outranks_the_resolved_key() {
-    // The opposite precedence, and deliberate: Anthropic's own transform
-    // honours a forwarded OAuth bearer, so the Bedrock fix above must not be
-    // generalized into a rule that the configured key always wins.
-    //
-    // An OAuth bearer is the whole of that exception. This forwarded a plain
-    // `x-api-key` until round 17, which read as the same claim and was not:
-    // Python overwrites a forwarded `x-api-key` with the deployment's.
-    let mut call = request(
-        "claude-sonnet-4-5",
-        Some("anthropic"),
-        json!([{"role": "user", "content": "hi"}]),
-        json!({}),
-    );
-    call.extra_headers = Some(Map::from_iter([(
-        "authorization".to_string(),
-        json!("Bearer sk-ant-oat01-forwarded"),
-    )]));
-    let prepared = prepare_chat_completions_call(call).expect("prepares");
-    let keys: Vec<_> = prepared
-        .upstream_headers
-        .iter()
-        .filter(|(name, _)| name.eq_ignore_ascii_case("x-api-key"))
-        .map(|(_, value)| value.as_str())
-        .collect();
-    assert!(keys.is_empty(), "got {:?}", prepared.upstream_headers);
-    assert!(
-        prepared
-            .upstream_headers
-            .iter()
-            .any(|(name, value)| name.eq_ignore_ascii_case("authorization")
-                && value == "Bearer sk-ant-oat01-forwarded")
-    );
-}
-
-#[test]
-fn a_bedrock_api_key_is_sent_as_a_bearer_token_instead_of_being_signed() {
-    // The configured bearer identity has its own account and quota boundary,
-    // so a request carrying one must not be signed as whatever principal the
-    // host's AWS credentials resolve to.
-    let prepared = prepare_chat_completions_call(request(
-        "bedrock/us-east-1/anthropic.claude-v2",
-        None,
-        json!([{"role": "user", "content": "hi"}]),
-        json!({"maxTokens": 16}),
-    ))
-    .expect("prepares");
-    assert_eq!(
-        prepared.auth,
-        RequestAuth::Bearer {
-            token: "sk-test".to_string()
-        }
-    );
-    assert!(
-        prepared
-            .upstream_headers
-            .iter()
-            .any(|(name, value)| name.eq_ignore_ascii_case("authorization")
-                && value == "Bearer sk-test"),
-        "prepare did not carry the bearer token"
-    );
-}
-
-fn decline_reason(
-    model: &str,
-    provider: Option<&str>,
-    messages: Value,
-    params: Value,
-) -> Option<&'static str> {
-    let params = match params {
-        Value::Object(map) => map,
-        other => panic!("params must be an object, got {other}"),
-    };
-    super::chat_completions_decline_reason(model, provider, messages, &params)
-}
-
-#[test]
-fn the_gate_accepts_what_prepare_accepts() {
-    assert_eq!(
-        decline_reason(
-            "anthropic/claude-sonnet-4-5",
-            None,
-            json!([{"role": "user", "content": "hi"}]),
-            json!({"max_tokens": 16}),
-        ),
-        None
-    );
-}
-
-#[test]
-fn the_gate_declines_without_resolving_credentials_or_calling_out() {
-    assert_eq!(
-        decline_reason(
-            "anthropic/claude-sonnet-4-5",
-            None,
-            json!([{"role": "user", "content": "hi"}]),
-            json!({"stream": true}),
-        ),
-        Some("streaming")
-    );
-    assert_eq!(
-        decline_reason(
-            "openai/gpt-4o",
-            None,
-            json!([{"role": "user", "content": "hi"}]),
-            json!({}),
-        ),
-        Some("provider is not on the rust chat completions path")
-    );
-    assert_eq!(
-        decline_reason(
-            "claude-sonnet-4-5",
-            None,
-            json!([{"role": "user", "content": "hi"}]),
-            json!({}),
-        ),
-        Some("provider is not on the rust chat completions path")
-    );
-    assert_eq!(
-        decline_reason(
-            "anthropic/claude-sonnet-4-5",
-            None,
-            json!("nope"),
-            json!({})
-        ),
-        Some("unreadable message list")
-    );
-    assert_eq!(
-        decline_reason("anthropic/claude-sonnet-4-5", None, json!([]), json!({})),
-        Some("empty message list")
-    );
-}
-
-#[test]
-fn the_gate_agrees_with_prepare_on_every_case_it_accepts() {
-    // A gate that accepts what prepare then declines would make the host emit
-    // its pre-call logging on a path that falls back, so pin the agreement.
-    for (messages, params) in [
-        (
-            json!([{"role": "user", "content": "hi"}]),
-            json!({"max_tokens": 8}),
-        ),
-        (
-            json!([{"role": "system", "content": "s"}, {"role": "user", "content": "hi"}]),
-            json!({"temperature": 0.1}),
-        ),
-        (
-            json!([{"role": "user", "content": "hi"}, {"role": "assistant", "content": "yo"}]),
-            json!({}),
-        ),
-    ] {
-        assert_eq!(
-            decline_reason(
-                "anthropic/claude-sonnet-4-5",
-                None,
-                messages.clone(),
-                params.clone()
-            ),
-            None,
-            "gate declined {messages}"
-        );
-        prepare_chat_completions_call(request(
-            "anthropic/claude-sonnet-4-5",
-            None,
-            messages.clone(),
-            params,
-        ))
-        .unwrap_or_else(|error| panic!("prepare declined {messages}: {error}"));
-    }
-}
-
-mod round_trip {
-    use tokio::{
-        io::{AsyncReadExt, AsyncWriteExt},
-        net::{TcpListener, TcpStream},
-    };
-
-    use super::*;
-    use crate::chat_completions::chat_completions;
-
-    async fn read_http_request(socket: &mut TcpStream) -> String {
-        let mut request = Vec::new();
-        let mut buffer = [0_u8; 1024];
-        let header_end = loop {
-            let n = socket.read(&mut buffer).await.expect("reads request");
-            if n == 0 {
-                break request.len();
-            }
-            request.extend_from_slice(&buffer[..n]);
-            if let Some(position) = request.windows(4).position(|window| window == b"\r\n\r\n") {
-                break position + 4;
-            }
-        };
-        let headers = String::from_utf8_lossy(&request[..header_end]);
-        let content_length = headers
-            .lines()
-            .find_map(|line| {
-                let (name, value) = line.split_once(':')?;
-                name.eq_ignore_ascii_case("content-length")
-                    .then(|| value.trim().parse::<usize>().ok())
-                    .flatten()
-            })
-            .unwrap_or(0);
-        while request.len().saturating_sub(header_end) < content_length {
-            let n = socket.read(&mut buffer).await.expect("reads body");
-            if n == 0 {
-                break;
-            }
-            request.extend_from_slice(&buffer[..n]);
-        }
-        String::from_utf8(request).expect("request is utf8")
-    }
-
-    fn http_response(status: &str, body: &str) -> String {
-        format!(
-            "HTTP/1.1 {status}\r\ncontent-type: application/json\r\ncontent-length: {}\r\nconnection: close\r\n\r\n{}",
-            body.len(),
-            body
-        )
-    }
-
-    /// Serve one request from a stub upstream and hand back what it received.
-    async fn serve_once(
-        status: &'static str,
-        body: &'static str,
-    ) -> (String, tokio::task::JoinHandle<String>) {
-        let listener = TcpListener::bind("127.0.0.1:0").await.expect("binds");
-        let port = listener.local_addr().expect("addr").port();
-        let handle = tokio::spawn(async move {
-            let (mut socket, _) = listener.accept().await.expect("accepts");
-            let received = read_http_request(&mut socket).await;
-            socket
-                .write_all(http_response(status, body).as_bytes())
-                .await
-                .expect("writes response");
-            socket.flush().await.expect("flushes");
-            received
-        });
-        (format!("http://127.0.0.1:{port}/v1/messages"), handle)
-    }
-
-    fn call(api_base: &str, messages: Value, params: Value) -> ChatCompletionsRequest<'_> {
-        ChatCompletionsRequest {
-            model: "anthropic/claude-sonnet-4-5",
-            messages,
-            optional_params: match params {
-                Value::Object(map) => map,
-                other => panic!("params must be an object, got {other}"),
-            },
-            api_key: Some("sk-test"),
-            api_base: Some(api_base),
-            custom_llm_provider: None,
-            extra_headers: None,
-            timeout: Some(std::time::Duration::from_secs(10)),
-        }
-    }
-
-    const GOOD_BODY: &str = r#"{"id":"msg_1","type":"message","role":"assistant","model":"claude-sonnet-4-5-20260101","content":[{"type":"text","text":"hello"}],"stop_reason":"end_turn","stop_sequence":null,"usage":{"input_tokens":11,"output_tokens":4}}"#;
-
-    #[tokio::test]
-    async fn round_trip_sends_the_translated_body_and_normalizes_the_response() {
-        let (api_base, handle) = serve_once("200 OK", GOOD_BODY).await;
-        let response = chat_completions(call(
-            &api_base,
-            json!([
-                {"role": "system", "content": "be terse"},
-                {"role": "user", "content": "hi"}
-            ]),
-            json!({"max_tokens": 16}),
-        ))
-        .await
-        .expect("call succeeds");
-
-        let received = handle.await.expect("server task");
-        let sent: Value = serde_json::from_str(
-            received
-                .split_once("\r\n\r\n")
-                .expect("request has a body")
-                .1,
-        )
-        .expect("body is json");
-        assert_eq!(
-            sent["messages"],
-            json!([{"role": "user", "content": [{"type": "text", "text": "hi"}]}])
-        );
-        assert_eq!(
-            sent["system"],
-            json!([{"type": "text", "text": "be terse"}])
-        );
-        assert_eq!(sent["max_tokens"], json!(16));
-        assert!(received.to_lowercase().contains("x-api-key: sk-test"));
-
-        assert_eq!(
-            response.choices[0].message.content.as_deref(),
-            Some("hello")
-        );
-        assert_eq!(response.usage.total_tokens, 15);
-    }
-
-    #[tokio::test]
-    async fn a_response_it_cannot_normalize_is_reported_as_already_sent() {
-        // The provider was called and billed, so the host must not retry this
-        // on its own path. `MissingField` here would read as a pre-send
-        // decline and be retried; `InvalidResponse` cannot.
-        const NO_USAGE: &str =
-            r#"{"model":"m","content":[{"type":"text","text":"hi"}],"stop_reason":"end_turn"}"#;
-        let (api_base, handle) = serve_once("200 OK", NO_USAGE).await;
-        let err = chat_completions(call(
-            &api_base,
-            json!([{"role": "user", "content": "hi"}]),
-            json!({"max_tokens": 16}),
-        ))
-        .await
-        .expect_err("response cannot be normalized");
-        handle.await.expect("server task");
-        assert!(
-            matches!(err, Error::InvalidResponse(_)),
-            "expected a post-send error, got {err:?}"
-        );
-    }
-
-    #[tokio::test]
-    async fn a_tool_use_block_in_the_response_is_also_reported_as_already_sent() {
-        const TOOL_USE: &str = r#"{"model":"m","content":[{"type":"tool_use","id":"t","name":"f","input":{}}],"stop_reason":"tool_use","usage":{"input_tokens":1,"output_tokens":1}}"#;
-        let (api_base, handle) = serve_once("200 OK", TOOL_USE).await;
-        let err = chat_completions(call(
-            &api_base,
-            json!([{"role": "user", "content": "hi"}]),
-            json!({"max_tokens": 16}),
-        ))
-        .await
-        .expect_err("response cannot be normalized");
-        handle.await.expect("server task");
-        assert!(
-            matches!(err, Error::InvalidResponse(_)),
-            "expected a post-send error, got {err:?}"
-        );
-    }
-
-    #[tokio::test]
-    async fn an_upstream_error_status_keeps_its_code() {
-        let (api_base, handle) =
-            serve_once("429 Too Many Requests", r#"{"error":"slow down"}"#).await;
-        let err = chat_completions(call(
-            &api_base,
-            json!([{"role": "user", "content": "hi"}]),
-            json!({"max_tokens": 16}),
-        ))
-        .await
-        .expect_err("upstream rejects");
-        handle.await.expect("server task");
-        assert!(
-            matches!(
-                err,
-                Error::Transport(litellm_http::transport::Error::Http { status: 429, .. })
-            ),
-            "expected a 429, got {err:?}"
-        );
-    }
-
-    #[tokio::test]
-    async fn a_connection_that_is_never_established_declines_instead_of_failing() {
-        // Nothing was sent, so nothing was billed and the host can still serve
-        // the request. Classing this with the post-send failures would turn a
-        // recoverable fallback into a user-facing error on exactly the
-        // deployments whose transport is configured only on the Python client.
-        let port = {
-            let listener = TcpListener::bind("127.0.0.1:0").await.expect("binds");
-            listener.local_addr().expect("has an address").port()
-            // Dropped here, so the port is closed and the connect is refused.
-        };
-        let err = chat_completions(call(
-            &format!("http://127.0.0.1:{port}/v1/messages"),
-            json!([{"role": "user", "content": "hi"}]),
-            json!({"max_tokens": 16}),
-        ))
-        .await
-        .expect_err("nothing is listening");
-        assert!(
-            matches!(
-                err,
-                Error::Transport(litellm_http::transport::Error::Connect(_))
-            ),
-            "expected a pre-send connect failure, got {err:?}"
-        );
-    }
-
-    #[test]
-    fn response_errors_collapse_to_one_variant_that_can_only_mean_already_sent() {
-        use crate::chat_completions::handler::as_response_error;
-
-        for original in [
-            Error::MissingField("usage"),
-            Error::Unsupported("non-text response content block"),
-            Error::InvalidRequest("whatever".to_string()),
-            Error::Auth(litellm_auth::Error::InvalidHeader),
-        ] {
-            let label = format!("{original:?}");
-            assert!(
-                matches!(as_response_error(original), Error::InvalidResponse(_)),
-                "{label} must not stay retryable once the provider has answered"
-            );
-        }
-        // An upstream status is already unambiguous, so it survives intact.
-        assert!(matches!(
-            as_response_error(Error::Transport(litellm_http::transport::Error::Http {
-                status: 500,
-                body: "boom".to_string()
-            })),
-            Error::Transport(litellm_http::transport::Error::Http { status: 500, .. })
-        ));
-    }
-}

litellm-rust/crates/core/src/messages/common_utils.rs

@@ -1,5 +1,5 @@
 use litellm_http::request::string_headers as shared_string_headers;
-pub(super) use litellm_http::request::{has_bearer_auth, has_header, truncate_error_body};
+pub(super) use litellm_http::request::truncate_error_body;
 use litellm_llms::{
     anthropic::experimental_pass_through::messages::transformation::ANTHROPIC_MESSAGES_CONFIG,
     azure_ai::anthropic::messages_transformation::AZURE_ANTHROPIC_MESSAGES_CONFIG,
@@ -26,3 +26,45 @@ pub(super) fn string_headers(
 ) -> Result<Vec<(String, String)>, Error> {
     shared_string_headers(HEADER_CONTEXT, extra_headers).map_err(Error::from)
 }
+
+#[cfg(test)]
+mod tests {
+    use serde_json::json;
+
+    use super::{messages_provider_config, string_headers, truncate_error_body};
+    use crate::messages::Error;
+
+    #[test]
+    fn provider_config_resolves_anthropic_and_azure_ai() {
+        assert!(messages_provider_config("anthropic").is_some());
+        assert!(messages_provider_config("azure_ai").is_some());
+        assert!(messages_provider_config("openai").is_none());
+    }
+
+    #[test]
+    fn truncate_error_body_caps_long_payloads() {
+        let body = "x".repeat(400);
+        let truncated = truncate_error_body(&body);
+        assert!(truncated.ends_with("... (truncated)"));
+        let prefix_chars = truncated
+            .strip_suffix("... (truncated)")
+            .expect("truncated marker present")
+            .chars()
+            .count();
+        assert_eq!(prefix_chars, 256);
+    }
+
+    #[test]
+    fn string_headers_rejects_non_string_values() {
+        let headers = json!({"x-count": 3}).as_object().unwrap().clone();
+        let err = string_headers(Some(headers)).expect_err("non-string header rejected");
+        assert_eq!(
+            err,
+            Error::Headers(litellm_http::request::HeaderError {
+                context: "messages",
+                name: "x-count".to_string(),
+                actual: "number",
+            })
+        );
+    }
+}

litellm-rust/crates/core/src/messages/error.rs

@@ -1,3 +1,5 @@
+use std::sync::Arc;
+
 use litellm_llms::base_llm::chat::transformation::Error as LlmError;
 
 #[derive(Clone, Debug, PartialEq, Eq, thiserror::Error)]
@@ -18,8 +20,34 @@ pub enum Error {
     Transport(#[from] litellm_http::transport::Error),
     #[error(transparent)]
     Headers(#[from] litellm_http::request::HeaderError),
+    #[error(transparent)]
+    Secret(#[from] SecretError),
 }
 
+#[derive(Clone, Debug, thiserror::Error)]
+#[error(transparent)]
+pub struct SecretError(Arc<litellm_secrets::Error>);
+
+impl SecretError {
+    pub fn source_error(&self) -> &litellm_secrets::Error {
+        &self.0
+    }
+}
+
+impl From<litellm_secrets::Error> for Error {
+    fn from(error: litellm_secrets::Error) -> Self {
+        Self::Secret(SecretError(Arc::new(error)))
+    }
+}
+
+impl PartialEq for SecretError {
+    fn eq(&self, other: &Self) -> bool {
+        Arc::ptr_eq(&self.0, &other.0)
+    }
+}
+
+impl Eq for SecretError {}
+
 impl From<LlmError> for Error {
     fn from(error: LlmError) -> Self {
         match error {

litellm-rust/crates/core/src/messages/handler.rs

@@ -17,8 +17,10 @@ pub(super) async fn send(
     body: &Value,
     timeout: Option<Duration>,
 ) -> Result<reqwest::Response, Error> {
+    let encoded = serde_json::to_vec(body)
+        .map_err(|err| Error::InvalidRequest(format!("failed to encode messages body: {err}")))?;
     let builder = headers.iter().fold(
-        http_client().post(url).json(body),
+        http_client().post(url).body(encoded),
         |builder, (key, value)| builder.header(key, value),
     );
     let builder = match timeout {

litellm-rust/crates/core/src/messages/mod.rs

@@ -12,6 +12,9 @@ mod common_utils;
 mod handler;
 mod prepare;
 pub mod route;
+use std::sync::Arc;
+
+use litellm_secrets::source::EnvironmentSecrets;
 use litellm_types::llms::anthropic_messages::anthropic_response::AnthropicMessagesResponse;
 use route::{LocalMessagesHost, MessagesCall, MessagesOutput, messages_machine};
 use serde_json::Value;
@@ -31,15 +34,15 @@ pub async fn messages(request: MessagesRequest<'_>) -> Result<AnthropicMessagesR
         api_base: request.api_base.map(Into::into),
         custom_llm_provider: request.custom_llm_provider.map(Into::into),
         extra_headers: request.extra_headers,
+        provider_specific_header: request.provider_specific_header,
         timeout: request.timeout,
+        shaping: request.shaping,
     };
-    match litellm_host::run::run(messages_machine(), &LocalMessagesHost::new(call)).await? {
+    let secrets = Arc::new(EnvironmentSecrets::python_compatible());
+    match litellm_host::run::run(messages_machine(secrets), &LocalMessagesHost::new(call)).await? {
         MessagesOutput::Message(message) => Ok(*message),
         MessagesOutput::Streamed => Err(Error::Unsupported(
             "streamed responses need a streaming host",
         )),
     }
 }
-
-#[cfg(test)]
-mod tests;

litellm-rust/crates/core/src/messages/prepare.rs

@@ -1,51 +1,102 @@
-use litellm_core_utils::get_llm_provider_logic::{CustomLlmProvider, get_custom_llm_provider};
-use litellm_llms::base_llm::anthropic_messages::transformation::{
-    BaseAnthropicMessagesConfig, MessagesAuthStrategy,
+use litellm_core_utils::{
+    dot_notation_indexing::delete_nested_value,
+    get_llm_provider_logic::{CustomLlmProvider, get_custom_llm_provider},
+    get_provider_specific_headers::get_provider_specific_headers,
+    settings::Lookup,
+};
+use litellm_llms::{
+    anthropic::experimental_pass_through::messages::handler::shape_anthropic_messages_request,
+    base_llm::anthropic_messages::transformation::{
+        BaseAnthropicMessagesConfig, MessagesTransformContext,
+    },
 };
 use litellm_types::llms::anthropic_messages::anthropic_request::AnthropicMessagesRequest;
 use serde_json::{Map, Value};
 
 use super::{
     Error,
-    common_utils::{has_bearer_auth, has_header, messages_provider_config, string_headers},
+    common_utils::{messages_provider_config, string_headers},
 };
 use crate::messages::types::{MessagesRequest, ProviderMessagesRequest};
 
-pub(super) fn prepare_provider_request(
-    request: MessagesRequest<'_>,
-) -> Result<ProviderMessagesRequest, Error> {
-    let provider_info = get_custom_llm_provider(request.model, request.custom_llm_provider)
+pub(super) struct ResolvedProvider<'a> {
+    pub(super) model: &'a str,
+    pub(super) provider: &'a str,
+    pub(super) config: &'static dyn BaseAnthropicMessagesConfig,
+}
+
+pub(super) fn resolve_provider<'a>(
+    model: &'a str,
+    custom_llm_provider: Option<&'a str>,
+) -> Result<ResolvedProvider<'a>, Error> {
+    let CustomLlmProvider {
+        model,
+        custom_llm_provider: provider,
+    } = get_custom_llm_provider(model, custom_llm_provider)
         .or_else(|| {
-            request
-                .custom_llm_provider
-                .map(|provider| CustomLlmProvider {
-                    model: request.model,
-                    custom_llm_provider: provider,
-                })
+            custom_llm_provider.map(|provider| CustomLlmProvider {
+                model,
+                custom_llm_provider: provider,
+            })
         })
         .ok_or_else(|| {
             Error::InvalidProvider(
                 "unable to resolve custom_llm_provider for messages request".to_string(),
             )
         })?;
-    let model = provider_info.model.to_string();
-    let provider = provider_info.custom_llm_provider;
-
     let config = messages_provider_config(provider)
         .ok_or_else(|| Error::InvalidProvider(provider.to_string()))?;
-    let env_lookup = |key: &str| std::env::var(key).ok();
+    Ok(ResolvedProvider {
+        model,
+        provider,
+        config,
+    })
+}
 
-    let headers =
-        validate_environment(config, request.extra_headers, request.api_key, &env_lookup)?;
+pub(super) fn prepare_provider_request(
+    request: MessagesRequest<'_>,
+    resolved: ResolvedProvider<'_>,
+    secrets: &dyn Lookup,
+) -> Result<ProviderMessagesRequest, Error> {
+    let ResolvedProvider {
+        model,
+        provider,
+        config,
+    } = resolved;
+    let model = model.to_string();
+    let env_lookup = |key: &str| secrets.get(key);
 
     let typed_request: AnthropicMessagesRequest =
-        serde_json::from_value(request.body).map_err(|err| {
-            Error::InvalidRequest(format!("invalid Anthropic messages request: {err}"))
-        })?;
-    let transformed = config.transform_anthropic_messages_request(AnthropicMessagesRequest {
-        model: model.clone(),
-        ..typed_request
-    })?;
+        serde_json::from_value(request.body).map_err(invalid_request)?;
+    let sanitized = shape_anthropic_messages_request(
+        AnthropicMessagesRequest {
+            model: model.clone(),
+            ..typed_request
+        },
+        request.shaping.reasoning_auto_summary,
+    )?;
+    let trimmed =
+        without_additional_drop_params(sanitized, &request.shaping.additional_drop_params)?;
+    let transformed = config.transform_anthropic_messages_request(
+        trimmed,
+        &MessagesTransformContext::new(request.shaping.capabilities, request.shaping.drop_params),
+    )?;
+
+    let scoped = get_provider_specific_headers(request.provider_specific_header.as_ref(), provider);
+    let forwarded = string_headers(Some(
+        request
+            .extra_headers
+            .into_iter()
+            .flatten()
+            .chain(scoped)
+            .collect(),
+    ))?;
+    let authenticated = config.authenticate(forwarded, request.api_key, &env_lookup)?;
+    let headers = config.request_headers(
+        with_default_headers(authenticated, config.default_headers()),
+        &transformed,
+    );
+
     let body = serde_json::to_value(transformed).map_err(|err| {
         Error::InvalidRequest(format!(
             "failed to serialize Anthropic messages request: {err}"
@@ -65,33 +116,371 @@ pub(super) fn prepare_provider_request(
     })
 }
 
-fn validate_environment(
-    config: &dyn BaseAnthropicMessagesConfig,
-    extra_headers: Option<Map<String, Value>>,
-    api_key: Option<&str>,
-    env_lookup: &dyn Fn(&str) -> Option<String>,
-) -> Result<Vec<(String, String)>, Error> {
-    let mut headers = string_headers(extra_headers)?;
-
-    let auth_strategy = config.auth_strategy();
-    let already_authorized = has_header(&headers, auth_strategy.header_name())
-        || (config.accepts_bearer_auth() && has_bearer_auth(&headers));
-    if !already_authorized {
-        let api_key = config.resolve_api_key(api_key, env_lookup)?;
-        let auth_header = match auth_strategy {
-            MessagesAuthStrategy::Bearer => {
-                ("authorization".to_string(), format!("Bearer {api_key}"))
-            }
-            MessagesAuthStrategy::Header(name) => (name.to_string(), api_key),
-        };
-        headers.push(auth_header);
-    }
-
-    for (name, value) in config.default_headers() {
-        if !has_header(&headers, name) {
-            headers.push((name.to_string(), value.to_string()));
-        }
-    }
-
-    Ok(headers)
+fn invalid_request(err: serde_json::Error) -> Error {
+    Error::InvalidRequest(format!("invalid Anthropic messages request: {err}"))
+}
+
+fn without_additional_drop_params(
+    request: AnthropicMessagesRequest,
+    paths: &[String],
+) -> Result<AnthropicMessagesRequest, Error> {
+    if paths.is_empty() {
+        return Ok(request);
+    }
+    let Value::Object(fields) = serde_json::to_value(request).map_err(invalid_request)? else {
+        return Err(Error::InvalidRequest(
+            "Anthropic messages request did not serialize to an object".to_string(),
+        ));
+    };
+    let (required, optional): (Map<String, Value>, Map<String, Value>) = fields
+        .into_iter()
+        .partition(|(key, _)| matches!(key.as_str(), "model" | "messages"));
+    let trimmed = paths.iter().fold(Value::Object(optional), |body, path| {
+        delete_nested_value(body, path)
+    });
+    let merged: Map<String, Value> = required
+        .into_iter()
+        .chain(trimmed.as_object().cloned().unwrap_or_default())
+        .collect();
+    serde_json::from_value(Value::Object(merged)).map_err(invalid_request)
+}
+
+fn with_default_headers(
+    headers: Vec<(String, String)>,
+    defaults: &[(&str, &str)],
+) -> Vec<(String, String)> {
+    let missing: Vec<(String, String)> = defaults
+        .iter()
+        .filter(|(name, _)| {
+            !headers
+                .iter()
+                .any(|(header, _)| header.eq_ignore_ascii_case(name))
+        })
+        .map(|(name, value)| ((*name).to_string(), (*value).to_string()))
+        .collect();
+    headers.into_iter().chain(missing).collect()
+}
+
+#[cfg(test)]
+mod tests {
+    use litellm_types::utils::ProviderSpecificHeaders;
+    use rstest::{fixture, rstest};
+    use serde_json::json;
+
+    use super::*;
+    use crate::messages::types::MessagesShaping;
+
+    #[fixture]
+    fn shaping() -> MessagesShaping {
+        MessagesShaping::default()
+    }
+
+    fn prepare(request: MessagesRequest<'_>) -> Result<ProviderMessagesRequest, Error> {
+        prepare_with_secrets(request, &|_: &str| None)
+    }
+
+    fn prepare_with_secrets(
+        request: MessagesRequest<'_>,
+        secrets: &dyn Lookup,
+    ) -> Result<ProviderMessagesRequest, Error> {
+        let resolved = resolve_provider(request.model, request.custom_llm_provider)?;
+        prepare_provider_request(request, resolved, secrets)
+    }
+
+    #[rstest]
+    #[case::api_key(
+        &[("ANTHROPIC_API_KEY", "sk-secret")],
+        &[("x-api-key", "sk-secret")],
+        "https://api.anthropic.com/v1/messages"
+    )]
+    #[case::auth_token(
+        &[("ANTHROPIC_AUTH_TOKEN", "token")],
+        &[("authorization", "Bearer token")],
+        "https://api.anthropic.com/v1/messages"
+    )]
+    #[case::api_base(
+        &[("ANTHROPIC_API_KEY", "sk-secret"), ("ANTHROPIC_API_BASE", "https://gateway.test")],
+        &[("x-api-key", "sk-secret")],
+        "https://gateway.test/v1/messages"
+    )]
+    #[case::sdk_base_url(
+        &[("ANTHROPIC_API_KEY", "sk-secret"), ("ANTHROPIC_BASE_URL", "https://sdk.test")],
+        &[("x-api-key", "sk-secret")],
+        "https://sdk.test/v1/messages"
+    )]
+    fn credentials_and_base_come_from_the_resolved_secrets(
+        shaping: MessagesShaping,
+        #[case] secrets: &[(&str, &str)],
+        #[case] expected_auth: &[(&str, &str)],
+        #[case] expected_url: &str,
+    ) {
+        let lookup = |name: &str| {
+            secrets
+                .iter()
+                .find(|(key, _)| *key == name)
+                .map(|(_, value)| value.to_string())
+        };
+        let prepared = prepare_with_secrets(
+            MessagesRequest {
+                model: "claude-test",
+                body: json!({"model": "claude-test", "messages": [{"role": "user", "content": "hi"}], "max_tokens": 16}),
+                api_key: None,
+                api_base: None,
+                custom_llm_provider: Some("anthropic"),
+                extra_headers: None,
+                provider_specific_header: None,
+                timeout: None,
+                shaping,
+            },
+            &lookup,
+        )
+        .unwrap();
+        let auth: Vec<(&str, &str)> = prepared
+            .upstream_headers
+            .iter()
+            .filter(|(name, _)| matches!(name.as_str(), "x-api-key" | "authorization"))
+            .map(|(name, value)| (name.as_str(), value.as_str()))
+            .collect();
+        assert_eq!(
+            (auth.as_slice(), prepared.url.as_str()),
+            (expected_auth, expected_url)
+        );
+    }
+
+    fn prepared_body(body: Value, shaping: MessagesShaping) -> Result<Value, Error> {
+        prepare(MessagesRequest {
+            model: "anthropic/claude-test",
+            body,
+            api_key: Some("sk-test"),
+            api_base: Some("https://anthropic.test"),
+            custom_llm_provider: Some("anthropic"),
+            extra_headers: None,
+            provider_specific_header: None,
+            timeout: None,
+            shaping,
+        })
+        .map(|prepared| prepared.body)
+    }
+
+    #[rstest]
+    #[case::nothing_forwarded(
+        &[],
+        &[("x-version", "1"), ("content-type", "application/json")],
+        &[("x-version", "1"), ("content-type", "application/json")],
+    )]
+    #[case::forwarded_header_wins_in_any_case(
+        &[("X-Version", "custom"), ("x-api-key", "k")],
+        &[("x-version", "1"), ("content-type", "application/json")],
+        &[("X-Version", "custom"), ("x-api-key", "k"), ("content-type", "application/json")],
+    )]
+    #[case::no_defaults(&[("x-api-key", "k")], &[], &[("x-api-key", "k")])]
+    fn default_headers_fill_only_missing_names(
+        #[case] forwarded: &[(&str, &str)],
+        #[case] defaults: &[(&str, &str)],
+        #[case] expected: &[(&str, &str)],
+    ) {
+        let owned = |headers: &[(&str, &str)]| -> Vec<(String, String)> {
+            headers
+                .iter()
+                .map(|(name, value)| ((*name).to_string(), (*value).to_string()))
+                .collect()
+        };
+        assert_eq!(
+            with_default_headers(owned(forwarded), defaults),
+            owned(expected)
+        );
+    }
+
+    #[rstest]
+    #[case::top_level_and_nested_paths(
+        json!({
+            "max_tokens": 1024,
+            "thinking": {"type": "enabled", "budget_tokens": 2048},
+            "context_management": {"edits": [{"type": "clear_thinking_20251015"}]},
+            "metadata": {"user_id": "u1"},
+            "tools": [{"name": "lookup", "input_schema": {"type": "object"}, "input_examples": [{"q": "x"}]}]
+        }),
+        &["thinking", "context_management", "tools[*].input_examples"],
+        json!({
+            "max_tokens": 1024,
+            "metadata": {"user_id": "u1"},
+            "tools": [{"name": "lookup", "input_schema": {"type": "object"}}]
+        }),
+    )]
+    #[case::no_paths(
+        json!({"max_tokens": 16, "safeguards": [{"type": "dangerous_tool_use"}]}),
+        &[],
+        json!({"max_tokens": 16, "safeguards": [{"type": "dangerous_tool_use"}]}),
+    )]
+    #[case::model_and_messages_are_never_dropped(
+        json!({"max_tokens": 16}),
+        &["model", "messages", "messages[0].content"],
+        json!({"max_tokens": 16}),
+    )]
+    fn prepared_body_drops_configured_paths(
+        shaping: MessagesShaping,
+        #[case] fields: Value,
+        #[case] additional_drop_params: &[&str],
+        #[case] expected_fields: Value,
+    ) {
+        let with_messages = |fields: Value| -> Value {
+            let Value::Object(fields) = fields else {
+                unreachable!()
+            };
+            Value::Object(
+                [
+                    ("model".to_string(), json!("claude-test")),
+                    (
+                        "messages".to_string(),
+                        json!([{"role": "user", "content": "hi"}]),
+                    ),
+                ]
+                .into_iter()
+                .chain(fields)
+                .collect(),
+            )
+        };
+        let shaping = MessagesShaping {
+            additional_drop_params: additional_drop_params
+                .iter()
+                .map(ToString::to_string)
+                .collect(),
+            ..shaping
+        };
+        assert_eq!(
+            prepared_body(with_messages(fields), shaping),
+            Ok(with_messages(expected_fields))
+        );
+    }
+
+    #[rstest]
+    #[case::model_prefix_picks_the_provider(
+        "azure_ai/claude-test",
+        None,
+        &[("x-priority", "extra"), ("x-scoped", "azure_ai")]
+    )]
+    #[case::explicit_provider(
+        "claude-test",
+        Some("anthropic"),
+        &[("x-priority", "scoped"), ("x-scoped", "anthropic")]
+    )]
+    #[case::provider_prefix_on_an_anthropic_model(
+        "anthropic/claude-test",
+        None,
+        &[("x-priority", "scoped"), ("x-scoped", "anthropic")]
+    )]
+    fn provider_specific_headers_follow_the_resolved_provider(
+        shaping: MessagesShaping,
+        #[case] model: &str,
+        #[case] custom_llm_provider: Option<&str>,
+        #[case] expected: &[(&str, &str)],
+    ) {
+        let configured: ProviderSpecificHeaders = serde_json::from_value(json!([
+            {"custom_llm_provider": "azure_ai", "extra_headers": {"x-scoped": "azure_ai"}},
+            {"custom_llm_provider": "anthropic", "extra_headers": {"x-scoped": "anthropic", "x-priority": "scoped"}}
+        ]))
+        .unwrap();
+        let prepared = prepare(MessagesRequest {
+            model,
+            body: json!({"model": model, "messages": [{"role": "user", "content": "hi"}], "max_tokens": 16}),
+            api_key: Some("sk-test"),
+            api_base: Some("https://resource.services.ai.azure.com"),
+            custom_llm_provider,
+            extra_headers: Some(serde_json::from_value(json!({"x-priority": "extra"})).unwrap()),
+            provider_specific_header: Some(configured),
+            timeout: None,
+            shaping,
+        })
+        .unwrap();
+        let caller_headers: Vec<(&str, &str)> = prepared
+            .upstream_headers
+            .iter()
+            .filter(|(name, _)| matches!(name.as_str(), "x-priority" | "x-scoped"))
+            .map(|(name, value)| (name.as_str(), value.as_str()))
+            .collect();
+        assert_eq!(caller_headers, expected);
+    }
+
+    #[rstest]
+    fn prepared_body_carries_the_provider_stripped_model(shaping: MessagesShaping) {
+        assert_eq!(
+            prepared_body(
+                json!({
+                    "model": "anthropic/claude-test",
+                    "messages": [{"role": "user", "content": "hi"}],
+                    "max_tokens": 16
+                }),
+                shaping,
+            ),
+            Ok(json!({
+                "model": "claude-test",
+                "messages": [{"role": "user", "content": "hi"}],
+                "max_tokens": 16
+            }))
+        );
+    }
+
+    #[rstest]
+    fn dropped_thinking_display_is_not_restored_by_auto_summary(shaping: MessagesShaping) {
+        let shaping = MessagesShaping {
+            reasoning_auto_summary: true,
+            additional_drop_params: vec!["thinking.display".to_string()],
+            ..shaping
+        };
+        assert_eq!(
+            prepared_body(
+                json!({
+                    "model": "claude-test",
+                    "messages": [{"role": "user", "content": "hi"}],
+                    "max_tokens": 4096,
+                    "thinking": {"type": "enabled", "budget_tokens": 2048}
+                }),
+                shaping,
+            ),
+            Ok(json!({
+                "model": "claude-test",
+                "messages": [{"role": "user", "content": "hi"}],
+                "max_tokens": 4096,
+                "thinking": {"type": "enabled", "budget_tokens": 2048}
+            }))
+        );
+    }
+
+    #[rstest]
+    fn dropping_an_invalid_metadata_user_id_does_not_skip_its_validation(shaping: MessagesShaping) {
+        let shaping = MessagesShaping {
+            additional_drop_params: vec!["metadata.user_id".to_string()],
+            ..shaping
+        };
+        assert!(matches!(
+            prepared_body(
+                json!({
+                    "model": "claude-test",
+                    "messages": [{"role": "user", "content": "hi"}],
+                    "max_tokens": 16,
+                    "metadata": {"user_id": 123}
+                }),
+                shaping,
+            ),
+            Err(Error::InvalidRequest(_))
+        ));
+    }
+
+    #[rstest]
+    fn prepared_body_rejects_invalid_metadata_before_the_call(shaping: MessagesShaping) {
+        assert_eq!(
+            prepared_body(
+                json!({
+                    "model": "claude-test",
+                    "messages": [{"role": "user", "content": "hi"}],
+                    "max_tokens": 16,
+                    "metadata": {"user_id": 123}
+                }),
+                shaping,
+            ),
+            Err(Error::InvalidRequest(
+                "metadata.user_id must be a string, got 123".to_string()
+            ))
+        );
+    }
 }

litellm-rust/crates/core/src/messages/route.rs

@@ -1,35 +1,32 @@
-use std::{sync::Mutex, time::Duration};
+use std::{
+    convert::Infallible,
+    sync::{Arc, Mutex},
+    time::Duration,
+};
 
 use bytes::Bytes;
 use litellm_auth::SecretValue;
-use litellm_core_utils::get_llm_provider_logic::get_custom_llm_provider;
 use litellm_host::{
     event::{MachineEvent, RawResponse, RequestContext, WireRequest},
     host::{Demand, Host},
-    machine::{HostChannel, MachineFault, RouteMachine},
-    route::Route,
+    machine::{CallMachine, HostChannel, MachineFault},
+    protocol::Protocol,
+};
+use litellm_secrets::source::SecretSource;
+use litellm_types::{
+    llms::anthropic_messages::anthropic_response::AnthropicMessagesResponse,
+    utils::ProviderSpecificHeaders,
 };
-use litellm_types::llms::anthropic_messages::anthropic_response::AnthropicMessagesResponse;
 use serde_json::{Map, Value};
 
 use super::{
     Error,
-    common_utils::messages_provider_config,
     handler::{decode_response, network, provider_error, send},
-    prepare::prepare_provider_request,
-    types::MessagesRequest,
+    prepare::{prepare_provider_request, resolve_provider},
+    types::{MessagesRequest, MessagesShaping},
 };
 use crate::constants::ANTHROPIC_MESSAGES_PROVIDER;
 
-#[derive(Clone, Copy, Debug, PartialEq, Eq)]
-pub enum MessagesOp {
-    ProjectRequest,
-}
-
-pub enum MessagesOpResult {
-    Request(Box<MessagesCall>),
-}
-
 /// The caller's request as the host projects it.
 pub struct MessagesCall {
     pub model: String,
@@ -38,7 +35,9 @@ pub struct MessagesCall {
     pub api_base: Option<String>,
     pub custom_llm_provider: Option<String>,
     pub extra_headers: Option<Map<String, Value>>,
+    pub provider_specific_header: Option<ProviderSpecificHeaders>,
     pub timeout: Option<Duration>,
+    pub shaping: MessagesShaping,
 }
 
 impl MessagesCall {
@@ -53,15 +52,20 @@ pub enum MessagesOutput {
     Streamed,
 }
 
+/// The upstream response as the caller sees it at stream hand-off, before any chunk.
+pub struct MessagesStreamHead {
+    pub headers: Vec<(String, String)>,
+}
+
 pub struct Messages;
 
-impl Route for Messages {
+impl Protocol for Messages {
     type Response = MessagesOutput;
     type Error = Error;
-    type Op = MessagesOp;
-    type OpResult = MessagesOpResult;
+    type Projection = MessagesCall;
+    type Op = Infallible;
     type Chunk = Bytes;
-    type StreamHead = ();
+    type StreamHead = MessagesStreamHead;
 }
 
 impl From<MachineFault> for Error {
@@ -69,26 +73,12 @@ impl From<MachineFault> for Error {
         Self::InvalidRequest(match fault {
             MachineFault::Abandoned => "messages host driver was abandoned".into(),
             MachineFault::Protocol(message) => format!("messages {message}"),
-            MachineFault::Mismatch => "invalid messages host operation result".into(),
         })
     }
 }
 
 pub type MessagesHost = HostChannel<Messages>;
-pub type MessagesMachine = RouteMachine<Messages>;
-
-/// Whether this route serves the request, decided before any callback runs so a host
-/// can still run its own path.
-pub fn supports(model: &str, custom_llm_provider: Option<&str>, stream: bool) -> bool {
-    let provider = get_custom_llm_provider(model, custom_llm_provider)
-        .map(|resolved| resolved.custom_llm_provider)
-        .or(custom_llm_provider);
-    match provider {
-        Some(ANTHROPIC_MESSAGES_PROVIDER) => true,
-        Some(provider) => !stream && messages_provider_config(provider).is_some(),
-        None => false,
-    }
-}
+pub type MessagesMachine = CallMachine<Messages>;
 
 /// The in-process host for a request already in hand. It answers projection once and
 /// observes nothing.
@@ -105,37 +95,46 @@ impl LocalMessagesHost {
 }
 
 impl Host<Messages> for LocalMessagesHost {
-    async fn route(&self, op: MessagesOp) -> Result<MessagesOpResult, Error> {
-        match op {
-            MessagesOp::ProjectRequest => self
-                .call
-                .lock()
-                .unwrap_or_else(|error| error.into_inner())
-                .take()
-                .map(|call| MessagesOpResult::Request(Box::new(call)))
-                .ok_or_else(|| {
-                    Error::InvalidRequest("messages request was already projected".into())
-                }),
-        }
+    async fn project(&self) -> Result<MessagesCall, Error> {
+        self.call
+            .lock()
+            .unwrap_or_else(|error| error.into_inner())
+            .take()
+            .ok_or_else(|| Error::InvalidRequest("messages request was already projected".into()))
+    }
+
+    async fn custom_op(&self, op: Infallible) -> Result<(), Error> {
+        match op {}
     }
 }
 
-pub fn messages_machine() -> MessagesMachine {
-    RouteMachine::new(|host| Box::pin(execute(host)))
+pub fn messages_machine(secrets: Arc<dyn SecretSource>) -> MessagesMachine {
+    CallMachine::new(move |host| Box::pin(execute(host, secrets.clone())))
 }
 
-async fn execute(host: MessagesHost) -> Result<MessagesOutput, Error> {
-    let MessagesOpResult::Request(call) = host.route(MessagesOp::ProjectRequest).await?;
+async fn execute(
+    host: MessagesHost,
+    secrets: Arc<dyn SecretSource>,
+) -> Result<MessagesOutput, Error> {
+    let call = host.project().await?;
     let stream = call.streams();
-    let request = prepare_provider_request(MessagesRequest {
-        model: &call.model,
-        body: Value::Object(call.body.clone()),
-        api_key: call.api_key.as_deref(),
-        api_base: call.api_base.as_deref(),
-        custom_llm_provider: call.custom_llm_provider.as_deref(),
-        extra_headers: call.extra_headers.clone(),
-        timeout: call.timeout,
-    })?;
+    let resolved = resolve_provider(&call.model, call.custom_llm_provider.as_deref())?;
+    let secrets = secrets.resolve(resolved.config.secret_names()).await?;
+    let request = prepare_provider_request(
+        MessagesRequest {
+            model: &call.model,
+            body: Value::Object(call.body.clone()),
+            api_key: call.api_key.as_deref(),
+            api_base: call.api_base.as_deref(),
+            custom_llm_provider: call.custom_llm_provider.as_deref(),
+            extra_headers: call.extra_headers.clone(),
+            provider_specific_header: call.provider_specific_header.clone(),
+            timeout: call.timeout,
+            shaping: call.shaping.clone(),
+        },
+        resolved,
+        secrets.as_ref(),
+    )?;
     if stream && request.provider != ANTHROPIC_MESSAGES_PROVIDER {
         return Err(Error::Unsupported("streaming messages for this provider"));
     }
@@ -143,8 +142,11 @@ async fn execute(host: MessagesHost) -> Result<MessagesOutput, Error> {
         model: request.model.clone(),
         custom_llm_provider: request.provider.clone(),
         optional_params: Value::Object(
-            call.body
-                .iter()
+            request
+                .body
+                .as_object()
+                .into_iter()
+                .flatten()
                 .filter(|(name, _)| !matches!(name.as_str(), "model" | "messages"))
                 .map(|(name, value)| (name.clone(), value.clone()))
                 .collect(),
@@ -184,7 +186,14 @@ async fn relay(
     host: &MessagesHost,
     mut response: reqwest::Response,
 ) -> Result<MessagesOutput, Error> {
-    if host.open(()).await? == Demand::Detached {
+    let head = MessagesStreamHead {
+        headers: response
+            .headers()
+            .iter()
+            .filter_map(|(name, value)| Some((name.to_string(), value.to_str().ok()?.to_string())))
+            .collect(),
+    };
+    if host.open(head).await? == Demand::Detached {
         return Ok(MessagesOutput::Streamed);
     }
     while let Some(chunk) = response.chunk().await.map_err(network)? {

litellm-rust/crates/core/src/messages/tests.rs

@@ -1,454 +0,0 @@
-use std::time::Duration;
-
-use serde_json::{Map, Value, json};
-use tokio::{
-    io::{AsyncReadExt, AsyncWriteExt},
-    net::{TcpListener, TcpStream},
-};
-
-use super::{
-    Error,
-    common_utils::{
-        has_bearer_auth, has_header, messages_provider_config, string_headers, truncate_error_body,
-    },
-    messages,
-};
-use crate::messages::types::MessagesRequest;
-
-async fn read_http_request(socket: &mut TcpStream) -> String {
-    let mut request = Vec::new();
-    let mut buffer = [0_u8; 1024];
-    let header_end = loop {
-        let n = socket.read(&mut buffer).await.expect("reads request");
-        if n == 0 {
-            break request.len();
-        }
-        request.extend_from_slice(&buffer[..n]);
-        if let Some(position) = request.windows(4).position(|window| window == b"\r\n\r\n") {
-            break position + 4;
-        }
-    };
-    let headers = String::from_utf8_lossy(&request[..header_end]);
-    let content_length = headers
-        .lines()
-        .find_map(|line| {
-            let (name, value) = line.split_once(':')?;
-            name.eq_ignore_ascii_case("content-length")
-                .then(|| value.trim().parse::<usize>().ok())
-                .flatten()
-        })
-        .unwrap_or(0);
-    while request.len().saturating_sub(header_end) < content_length {
-        let n = socket.read(&mut buffer).await.expect("reads body");
-        if n == 0 {
-            break;
-        }
-        request.extend_from_slice(&buffer[..n]);
-    }
-    String::from_utf8(request).expect("request is utf8")
-}
-
-fn write_response(body: &str) -> String {
-    format!(
-        "HTTP/1.1 200 OK\r\ncontent-type: application/json\r\ncontent-length: {}\r\nconnection: close\r\n\r\n{}",
-        body.len(),
-        body
-    )
-}
-
-#[test]
-fn provider_config_resolves_anthropic_and_azure_ai() {
-    assert!(messages_provider_config("anthropic").is_some());
-    assert!(messages_provider_config("azure_ai").is_some());
-    assert!(messages_provider_config("openai").is_none());
-}
-
-#[test]
-fn truncate_error_body_caps_long_payloads() {
-    let body = "x".repeat(400);
-    let truncated = truncate_error_body(&body);
-    assert!(truncated.ends_with("... (truncated)"));
-    let prefix_chars = truncated
-        .strip_suffix("... (truncated)")
-        .expect("truncated marker present")
-        .chars()
-        .count();
-    assert_eq!(prefix_chars, 256);
-}
-
-#[test]
-fn string_headers_rejects_non_string_values() {
-    let headers = json!({"x-count": 3}).as_object().unwrap().clone();
-    let err = string_headers(Some(headers)).expect_err("non-string header rejected");
-    assert_eq!(
-        err,
-        Error::Headers(litellm_http::request::HeaderError {
-            context: "messages",
-            name: "x-count".to_string(),
-            actual: "number",
-        })
-    );
-}
-
-#[test]
-fn has_header_is_case_insensitive() {
-    let headers = vec![("X-Api-Key".to_string(), "secret".to_string())];
-    assert!(has_header(&headers, "x-api-key"));
-    assert!(!has_header(&headers, "authorization"));
-}
-
-#[test]
-fn has_bearer_auth_requires_a_nonempty_bearer_token() {
-    assert!(has_bearer_auth(&[(
-        "Authorization".to_string(),
-        "Bearer tok".to_string()
-    )]));
-    assert!(has_bearer_auth(&[(
-        "authorization".to_string(),
-        "bearer tok".to_string()
-    )]));
-    assert!(!has_bearer_auth(&[(
-        "authorization".to_string(),
-        "Bearer ".to_string()
-    )]));
-    assert!(!has_bearer_auth(&[(
-        "authorization".to_string(),
-        String::new()
-    )]));
-    assert!(!has_bearer_auth(&[(
-        "authorization".to_string(),
-        "Basic abc".to_string()
-    )]));
-    assert!(!has_bearer_auth(&[(
-        "x-api-key".to_string(),
-        "sk".to_string()
-    )]));
-}
-
-#[tokio::test]
-async fn messages_round_trip_builds_azure_request_and_passes_response_through() {
-    let listener = TcpListener::bind("127.0.0.1:0").await.expect("binds");
-    let addr = listener.local_addr().expect("addr");
-
-    let server = tokio::spawn(async move {
-        let (mut socket, _) = listener.accept().await.expect("accepts request");
-        let request = read_http_request(&mut socket).await;
-        let response_body = r#"{"id":"msg_1","type":"message","role":"assistant","content":[{"type":"text","text":"hi"}],"model":"claude-sonnet-4-5","stop_reason":"end_turn","usage":{"input_tokens":1,"output_tokens":2}}"#;
-        socket
-            .write_all(write_response(response_body).as_bytes())
-            .await
-            .expect("writes response");
-        request
-    });
-
-    let response = messages(MessagesRequest {
-        model: "claude-sonnet-4-5",
-        body: json!({
-            "model": "claude-sonnet-4-5",
-            "max_tokens": 1024,
-            "messages": [{
-                "role": "user",
-                "content": [{
-                    "type": "text",
-                    "text": "hi",
-                    "cache_control": {"type": "ephemeral", "scope": "global"}
-                }]
-            }]
-        }),
-        api_key: Some("sk-azure"),
-        api_base: Some(&format!("http://{addr}")),
-        custom_llm_provider: Some("azure_ai"),
-        extra_headers: None,
-        timeout: Some(Duration::from_secs(5)),
-    })
-    .await
-    .expect("messages request succeeds");
-
-    assert_eq!(response.content[0]["text"], "hi");
-    assert_eq!(response.stop_reason.as_deref(), Some("end_turn"));
-
-    let request = server.await.expect("server task completes");
-    let (head, body) = request.split_once("\r\n\r\n").expect("has body");
-    assert!(head.starts_with("POST /anthropic/v1/messages "), "{head}");
-    let head_lower = head.to_ascii_lowercase();
-    assert!(head_lower.contains("x-api-key: sk-azure"), "{head}");
-    assert!(
-        head_lower.contains("anthropic-version: 2023-06-01"),
-        "{head}"
-    );
-    assert!(
-        head_lower.contains("content-type: application/json"),
-        "{head}"
-    );
-
-    let sent_body: Value = serde_json::from_str(body).expect("body is json");
-    assert_eq!(
-        sent_body["messages"][0]["content"][0]["cache_control"],
-        json!({"type": "ephemeral"})
-    );
-}
-
-#[tokio::test]
-async fn messages_round_trip_builds_native_anthropic_request() {
-    let listener = TcpListener::bind("127.0.0.1:0").await.expect("binds");
-    let addr = listener.local_addr().expect("addr");
-
-    let server = tokio::spawn(async move {
-        let (mut socket, _) = listener.accept().await.expect("accepts request");
-        let request = read_http_request(&mut socket).await;
-        let response_body = r#"{"id":"msg_1","type":"message","role":"assistant","content":[{"type":"text","text":"hi"}],"model":"claude-sonnet-4-5","stop_reason":"end_turn","usage":{"input_tokens":1,"output_tokens":2}}"#;
-        socket
-            .write_all(write_response(response_body).as_bytes())
-            .await
-            .expect("writes response");
-        request
-    });
-
-    let response = messages(MessagesRequest {
-        model: "claude-sonnet-4-5",
-        body: json!({
-            "model": "claude-sonnet-4-5",
-            "max_tokens": 1024,
-            "messages": [{"role": "user", "content": "hi"}]
-        }),
-        api_key: Some("sk-ant"),
-        api_base: Some(&format!("http://{addr}")),
-        custom_llm_provider: Some("anthropic"),
-        extra_headers: None,
-        timeout: Some(Duration::from_secs(5)),
-    })
-    .await
-    .expect("messages request succeeds");
-
-    assert_eq!(response.content[0]["text"], "hi");
-    assert_eq!(response.stop_reason.as_deref(), Some("end_turn"));
-
-    let request = server.await.expect("server task completes");
-    let (head, _) = request.split_once("\r\n\r\n").expect("has body");
-    assert!(head.starts_with("POST /v1/messages "), "{head}");
-    let head_lower = head.to_ascii_lowercase();
-    assert!(head_lower.contains("x-api-key: sk-ant"), "{head}");
-    assert!(
-        head_lower.contains("anthropic-version: 2023-06-01"),
-        "{head}"
-    );
-}
-
-#[tokio::test]
-async fn messages_does_not_duplicate_auth_when_x_api_key_supplied() {
-    let listener = TcpListener::bind("127.0.0.1:0").await.expect("binds");
-    let addr = listener.local_addr().expect("addr");
-
-    let server = tokio::spawn(async move {
-        let (mut socket, _) = listener.accept().await.expect("accepts request");
-        let request = read_http_request(&mut socket).await;
-        let response_body =
-            r#"{"id":"msg_2","type":"message","role":"assistant","content":[],"model":"m"}"#;
-        socket
-            .write_all(write_response(response_body).as_bytes())
-            .await
-            .expect("writes response");
-        request
-    });
-
-    let mut headers = Map::new();
-    headers.insert(
-        "x-api-key".to_string(),
-        Value::String("from-python".to_string()),
-    );
-    headers.insert(
-        "anthropic-beta".to_string(),
-        Value::String("token-efficient-tools-2025-02-19".to_string()),
-    );
-
-    messages(MessagesRequest {
-        model: "claude-sonnet-4-5",
-        body: json!({"model": "claude-sonnet-4-5", "max_tokens": 8, "messages": []}),
-        api_key: Some("rust-fallback-key"),
-        api_base: Some(&format!("http://{addr}")),
-        custom_llm_provider: Some("azure_ai"),
-        extra_headers: Some(headers),
-        timeout: Some(Duration::from_secs(5)),
-    })
-    .await
-    .expect("messages request succeeds");
-
-    let request = server.await.expect("server task completes");
-    let head = request
-        .split_once("\r\n\r\n")
-        .expect("has body")
-        .0
-        .to_ascii_lowercase();
-    let api_key_count = head
-        .lines()
-        .filter(|line| line.starts_with("x-api-key:"))
-        .count();
-    assert_eq!(api_key_count, 1, "{head}");
-    assert!(head.contains("x-api-key: from-python"), "{head}");
-    assert!(
-        head.contains("anthropic-beta: token-efficient-tools-2025-02-19"),
-        "{head}"
-    );
-    assert!(!head.contains("rust-fallback-key"), "{head}");
-}
-
-#[tokio::test]
-async fn messages_forwards_entra_id_bearer_without_requiring_api_key() {
-    let listener = TcpListener::bind("127.0.0.1:0").await.expect("binds");
-    let addr = listener.local_addr().expect("addr");
-
-    let server = tokio::spawn(async move {
-        let (mut socket, _) = listener.accept().await.expect("accepts request");
-        let request = read_http_request(&mut socket).await;
-        let response_body =
-            r#"{"id":"msg_3","type":"message","role":"assistant","content":[],"model":"m"}"#;
-        socket
-            .write_all(write_response(response_body).as_bytes())
-            .await
-            .expect("writes response");
-        request
-    });
-
-    let mut headers = Map::new();
-    headers.insert(
-        "Authorization".to_string(),
-        Value::String("Bearer entra-token".to_string()),
-    );
-
-    messages(MessagesRequest {
-        model: "claude-sonnet-4-5",
-        body: json!({"model": "claude-sonnet-4-5", "max_tokens": 8, "messages": []}),
-        api_key: None,
-        api_base: Some(&format!("http://{addr}")),
-        custom_llm_provider: Some("azure_ai"),
-        extra_headers: Some(headers),
-        timeout: Some(Duration::from_secs(5)),
-    })
-    .await
-    .expect("entra id request succeeds without api key");
-
-    let request = server.await.expect("server task completes");
-    let head = request
-        .split_once("\r\n\r\n")
-        .expect("has body")
-        .0
-        .to_ascii_lowercase();
-    assert!(head.contains("authorization: bearer entra-token"), "{head}");
-    assert!(!head.contains("x-api-key"), "{head}");
-}
-
-#[tokio::test]
-async fn messages_requires_auth_when_no_key_and_no_header() {
-    let err = messages(MessagesRequest {
-        model: "claude-sonnet-4-5",
-        body: json!({"model": "claude-sonnet-4-5", "max_tokens": 8, "messages": []}),
-        api_key: None,
-        api_base: Some("http://127.0.0.1:1"),
-        custom_llm_provider: Some("azure_ai"),
-        extra_headers: None,
-        timeout: Some(Duration::from_millis(50)),
-    })
-    .await
-    .expect_err("missing auth errors");
-
-    assert!(matches!(err, Error::Auth(_)));
-}
-
-#[tokio::test]
-async fn messages_ignores_malformed_authorization_and_uses_api_key() {
-    let listener = TcpListener::bind("127.0.0.1:0").await.expect("binds");
-    let addr = listener.local_addr().expect("addr");
-
-    let server = tokio::spawn(async move {
-        let (mut socket, _) = listener.accept().await.expect("accepts request");
-        let request = read_http_request(&mut socket).await;
-        let response_body =
-            r#"{"id":"msg_4","type":"message","role":"assistant","content":[],"model":"m"}"#;
-        socket
-            .write_all(write_response(response_body).as_bytes())
-            .await
-            .expect("writes response");
-        request
-    });
-
-    let mut headers = Map::new();
-    headers.insert(
-        "Authorization".to_string(),
-        Value::String("Bearer ".to_string()),
-    );
-
-    messages(MessagesRequest {
-        model: "claude-sonnet-4-5",
-        body: json!({"model": "claude-sonnet-4-5", "max_tokens": 8, "messages": []}),
-        api_key: Some("sk-azure"),
-        api_base: Some(&format!("http://{addr}")),
-        custom_llm_provider: Some("azure_ai"),
-        extra_headers: Some(headers),
-        timeout: Some(Duration::from_secs(5)),
-    })
-    .await
-    .expect("falls back to api key");
-
-    let request = server.await.expect("server task completes");
-    let head = request
-        .split_once("\r\n\r\n")
-        .expect("has body")
-        .0
-        .to_ascii_lowercase();
-    assert!(head.contains("x-api-key: sk-azure"), "{head}");
-}
-
-#[tokio::test]
-async fn messages_maps_provider_error_status_to_http_error() {
-    let listener = TcpListener::bind("127.0.0.1:0").await.expect("binds");
-    let addr = listener.local_addr().expect("addr");
-
-    tokio::spawn(async move {
-        let (mut socket, _) = listener.accept().await.expect("accepts request");
-        let _ = read_http_request(&mut socket).await;
-        let body = "unauthorized";
-        let response = format!(
-            "HTTP/1.1 401 Unauthorized\r\ncontent-length: {}\r\nconnection: close\r\n\r\n{}",
-            body.len(),
-            body
-        );
-        socket
-            .write_all(response.as_bytes())
-            .await
-            .expect("writes response");
-    });
-
-    let err = messages(MessagesRequest {
-        model: "claude-sonnet-4-5",
-        body: json!({"model": "claude-sonnet-4-5", "max_tokens": 8, "messages": []}),
-        api_key: Some("sk-azure"),
-        api_base: Some(&format!("http://{addr}")),
-        custom_llm_provider: Some("azure_ai"),
-        extra_headers: None,
-        timeout: Some(Duration::from_secs(5)),
-    })
-    .await
-    .expect_err("provider error propagates");
-
-    assert!(matches!(
-        err,
-        Error::Transport(litellm_http::transport::Error::Http { status: 401, .. })
-    ));
-}
-
-#[tokio::test]
-async fn messages_rejects_unsupported_provider() {
-    let err = messages(MessagesRequest {
-        model: "claude-3-5-sonnet",
-        body: json!({"model": "claude-3-5-sonnet", "max_tokens": 8, "messages": []}),
-        api_key: Some("sk"),
-        api_base: Some("http://127.0.0.1:1"),
-        custom_llm_provider: Some("openai"),
-        extra_headers: None,
-        timeout: Some(Duration::from_millis(50)),
-    })
-    .await
-    .expect_err("unsupported provider errors");
-
-    assert!(matches!(err, Error::InvalidProvider(provider) if provider == "openai"));
-}

litellm-rust/crates/core/src/messages/types.rs

@@ -1,8 +1,25 @@
 use std::time::Duration;
 
-use litellm_llms::base_llm::anthropic_messages::transformation::BaseAnthropicMessagesConfig;
+use litellm_llms::{
+    anthropic::common_utils::AnthropicModelCapabilities,
+    base_llm::anthropic_messages::transformation::BaseAnthropicMessagesConfig,
+};
+use litellm_types::utils::ProviderSpecificHeaders;
+use serde::{Deserialize, Serialize};
 use serde_json::{Map, Value};
 
+#[derive(Clone, Debug, Default, PartialEq, Serialize, Deserialize)]
+pub struct MessagesShaping {
+    #[serde(default)]
+    pub capabilities: AnthropicModelCapabilities,
+    #[serde(default)]
+    pub drop_params: bool,
+    #[serde(default)]
+    pub reasoning_auto_summary: bool,
+    #[serde(default)]
+    pub additional_drop_params: Vec<String>,
+}
+
 pub struct MessagesRequest<'a> {
     pub model: &'a str,
     pub body: Value,
@@ -10,7 +27,9 @@ pub struct MessagesRequest<'a> {
     pub api_base: Option<&'a str>,
     pub custom_llm_provider: Option<&'a str>,
     pub extra_headers: Option<Map<String, Value>>,
+    pub provider_specific_header: Option<ProviderSpecificHeaders>,
     pub timeout: Option<Duration>,
+    pub shaping: MessagesShaping,
 }
 
 pub struct ProviderMessagesRequest {
@@ -22,3 +41,86 @@ pub struct ProviderMessagesRequest {
     pub upstream_headers: Vec<(String, String)>,
     pub timeout: Option<Duration>,
 }
+
+#[cfg(test)]
+mod tests {
+    use litellm_llms::anthropic::common_utils::SupportedEffortTiers;
+    use rstest::rstest;
+    use serde_json::json;
+
+    use super::*;
+
+    #[rstest]
+    #[case::nothing_projected(json!({}), MessagesShaping::default())]
+    #[case::only_drop_params(
+        json!({"drop_params": true}),
+        MessagesShaping { drop_params: true, ..MessagesShaping::default() },
+    )]
+    #[case::only_reasoning_auto_summary(
+        json!({"reasoning_auto_summary": true}),
+        MessagesShaping { reasoning_auto_summary: true, ..MessagesShaping::default() },
+    )]
+    #[case::only_additional_drop_params(
+        json!({"additional_drop_params": ["tools[*].input_examples"]}),
+        MessagesShaping {
+            additional_drop_params: vec!["tools[*].input_examples".to_string()],
+            ..MessagesShaping::default()
+        },
+    )]
+    #[case::partial_capabilities(
+        json!({"capabilities": {"supports_reasoning": true}}),
+        MessagesShaping {
+            capabilities: AnthropicModelCapabilities {
+                supports_reasoning: true,
+                ..AnthropicModelCapabilities::default()
+            },
+            ..MessagesShaping::default()
+        },
+    )]
+    #[case::everything_the_python_host_projects(
+        json!({
+            "capabilities": {
+                "supports_reasoning": true,
+                "supports_adaptive_thinking": true,
+                "thinking_always_on": false,
+                "supports_legacy_thinking": false,
+                "supports_output_config": true,
+                "supports_sampling_params": false,
+                "supports_speed": true,
+                "effort_tiers": {"minimal": false, "low": true, "medium": true, "high": true, "xhigh": true, "max": false}
+            },
+            "drop_params": true,
+            "reasoning_auto_summary": true,
+            "additional_drop_params": ["metadata.user_id", "thinking"]
+        }),
+        MessagesShaping {
+            capabilities: AnthropicModelCapabilities {
+                supports_reasoning: true,
+                supports_adaptive_thinking: true,
+                thinking_always_on: false,
+                supports_legacy_thinking: false,
+                supports_output_config: true,
+                supports_sampling_params: false,
+                supports_speed: true,
+                effort_tiers: SupportedEffortTiers {
+                    minimal: false,
+                    low: true,
+                    medium: true,
+                    high: true,
+                    xhigh: true,
+                    max: false,
+                },
+            },
+            drop_params: true,
+            reasoning_auto_summary: true,
+            additional_drop_params: vec!["metadata.user_id".to_string(), "thinking".to_string()],
+        },
+    )]
+    fn shaping_deserializes_with_defaults_for_absent_fields(
+        #[case] projected: Value,
+        #[case] expected: MessagesShaping,
+    ) {
+        let shaping: MessagesShaping = serde_json::from_value(projected).unwrap();
+        assert_eq!(shaping, expected);
+    }
+}

litellm-rust/crates/core/src/ocr/document.rs

@@ -23,9 +23,6 @@ pub fn prepare_document(input: OcrDocumentInput) -> Result<OcrDocument, Error> {
             file_name.as_deref(),
             mime_type.as_deref(),
         )?),
-        OcrDocumentInput::HostReader { .. } => Err(Error::InvalidRequest(
-            "OCR file reader was not read by the host".into(),
-        )),
     }
 }
 
@@ -207,7 +204,7 @@ mod tests {
     }
 
     #[test]
-    fn byte_documents_are_encoded_and_host_readers_must_be_read_first() {
+    fn byte_documents_are_encoded() {
         assert_eq!(
             prepare_document(OcrDocumentInput::Bytes {
                 bytes: b"abc".as_slice().into(),
@@ -217,7 +214,6 @@ mod tests {
             .unwrap(),
             document("data:application/pdf;base64,YWJj")
         );
-        assert!(prepare_document(OcrDocumentInput::HostReader { mime_type: None }).is_err());
     }
 
     #[test]

litellm-rust/crates/core/src/ocr/mod.rs

@@ -7,38 +7,3 @@ pub mod provider_config;
 pub mod route;
 pub mod types;
 pub mod wire;
-
-#[cfg(test)]
-#[path = "../../tests/aws_textract_ocr.rs"]
-mod aws_textract_tests;
-
-#[cfg(test)]
-#[path = "../../tests/azure_ai_ocr.rs"]
-mod azure_ai_tests;
-#[cfg(test)]
-#[path = "../../tests/azure_document_intelligence_ocr.rs"]
-mod azure_document_intelligence_tests;
-#[cfg(test)]
-#[path = "../../tests/cohere_ocr.rs"]
-mod cohere_tests;
-#[cfg(test)]
-#[path = "../../tests/deepseek_ocr.rs"]
-mod deepseek_tests;
-#[cfg(test)]
-#[path = "../../tests/ocr/document.rs"]
-mod document_tests;
-#[cfg(test)]
-#[path = "../../tests/reducto_ocr.rs"]
-mod reducto_tests;
-#[cfg(test)]
-#[path = "../../tests/ocr/support.rs"]
-pub(crate) mod test_support;
-#[cfg(test)]
-#[path = "../../tests/ocr.rs"]
-pub(crate) mod tests;
-#[cfg(test)]
-#[path = "../../tests/vertex_ai_deepseek_ocr.rs"]
-mod vertex_ai_deepseek_tests;
-#[cfg(test)]
-#[path = "../../tests/vertex_ai_ocr.rs"]
-mod vertex_ai_tests;

litellm-rust/crates/core/src/ocr/prepare.rs

@@ -1,11 +1,9 @@
 use litellm_auth::{InputSource, SecretValue, Sourced};
-use litellm_llms::base_llm::{
-    inference::secrets::Secrets,
-    ocr::{
-        handler::OcrClient,
-        transformation::{OcrConnection, OcrCredentialInputs, PreparedOcrRequest},
-    },
+use litellm_llms::base_llm::ocr::{
+    handler::OcrClient,
+    transformation::{OcrConnection, OcrCredentialInputs, PreparedOcrRequest},
 };
+use litellm_secrets::source::Secrets;
 
 use super::provider_config::OcrProvider;
 use crate::ocr::types::{LiteLLMOcrRequest, ResolvedOcrRequest};
@@ -72,20 +70,203 @@ pub(crate) fn prepare_request(
     }
 }
 
-#[cfg(test)]
-pub(crate) fn prepare_request_for_test(request: ResolvedOcrRequest) -> PreparedOcrRequest {
-    prepare_request(
-        request,
-        true,
-        &OcrClient::for_test(reqwest::Client::new(), reqwest::Client::new()),
-        std::sync::Arc::new(litellm_core_utils::settings::ProcessEnvironment),
-    )
-}
-
 #[cfg(test)]
 mod tests {
+    use std::time::Duration;
+
+    use futures_util::future::BoxFuture;
     use litellm_core_utils::call_arguments::{CallArguments, compose_body, parse_options};
-    use serde_json::json;
+    use litellm_host::event::WireRequest;
+    use litellm_llms::{
+        base_llm::ocr::{
+            error::Error,
+            handler::{CallHooks, OcrClient},
+            transformation::{BaseOcrConfig, OcrResponseFormat},
+        },
+        cohere::ocr::transformation::CohereParseConfig,
+        mistral::ocr::transformation::MistralOcrConfig,
+        vertex_ai::ocr::transformation::VertexAiOcrConfig,
+    };
+    use serde_json::{Value, json};
+
+    use super::*;
+    use crate::ocr::{
+        document::prepare_document,
+        types::LiteLLMOcrRequest,
+        wire::{OcrWireRequest, decode_request},
+    };
+
+    /// Stands in for a host with no hooks registered.
+    struct NoHooks;
+
+    impl CallHooks<Error> for NoHooks {
+        fn before_send(&self, wire: WireRequest) -> BoxFuture<'_, Result<WireRequest, Error>> {
+            Box::pin(async move { Ok(wire) })
+        }
+
+        fn response_received<'a>(&'a self, _body: &'a [u8]) -> BoxFuture<'a, Result<(), Error>> {
+            Box::pin(async { Ok(()) })
+        }
+    }
+
+    fn client() -> OcrClient {
+        OcrClient::for_test(reqwest::Client::new(), reqwest::Client::new())
+    }
+
+    fn request(model: &str, base: &str, document: Value, options: Value) -> LiteLLMOcrRequest {
+        decode_request(OcrWireRequest {
+            model: model.into(),
+            document,
+            api_key: Some(litellm_auth::SecretValue::new("test-key")),
+            api_base: Some(base.into()),
+            custom_llm_provider: None,
+            extra_headers: None,
+            optional_params: options.as_object().unwrap().clone(),
+            input_sources: Default::default(),
+            timeout_seconds: Some(2.0),
+        })
+        .unwrap()
+    }
+
+    fn prepared(request: LiteLLMOcrRequest) -> PreparedOcrRequest {
+        prepare_request(
+            request.map_document(prepare_document).unwrap(),
+            true,
+            &client(),
+            std::sync::Arc::new(litellm_core_utils::settings::ProcessEnvironment),
+        )
+    }
+
+    fn image(url: &str) -> Value {
+        json!({"type": "image_url", "image_url": url})
+    }
+
+    #[tokio::test]
+    async fn cohere_body_keeps_native_document_fields_and_untyped_overrides() {
+        let request = request(
+            "cohere/parse",
+            "https://example.com",
+            image("https://example.com/original.png"),
+            json!({
+                "output_format": "markdown", "timeout": 30,
+                "extra_body": {
+                    "output_format": {"future": true},
+                    "document": {"type": "image_url", "image_url": "https://example.com/a.png",
+                        "provider_options": {"nested": [false, 0, null]}}
+                }
+            }),
+        );
+
+        let http = CohereParseConfig
+            .prepare_request(&prepared(request), &client(), &NoHooks)
+            .await
+            .unwrap();
+
+        let body: Value = serde_json::from_slice(http.body()).unwrap();
+        assert_eq!(
+            body,
+            json!({
+                "model": "parse", "output_format": {"future": true},
+                "document": {"type": "image_url", "image_url": "https://example.com/a.png",
+                    "provider_options": {"nested": [false, 0, null]}}
+            })
+        );
+    }
+
+    #[tokio::test]
+    async fn explicit_null_options_use_defaults_before_http() {
+        let request = request(
+            "cohere/parse",
+            "https://example.com",
+            image("https://example.com/a.png"),
+            json!({"output_format": null, "req_format": null}),
+        );
+        assert_eq!(
+            request.response_format().unwrap(),
+            OcrResponseFormat::Litellm
+        );
+
+        let http = CohereParseConfig
+            .prepare_request(&prepared(request), &client(), &NoHooks)
+            .await
+            .unwrap();
+
+        let body: Value = serde_json::from_slice(http.body()).unwrap();
+        assert_eq!(body["output_format"], "markdown");
+        assert!(body.get("req_format").is_none());
+    }
+
+    #[tokio::test]
+    async fn direct_and_vertex_mistral_build_the_same_request_and_share_normalization() {
+        let options = json!({
+            "pages": [0, 2],
+            "include_image_base64": true,
+            "vertex_project": "project-1",
+            "vertex_location": "us-central1",
+            "unknown": "preserved"
+        });
+        let document =
+            json!({"type": "document_url", "document_url": "data:application/pdf;base64,YWJj"});
+        let direct = prepared(request(
+            "mistral/mistral-ocr-maas",
+            "https://mistral.test",
+            document.clone(),
+            options.clone(),
+        ));
+        let vertex = prepared(request(
+            "vertex_ai/mistral-ocr-maas",
+            "https://vertex.test",
+            document,
+            options,
+        ));
+
+        let direct_http = MistralOcrConfig
+            .prepare_request(&direct, &client(), &NoHooks)
+            .await
+            .unwrap();
+        let vertex_http = VertexAiOcrConfig
+            .prepare_request(&vertex, &client(), &NoHooks)
+            .await
+            .unwrap();
+
+        assert_eq!(direct_http.url(), "https://mistral.test/v1/ocr");
+        assert_eq!(
+            vertex_http.url(),
+            "https://vertex.test/v1/projects/project-1/locations/us-central1/publishers/mistralai/models/mistral-ocr-maas:rawPredict"
+        );
+        for http in [&direct_http, &vertex_http] {
+            assert_eq!(http.header("authorization").unwrap(), "Bearer test-key");
+            assert_eq!(http.header("content-type").unwrap(), "application/json");
+            assert_eq!(http.timeout(), Some(Duration::from_secs(2)));
+            let body: Value = serde_json::from_slice(http.body()).unwrap();
+            assert_eq!(
+                body,
+                json!({
+                    "model": "mistral-ocr-maas",
+                    "document": {"type": "document_url", "document_url": "data:application/pdf;base64,YWJj"},
+                    "pages": [0, 2],
+                    "include_image_base64": true,
+                    "unknown": "preserved"
+                })
+            );
+        }
+        let payload = serde_json::to_vec(
+            &json!({"pages": [{"index": 0, "markdown": "hello"}], "extra": "preserved"}),
+        )
+        .unwrap();
+        let direct_response = MistralOcrConfig
+            .transform_ocr_response(&direct.model, &payload, OcrResponseFormat::Litellm)
+            .unwrap()
+            .into_json();
+        let vertex_response = VertexAiOcrConfig
+            .transform_ocr_response(&vertex.model, &payload, OcrResponseFormat::Litellm)
+            .unwrap()
+            .into_json();
+        assert_eq!(direct_response, vertex_response);
+        assert_eq!(direct_response["model"], "mistral-ocr-maas");
+        assert_eq!(direct_response["object"], "ocr");
+        assert_eq!(direct_response["extra"], "preserved");
+    }
 
     #[derive(serde::Deserialize)]
     struct KnownParams {

litellm-rust/crates/core/src/ocr/provider_config.rs

@@ -5,15 +5,15 @@ use litellm_llms::{
         transformation::TextractDetectTextConfig,
     },
     azure_ai::ocr::{
-        cohere_parse_transformation::AzureAICohereParseConfig,
+        cohere_parse_transformation::{AZURE_COHERE_PARSE_PATH, AzureAICohereParseConfig},
         document_intelligence::transformation::AzureDocumentIntelligenceOcrConfig,
-        transformation::AzureAiOcrConfig,
+        transformation::{AZURE_AI_OCR_PATH, AzureAiOcrConfig},
     },
     base_llm::ocr::{
         error::Error,
         handler::{self, CallHooks, OcrClient},
         transformation::{
-            BaseOcrConfig, LiteLLMOcrResponse, OcrCredentialInputs, OcrDocument,
+            BaseOcrConfig, LiteLLMOcrResponse, OcrCredentialInputs, OcrDocument, OcrResponseFormat,
             PreparedOcrRequest, ResolvedOcrCredentials,
         },
     },
@@ -157,6 +157,36 @@ pub fn get_health_check_document(
         .get_health_check_document())
 }
 
+/// Normalize a relayed Azure AI response into the LiteLLM OCR shape when
+/// `endpoint` is the OCR route of the model's resolved config.
+pub fn passthrough_response(
+    model: &str,
+    endpoint: &str,
+    body: &[u8],
+) -> Result<Option<LiteLLMOcrResponse>, Error> {
+    let (model, config) = resolve_provider_config(model, Some("azure_ai"))?;
+    let segments: Vec<&str> = endpoint
+        .split('/')
+        .filter(|segment| !segment.is_empty())
+        .collect();
+    let is_ocr_endpoint = match config {
+        OcrConfigKind::AzureAi => segments == AZURE_AI_OCR_PATH,
+        OcrConfigKind::AzureCohere => segments == AZURE_COHERE_PARSE_PATH,
+        OcrConfigKind::AzureDocumentIntelligence => {
+            segments == AzureDocumentIntelligenceOcrConfig::analyze_path(&model)?
+        }
+        other => {
+            let provider: &'static str = other.provider().into();
+            return Err(Error::InvalidProvider(provider.to_owned()));
+        }
+    };
+    if !is_ocr_endpoint {
+        return Ok(None);
+    }
+    with_config!(config, config => config.transform_ocr_response(&model, body, OcrResponseFormat::Litellm))
+        .map(Some)
+}
+
 #[derive(Clone, Copy, Debug, EnumString, IntoStaticStr, PartialEq, Eq)]
 #[strum(serialize_all = "snake_case")]
 pub(crate) enum OcrProvider {
@@ -520,4 +550,68 @@ mod tests {
         assert!(matches!(&error, Error::InvalidProvider(provider) if provider == "not_a_provider"));
         assert_eq!(error.http_status_code(), Some(400));
     }
+
+    #[rstest]
+    #[case("azure_ai/mistral-document-ai-2512", "providers/mistral/azure/ocr")]
+    #[case("azure_ai/mistral-document-ai-2512", "/providers/mistral/azure/ocr/")]
+    #[case("azure_ai/Cohere-parse-v5", "providers/cohere/v2/parse")]
+    #[case(
+        "azure_ai/doc-intelligence/prebuilt-layout",
+        "documentintelligence/documentModels/prebuilt-layout:analyze"
+    )]
+    fn passthrough_response_recognizes_the_resolved_config_ocr_route(
+        #[case] model: &str,
+        #[case] endpoint: &str,
+    ) {
+        assert!(passthrough_response(model, endpoint, b"not json").is_err());
+    }
+
+    #[rstest]
+    #[case("azure_ai/mistral-document-ai-2512", "models/info")]
+    #[case("azure_ai/mistral-document-ai-2512", "providers/cohere/v2/parse")]
+    #[case("azure_ai/Cohere-parse-v5", "providers/mistral/azure/ocr")]
+    #[case(
+        "azure_ai/doc-intelligence/prebuilt-layout",
+        "documentintelligence/documentModels/prebuilt-read:analyze"
+    )]
+    fn passthrough_response_skips_other_routes(#[case] model: &str, #[case] endpoint: &str) {
+        assert!(
+            passthrough_response(model, endpoint, b"not json")
+                .unwrap()
+                .is_none()
+        );
+    }
+
+    #[test]
+    fn passthrough_response_normalizes_the_mistral_body() {
+        let body = br#"{
+            "pages": [{"index": 0, "markdown": "page one"}, {"index": 1, "markdown": "page two"}],
+            "model": "mistral-document-ai-2512",
+            "usage_info": {"pages_processed": 2}
+        }"#;
+        let json = passthrough_response(
+            "azure_ai/mistral-document-ai-2512",
+            "providers/mistral/azure/ocr",
+            body,
+        )
+        .unwrap()
+        .unwrap()
+        .into_json();
+        assert_eq!(json["usage_info"]["pages_processed"], 2);
+        assert_eq!(json["pages"][0]["markdown"], "page one");
+    }
+
+    #[test]
+    fn passthrough_response_counts_cohere_billed_pages() {
+        let body = br#"{"id": "parse-1", "pages": [], "meta": {"billed_units": {"pages": 3}}}"#;
+        let json = passthrough_response(
+            "azure_ai/Cohere-parse-v5",
+            "providers/cohere/v2/parse",
+            body,
+        )
+        .unwrap()
+        .unwrap()
+        .into_json();
+        assert_eq!(json["usage_info"]["pages_processed"], 3);
+    }
 }

litellm-rust/crates/core/src/ocr/route.rs

@@ -3,102 +3,73 @@ use std::sync::{Arc, Mutex};
 use litellm_auth::ResolvedCredential;
 use litellm_host::{
     event::{CallEvent, RequestContext, WireRequest},
-    machine::{HostChannel, HostTokenProvider, MachineFault, RouteMachine, TokenRoute},
-    route::Route,
+    host::Reply,
+    machine::{CallMachine, HostChannel, HostTokenProvider, TokenProtocol},
+    protocol::Protocol,
 };
 use litellm_llms::base_llm::ocr::{
     error::Error, handler::OcrClient, transformation::LiteLLMOcrResponse,
 };
 
 use super::handler::perform_ocr_request;
-use crate::ocr::types::{LiteLLMOcrRequest, OcrDocumentInput, OcrFileContent, ResolvedOcrRequest};
+use crate::ocr::types::{LiteLLMOcrRequest, OcrDocumentInput, ResolvedOcrRequest};
 
-#[derive(Clone, Copy, Debug, PartialEq, Eq)]
 pub enum OcrOp {
-    ProjectRequest,
-    ReadDocument,
-    AcquireAzureAdToken,
+    AcquireAzureAdToken(Reply<ResolvedCredential>),
 }
 
-pub enum OcrOpResult {
-    Request {
-        request: Box<LiteLLMOcrRequest<OcrDocumentInput>>,
-        caller_token: bool,
-    },
-    Document(OcrFileContent),
-    AzureAdToken(ResolvedCredential),
+/// The caller's request as the host projects it.
+pub struct OcrProjection {
+    pub request: LiteLLMOcrRequest<OcrDocumentInput>,
+    /// The caller passed its own Azure AD token provider, which the host keeps.
+    pub caller_token: bool,
 }
 
 pub struct Ocr;
 
-impl Route for Ocr {
+impl Protocol for Ocr {
     type Response = LiteLLMOcrResponse;
     type Error = Error;
+    type Projection = OcrProjection;
     type Op = OcrOp;
-    type OpResult = OcrOpResult;
     type Chunk = std::convert::Infallible;
     type StreamHead = std::convert::Infallible;
 }
 
-impl TokenRoute for Ocr {
-    fn acquire_token_op() -> OcrOp {
-        OcrOp::AcquireAzureAdToken
-    }
-
-    fn token_credential(result: OcrOpResult) -> Option<ResolvedCredential> {
-        match result {
-            OcrOpResult::AzureAdToken(credential) => Some(credential),
-            _ => None,
-        }
+impl TokenProtocol for Ocr {
+    fn acquire_token_op(reply: Reply<ResolvedCredential>) -> OcrOp {
+        OcrOp::AcquireAzureAdToken(reply)
     }
 }
 
 pub type OcrHost = HostChannel<Ocr>;
-pub type OcrMachine = RouteMachine<Ocr>;
+pub type OcrMachine = CallMachine<Ocr>;
 
-/// The OCR call as a machine: projection, document reading and token acquisition are
-/// host operations; everything else runs in Rust.
+/// The OCR call as a machine: projection and token acquisition are host operations;
+/// everything else runs in Rust.
 pub fn ocr_machine(client: OcrClient) -> OcrMachine {
-    RouteMachine::new(move |host| Box::pin(execute(client, host)))
+    CallMachine::new(move |host| Box::pin(execute(client, host)))
 }
 
 async fn execute(client: OcrClient, host: OcrHost) -> Result<LiteLLMOcrResponse, Error> {
-    let OcrOpResult::Request {
+    let OcrProjection {
         request,
         caller_token,
-    } = host.route(OcrOp::ProjectRequest).await?
-    else {
-        return Err(MachineFault::Mismatch.into());
-    };
+    } = host.project().await?;
     let request = LiteLLMOcrRequest {
         azure_ad_token_provider: caller_token
             .then(|| HostTokenProvider::handle(host.clone()))
             .or(request.azure_ad_token_provider),
-        ..*request
+        ..request
     };
     let caller_document = matches!(request.document, OcrDocumentInput::Document(_));
-    let request = prepare_request_document(request, &host).await?;
+    let request = prepare_request_document(request).await?;
     perform_ocr_request(&client, request, &host, caller_document).await
 }
 
 async fn prepare_request_document(
     request: LiteLLMOcrRequest<OcrDocumentInput>,
-    host: &OcrHost,
 ) -> Result<ResolvedOcrRequest, Error> {
-    let request = match &request.document {
-        OcrDocumentInput::HostReader { mime_type } => {
-            let mime_type = mime_type.clone();
-            let OcrOpResult::Document(content) = host.route(OcrOp::ReadDocument).await? else {
-                return Err(MachineFault::Mismatch.into());
-            };
-            request.with_document(OcrDocumentInput::Bytes {
-                bytes: content.bytes,
-                file_name: content.file_name,
-                mime_type,
-            })
-        }
-        _ => request,
-    };
     if let OcrDocumentInput::Document(_) = &request.document {
         return request.map_document(super::document::prepare_document);
     }
@@ -107,7 +78,6 @@ async fn prepare_request_document(
         .map_err(|error| Error::DocumentTask(Arc::new(error)))?
 }
 
-type Reader = Box<dyn Fn() -> Result<OcrFileContent, Error> + Send + Sync>;
 type BeforeSend =
     Box<dyn Fn(WireRequest, &RequestContext) -> Result<WireRequest, Error> + Send + Sync>;
 type Observer = Box<dyn Fn(&CallEvent) + Send + Sync>;
@@ -116,7 +86,6 @@ type Observer = Box<dyn Fn(&CallEvent) + Send + Sync>;
 /// projection, and the optional observer sees and may rewrite the wire request.
 pub struct LocalOcrHost {
     request: Mutex<Option<LiteLLMOcrRequest<OcrDocumentInput>>>,
-    reader: Option<Reader>,
     before_send: Option<BeforeSend>,
     observer: Option<Observer>,
 }
@@ -125,22 +94,11 @@ impl LocalOcrHost {
     pub fn new(request: LiteLLMOcrRequest<OcrDocumentInput>) -> Self {
         Self {
             request: Mutex::new(Some(request)),
-            reader: None,
             before_send: None,
             observer: None,
         }
     }
 
-    pub fn with_reader(
-        self,
-        reader: impl Fn() -> Result<OcrFileContent, Error> + Send + Sync + 'static,
-    ) -> Self {
-        Self {
-            reader: Some(Box::new(reader)),
-            ..self
-        }
-    }
-
     pub fn with_before_send(
         self,
         before_send: impl Fn(WireRequest, &RequestContext) -> Result<WireRequest, Error>
@@ -163,25 +121,21 @@ impl LocalOcrHost {
 }
 
 impl litellm_host::host::Host<Ocr> for LocalOcrHost {
-    async fn route(&self, op: OcrOp) -> Result<OcrOpResult, Error> {
+    async fn project(&self) -> Result<OcrProjection, Error> {
+        self.request
+            .lock()
+            .unwrap_or_else(|error| error.into_inner())
+            .take()
+            .map(|request| OcrProjection {
+                request,
+                caller_token: false,
+            })
+            .ok_or_else(|| Error::InvalidRequest("OCR request was already projected".into()))
+    }
+
+    async fn custom_op(&self, op: OcrOp) -> Result<(), Error> {
         match op {
-            OcrOp::ProjectRequest => self
-                .request
-                .lock()
-                .unwrap_or_else(|error| error.into_inner())
-                .take()
-                .map(|request| OcrOpResult::Request {
-                    request: Box::new(request),
-                    caller_token: false,
-                })
-                .ok_or_else(|| Error::InvalidRequest("OCR request was already projected".into())),
-            OcrOp::ReadDocument => self
-                .reader
-                .as_ref()
-                .ok_or_else(|| Error::InvalidRequest("OCR host has no document reader".into()))
-                .and_then(|reader| reader())
-                .map(OcrOpResult::Document),
-            OcrOp::AcquireAzureAdToken => {
+            OcrOp::AcquireAzureAdToken(_) => {
                 Err(Error::Auth(litellm_auth::Error::AzureTokenAcquisition(
                     "OCR host has no Azure AD token provider".into(),
                 )))

litellm-rust/crates/core/src/ocr/types.rs

@@ -25,9 +25,6 @@ pub enum OcrDocumentInput {
         file_name: Option<String>,
         mime_type: Option<String>,
     },
-    HostReader {
-        mime_type: Option<String>,
-    },
 }
 
 impl From<OcrDocument> for OcrDocumentInput {
@@ -45,12 +42,6 @@ impl From<PathBuf> for OcrDocumentInput {
     }
 }
 
-#[derive(Clone, Debug, PartialEq, Eq)]
-pub struct OcrFileContent {
-    pub bytes: Bytes,
-    pub file_name: Option<String>,
-}
-
 /// Caller-supplied connection overrides for a [`LiteLLMOcrRequest`], in the
 /// shape hosts receive them: JSON-ish headers, optional timeout, optional
 /// credentials, and per-field provenance in `input_sources`.

litellm-rust/crates/core/tests/audio_transcription.rs

@@ -0,0 +1,250 @@
+use litellm_core::audio_transcription::{
+    Error, audio_transcription, types::AudioTranscriptionRequest,
+};
+use rstest::{fixture, rstest};
+use serde_json::{Map, Value, json};
+use wiremock::ResponseTemplate;
+
+mod support;
+use support::*;
+
+const MODEL: &str = "mistral.voxtral-mini-3b-2507";
+
+fn transcript_response(text: &str) -> ResponseTemplate {
+    json_response(json!({"output": {"message": {"content": [{"text": text}]}}}))
+}
+
+fn aws_params(region: &str) -> Map<String, Value> {
+    Map::from_iter([
+        ("aws_access_key_id".to_string(), json!("access-key")),
+        ("aws_secret_access_key".to_string(), json!("secret-key")),
+        ("aws_region_name".to_string(), json!(region)),
+    ])
+}
+
+#[fixture]
+fn request() -> AudioTranscriptionRequest<'static> {
+    AudioTranscriptionRequest {
+        model: MODEL,
+        audio: json!({"data": "AQI=", "format": "wav", "filename": "audio.wav"}),
+        api_key: None,
+        api_base: None,
+        custom_llm_provider: Some("bedrock"),
+        extra_headers: None,
+        optional_params: aws_params("us-east-1"),
+        timeout: None,
+    }
+}
+
+#[rstest]
+#[case::us_east_1("us-east-1")]
+#[case::eu_west_1("eu-west-1")]
+#[tokio::test]
+async fn bedrock_converse_request_is_signed_for_the_requested_region(
+    request: AudioTranscriptionRequest<'static>,
+    #[case] region: &str,
+) {
+    let upstream = upstream([transcript_response("hello")]).await;
+    let base = upstream.uri();
+
+    let response = audio_transcription(AudioTranscriptionRequest {
+        api_base: Some(&base),
+        optional_params: aws_params(region),
+        ..request
+    })
+    .await
+    .expect("transcription");
+
+    assert_eq!(response, json!({"text": "hello"}));
+    let sent = only_request(&upstream).await;
+    assert_eq!(sent.method.as_str(), "POST");
+    assert_eq!(sent.url.path(), format!("/model/{MODEL}/converse"));
+    let authorization = sent.header("authorization").expect("request is signed");
+    assert!(
+        authorization.starts_with("AWS4-HMAC-SHA256 Credential=access-key/"),
+        "{authorization}"
+    );
+    assert!(
+        authorization.contains(&format!("/{region}/bedrock/aws4_request")),
+        "{authorization}"
+    );
+    assert!(sent.header("x-amz-date").is_some());
+    assert!(!sent.body_text().contains("secret-key"));
+}
+
+#[rstest]
+#[tokio::test]
+async fn the_provider_can_come_from_the_model_prefix(request: AudioTranscriptionRequest<'static>) {
+    let upstream = upstream([transcript_response("hello")]).await;
+    let base = upstream.uri();
+    let model = format!("bedrock/{MODEL}");
+
+    audio_transcription(AudioTranscriptionRequest {
+        model: &model,
+        custom_llm_provider: None,
+        api_base: Some(&base),
+        ..request
+    })
+    .await
+    .expect("transcription");
+
+    assert_eq!(
+        only_request(&upstream).await.url.path(),
+        format!("/model/{MODEL}/converse")
+    );
+}
+
+#[rstest]
+#[tokio::test]
+async fn audio_and_transcription_params_reach_the_converse_body(
+    request: AudioTranscriptionRequest<'static>,
+    #[values("wav", "mp3", "flac", "ogg")] format: &str,
+) {
+    let upstream = upstream([transcript_response("hello")]).await;
+    let base = upstream.uri();
+    let optional_params = aws_params("us-east-1")
+        .into_iter()
+        .chain([
+            ("language".to_string(), json!("fr")),
+            ("temperature".to_string(), json!(0.2)),
+        ])
+        .collect();
+
+    audio_transcription(AudioTranscriptionRequest {
+        audio: json!({"data": "AQI=", "format": format}),
+        api_base: Some(&base),
+        optional_params,
+        ..request
+    })
+    .await
+    .expect("transcription");
+
+    let body = only_request(&upstream).await.json();
+    let content = &body["messages"][0]["content"];
+    assert_eq!(
+        content[0],
+        json!({"audio": {"format": format, "source": {"bytes": "AQI="}}})
+    );
+    let instruction = content[1]["text"].as_str().expect("instruction text");
+    assert!(instruction.contains("fr"), "{instruction}");
+    assert_eq!(body["inferenceConfig"]["temperature"], 0.2);
+}
+
+#[rstest]
+#[case::unknown_format(json!({"data": "AQI=", "format": "aac"}))]
+#[case::missing_data(json!({"format": "wav"}))]
+#[case::not_an_object(json!("AQI="))]
+#[tokio::test]
+async fn invalid_audio_is_rejected_before_sending(
+    request: AudioTranscriptionRequest<'static>,
+    #[case] audio: Value,
+) {
+    let upstream = upstream([transcript_response("hello")]).await;
+    let base = upstream.uri();
+
+    let error = audio_transcription(AudioTranscriptionRequest {
+        audio,
+        api_base: Some(&base),
+        ..request
+    })
+    .await
+    .expect_err("invalid audio is rejected");
+
+    assert!(
+        matches!(
+            error,
+            Error::InvalidRequest(_) | Error::MissingField(_) | Error::InvalidType { .. }
+        ),
+        "{error:?}"
+    );
+    assert!(received(&upstream).await.is_empty());
+}
+
+#[rstest]
+#[case::unknown_provider(MODEL, Some("openai"), "openai")]
+#[case::unresolvable_model(
+    "no-such-model",
+    None,
+    "unable to resolve custom_llm_provider for audio transcription request"
+)]
+#[tokio::test]
+async fn unsupported_providers_are_rejected_before_sending(
+    request: AudioTranscriptionRequest<'static>,
+    #[case] model: &'static str,
+    #[case] provider: Option<&'static str>,
+    #[case] reported: &str,
+) {
+    let error = audio_transcription(AudioTranscriptionRequest {
+        model,
+        custom_llm_provider: provider,
+        api_base: Some(UNREACHABLE_BASE),
+        ..request
+    })
+    .await
+    .expect_err("unsupported provider errors");
+
+    assert_eq!(error, Error::InvalidProvider(reported.into()));
+}
+
+#[rstest]
+#[tokio::test]
+async fn a_non_string_extra_header_is_rejected(request: AudioTranscriptionRequest<'static>) {
+    let error = audio_transcription(AudioTranscriptionRequest {
+        extra_headers: Some(Map::from_iter([("x-count".to_string(), json!(3))])),
+        api_base: Some(UNREACHABLE_BASE),
+        ..request
+    })
+    .await
+    .expect_err("a non-string header is rejected");
+
+    assert!(matches!(error, Error::Headers(_)), "{error:?}");
+}
+
+#[rstest]
+#[case::throttled(429)]
+#[case::server_error(500)]
+#[tokio::test]
+async fn an_upstream_error_keeps_its_status_and_body(
+    request: AudioTranscriptionRequest<'static>,
+    #[case] status: u16,
+) {
+    let upstream =
+        upstream([ResponseTemplate::new(status).set_body_string("upstream said no")]).await;
+    let base = upstream.uri();
+
+    let error = audio_transcription(AudioTranscriptionRequest {
+        api_base: Some(&base),
+        ..request
+    })
+    .await
+    .expect_err("upstream error propagates");
+
+    assert_eq!(
+        error,
+        Error::Transport(litellm_http::transport::Error::Http {
+            status,
+            body: "upstream said no".into()
+        })
+    );
+}
+
+#[rstest]
+#[case::not_json(ResponseTemplate::new(200).set_body_string("not json"))]
+#[case::no_output(json_response(json!({"unexpected": true})))]
+#[tokio::test]
+async fn an_unreadable_success_body_is_an_invalid_response(
+    request: AudioTranscriptionRequest<'static>,
+    #[case] response: ResponseTemplate,
+) {
+    let upstream = upstream([response]).await;
+    let base = upstream.uri();
+
+    let error = audio_transcription(AudioTranscriptionRequest {
+        api_base: Some(&base),
+        ..request
+    })
+    .await
+    .expect_err("an unreadable body fails");
+
+    assert!(matches!(error, Error::InvalidResponse(_)), "{error:?}");
+}

litellm-rust/crates/core/tests/aws_textract_ocr.rs

@@ -1,193 +0,0 @@
-use std::{collections::BTreeMap, time::SystemTime};
-
-use litellm_auth_aws::{Credentials, aws_signature_headers, sign_post};
-use litellm_llms::base_llm::ocr::error::Error;
-use serde_json::{Value, json};
-use time::{PrimitiveDateTime, format_description};
-
-use crate::ocr::{
-    route::LocalOcrHost,
-    test_support::{
-        MockResponse, header, mock_server, perform_ocr_with, request_body,
-        wire_request_with_document,
-    },
-    types::LiteLLMOcrRequest,
-};
-
-const ACCESS_KEY_ID: &str = "AKIDEXAMPLE";
-const SECRET_ACCESS_KEY: &str = "wJalrXUtnFEMI/K7MDENG+bPxRfiCYEXAMPLEKEY";
-
-fn textract_request(base: &str) -> LiteLLMOcrRequest {
-    textract_request_for("aws_textract/detect-document-text", base)
-}
-
-fn textract_request_for(model: &str, base: &str) -> LiteLLMOcrRequest {
-    wire_request_with_document(
-        model,
-        &format!("{base}/"),
-        json!({"type": "image_url", "image_url": "data:image/png;base64,b3JpZ2luYWw="}),
-        json!({
-            "aws_access_key_id": ACCESS_KEY_ID,
-            "aws_secret_access_key": SECRET_ACCESS_KEY,
-            "aws_region_name": "eu-west-1"
-        }),
-    )
-}
-
-fn textract_response() -> MockResponse {
-    MockResponse::json(json!({
-        "DocumentMetadata": {"Pages": 1},
-        "Blocks": [{"BlockType": "PAGE"}, {"BlockType": "LINE", "Text": "Invoice 12345"}]
-    }))
-}
-
-/// Recomputes SigV4 over the bytes the server received, at the time the client claimed.
-fn expected_authorization(url: &str, raw_request: &str) -> String {
-    let format =
-        format_description::parse_borrowed::<2>("[year][month][day]T[hour][minute][second]Z")
-            .unwrap();
-    let signed_at: SystemTime =
-        PrimitiveDateTime::parse(header(raw_request, "x-amz-date").unwrap(), &format)
-            .unwrap()
-            .assume_utc()
-            .into();
-    let headers: BTreeMap<String, String> = ["content-type", "x-amz-target"]
-        .into_iter()
-        .map(|name| {
-            (
-                name.to_string(),
-                header(raw_request, name).unwrap().to_string(),
-            )
-        })
-        .collect();
-    let body = raw_request.split_once("\r\n\r\n").unwrap().1;
-    sign_post(
-        url,
-        body.as_bytes(),
-        &aws_signature_headers(&headers),
-        "eu-west-1",
-        "textract",
-        &Credentials::new(ACCESS_KEY_ID, SECRET_ACCESS_KEY, None, None, "test"),
-        signed_at,
-    )
-    .unwrap()["Authorization"]
-        .clone()
-}
-
-#[tokio::test]
-async fn the_request_is_signed_for_textract_and_lines_become_the_page() {
-    let (base, seen, server) = mock_server(vec![textract_response()]).await;
-
-    let response = perform_ocr_with(LocalOcrHost::new(textract_request(&base)))
-        .await
-        .unwrap();
-    server.await.unwrap();
-
-    let raw = seen.lock().unwrap()[0].clone();
-    assert_eq!(
-        header(&raw, "x-amz-target"),
-        Some("Textract.DetectDocumentText")
-    );
-    assert_eq!(
-        header(&raw, "content-type"),
-        Some("application/x-amz-json-1.1")
-    );
-    assert_eq!(
-        request_body(&raw),
-        json!({"Document": {"Bytes": "b3JpZ2luYWw="}})
-    );
-    assert_eq!(
-        header(&raw, "authorization"),
-        Some(expected_authorization(&format!("{base}/"), &raw).as_str())
-    );
-    assert_eq!(response.pages[0].markdown, "Invoice 12345");
-    assert_eq!(response.usage_info.unwrap().pages_processed, Some(1));
-}
-
-#[tokio::test]
-async fn a_body_rewritten_by_before_send_is_what_gets_signed_and_sent() {
-    let (base, seen, server) = mock_server(vec![textract_response()]).await;
-    let host = LocalOcrHost::new(textract_request(&base)).with_before_send(|mut wire, _| {
-        assert!(
-            !wire
-                .headers
-                .iter()
-                .any(|(name, _)| name.eq_ignore_ascii_case("authorization")),
-            "the hook ran after signing"
-        );
-        wire.body["Document"]["Bytes"] = Value::from("cmVkYWN0ZWQ=");
-        Ok(wire)
-    });
-
-    perform_ocr_with(host).await.unwrap();
-    server.await.unwrap();
-
-    let raw = seen.lock().unwrap()[0].clone();
-    assert_eq!(
-        request_body(&raw),
-        json!({"Document": {"Bytes": "cmVkYWN0ZWQ="}})
-    );
-    assert_eq!(
-        header(&raw, "authorization"),
-        Some(expected_authorization(&format!("{base}/"), &raw).as_str())
-    );
-}
-
-#[tokio::test]
-async fn a_multi_page_rejection_reaches_the_caller_with_the_single_page_limit() {
-    let (base, _, server) = mock_server(vec![MockResponse {
-        status: 400,
-        headers: vec![],
-        body: json!({
-            "__type": "UnsupportedDocumentException",
-            "Message": "Request has unsupported document format"
-        }),
-    }])
-    .await;
-
-    let error = perform_ocr_with(LocalOcrHost::new(textract_request(&base)))
-        .await
-        .unwrap_err();
-    server.await.unwrap();
-
-    let Error::Provider { status, body, .. } = error else {
-        panic!("expected a provider error, got {error:?}");
-    };
-    assert_eq!(status, 400);
-    assert!(
-        body.contains("multi-page documents are not supported"),
-        "{body}"
-    );
-}
-
-#[tokio::test]
-async fn analyze_document_asks_for_layout_and_tables_and_returns_markdown() {
-    let (base, seen, server) = mock_server(vec![MockResponse::json(json!({
-        "DocumentMetadata": {"Pages": 1},
-        "Blocks": [
-            {"Id": "l1", "BlockType": "LINE", "Text": "Quarterly Report"},
-            {"Id": "t", "BlockType": "LAYOUT_TITLE",
-                "Relationships": [{"Type": "CHILD", "Ids": ["l1"]}]}
-        ]
-    }))])
-    .await;
-    let request = textract_request_for("aws_textract/analyze-document", &base);
-
-    let response = perform_ocr_with(LocalOcrHost::new(request)).await.unwrap();
-    server.await.unwrap();
-
-    let raw = seen.lock().unwrap()[0].clone();
-    assert_eq!(
-        header(&raw, "x-amz-target"),
-        Some("Textract.AnalyzeDocument")
-    );
-    assert_eq!(
-        request_body(&raw)["FeatureTypes"],
-        json!(["LAYOUT", "TABLES"])
-    );
-    assert_eq!(
-        header(&raw, "authorization"),
-        Some(expected_authorization(&format!("{base}/"), &raw).as_str())
-    );
-    assert_eq!(response.pages[0].markdown, "# Quarterly Report");
-}

litellm-rust/crates/core/tests/azure_ai_ocr.rs

@@ -1,293 +0,0 @@
-use litellm_llms::base_llm::ocr::error::Error;
-use serde_json::{Value, json};
-
-use super::test_support::{MockResponse, mock_server, perform_ocr, perform_ocr_with, wire_request};
-use crate::ocr::route::LocalOcrHost;
-
-#[tokio::test]
-async fn facade_executes_azure_mistral_with_prepared_auth() {
-    let (base, seen, server) = mock_server(vec![MockResponse::json(json!({
-        "pages":[{"index":0,"markdown":"hello"}],
-        "usage_info":{"pages_processed":1}
-    }))])
-    .await;
-    let mut request = wire_request(
-        "azure_ai/model",
-        &base,
-        json!({"include_image_base64":true}),
-    );
-    request.credentials.api_key = None;
-    request.transport.extra_headers = vec![(
-        "Authorization".into(),
-        "Bearer python-prepared-token".into(),
-    )];
-
-    let result = perform_ocr(request).await.unwrap();
-    server.await.unwrap();
-    assert_eq!(result.pages[0].markdown, "hello");
-    let requests = seen.lock().unwrap();
-    assert_eq!(requests.len(), 1);
-    assert!(requests[0].starts_with("POST /providers/mistral/azure/ocr "));
-    assert!(
-        requests[0]
-            .to_ascii_lowercase()
-            .contains("authorization: bearer python-prepared-token\r\n")
-    );
-    let body: Value = serde_json::from_str(requests[0].split_once("\r\n\r\n").unwrap().1).unwrap();
-    assert_eq!(
-        body,
-        json!({
-            "model":"model",
-            "document":{"type":"document_url","document_url":"data:application/pdf;base64,YWJj"},
-            "include_image_base64":true
-        })
-    );
-}
-
-#[tokio::test]
-async fn facade_acquires_supplied_entra_token_for_final_request() {
-    let (base, seen, server) = mock_server(vec![MockResponse::json(json!({"pages":[]}))]).await;
-    let mut request = wire_request(
-        "azure_ai/model",
-        &base,
-        json!({"azure_ad_token":"rust-owned-token"}),
-    );
-    request.credentials.api_key = None;
-
-    perform_ocr(request).await.unwrap();
-    server.await.unwrap();
-
-    let requests = seen.lock().unwrap();
-    assert_eq!(requests.len(), 1);
-    assert!(
-        requests[0]
-            .to_ascii_lowercase()
-            .contains("authorization: bearer rust-owned-token\r\n")
-    );
-}
-
-#[tokio::test]
-async fn rejects_non_inline_body_after_guardrails() {
-    let request = wire_request("azure_ai/model", "http://127.0.0.1:1", json!({}));
-    let host = LocalOcrHost::new(request).with_before_send(|mut wire, _| {
-        wire.body["document"] = json!({
-            "type":"document_url",
-            "document_url":"https://example.com/not-inline.pdf"
-        });
-        Ok(wire)
-    });
-    let error = perform_ocr_with(host).await.unwrap_err();
-    assert!(error.to_string().contains("data URI"));
-}
-
-mod transformation {
-    use std::sync::{
-        Arc,
-        atomic::{AtomicUsize, Ordering},
-    };
-
-    use litellm_auth::{
-        ResolvedCredential, SecretValue, TokenFuture, TokenProvider, TokenProviderHandle,
-    };
-    use rstest::rstest;
-    use serde_json::json;
-
-    use super::*;
-    use crate::ocr::{
-        test_support::{MockResponse, header, mock_server, perform_ocr},
-        types::LiteLLMOcrRequest,
-        wire::decode_request,
-    };
-
-    #[derive(Debug)]
-    struct CountingToken {
-        token: fn(usize) -> String,
-        calls: AtomicUsize,
-    }
-
-    impl CountingToken {
-        fn new(token: fn(usize) -> String) -> Arc<Self> {
-            Arc::new(Self {
-                token,
-                calls: AtomicUsize::new(0),
-            })
-        }
-
-        fn calls(&self) -> usize {
-            self.calls.load(Ordering::SeqCst)
-        }
-    }
-
-    impl TokenProvider for CountingToken {
-        fn acquire(&self) -> TokenFuture<'_> {
-            let call = self.calls.fetch_add(1, Ordering::SeqCst) + 1;
-            let token = SecretValue::new((self.token)(call));
-            Box::pin(async move {
-                Ok(ResolvedCredential::AccessToken {
-                    token,
-                    expires_on: None,
-                })
-            })
-        }
-    }
-
-    fn numbered_token(call: usize) -> String {
-        format!("callback-{call}")
-    }
-
-    fn azure_request(
-        provider: &Arc<CountingToken>,
-        api_base: Option<&str>,
-        api_key: Option<&str>,
-        extra_headers: Value,
-        optional_params: Value,
-    ) -> LiteLLMOcrRequest {
-        let wire = serde_json::from_value(json!({
-            "model": "azure_ai/mistral-ocr-latest",
-            "document": {"type":"document_url","document_url":"data:application/pdf;base64,YWJj"},
-            "api_key": api_key,
-            "api_base": api_base,
-            "custom_llm_provider": null,
-            "extra_headers": extra_headers,
-            "optional_params": optional_params,
-            "timeout_seconds": 2.0
-        }))
-        .unwrap();
-        LiteLLMOcrRequest {
-            azure_ad_token_provider: Some(TokenProviderHandle::new(provider.clone())),
-            ..decode_request(wire).unwrap()
-        }
-    }
-
-    fn ocr_page() -> MockResponse {
-        MockResponse::json(json!({"pages":[{"index":0,"markdown":"hello"}]}))
-    }
-
-    #[tokio::test]
-    async fn token_provider_result_is_the_bearer_and_is_acquired_for_each_request() {
-        let provider = CountingToken::new(numbered_token);
-        let (base, seen, server) = mock_server(vec![ocr_page(), ocr_page()]).await;
-
-        for _ in 0..2 {
-            perform_ocr(azure_request(
-                &provider,
-                Some(&base),
-                None,
-                Value::Null,
-                json!({}),
-            ))
-            .await
-            .unwrap();
-        }
-        server.await.unwrap();
-
-        assert_eq!(provider.calls(), 2);
-        let requests = seen.lock().unwrap();
-        assert_eq!(
-            requests
-                .iter()
-                .map(|request| header(request, "authorization"))
-                .collect::<Vec<_>>(),
-            [Some("Bearer callback-1"), Some("Bearer callback-2")]
-        );
-    }
-
-    #[rstest]
-    #[case::api_key_skips_provider(Some("resource-key"), Value::Null, json!({}), "Bearer resource-key", 0)]
-    #[case::provider_beats_static_token(
-        None,
-        Value::Null,
-        json!({"azure_ad_token":"static-token"}),
-        "Bearer callback-1",
-        1
-    )]
-    #[case::header_wins_on_the_wire_but_provider_still_runs(
-        None,
-        json!({"Authorization":"Bearer override"}),
-        json!({}),
-        "Bearer override",
-        1
-    )]
-    #[tokio::test]
-    async fn credential_precedence(
-        #[case] api_key: Option<&str>,
-        #[case] extra_headers: Value,
-        #[case] optional_params: Value,
-        #[case] expected_authorization: &str,
-        #[case] expected_calls: usize,
-    ) {
-        let provider = CountingToken::new(numbered_token);
-        let (base, seen, server) = mock_server(vec![ocr_page()]).await;
-
-        perform_ocr(azure_request(
-            &provider,
-            Some(&base),
-            api_key,
-            extra_headers,
-            optional_params,
-        ))
-        .await
-        .unwrap();
-        server.await.unwrap();
-
-        assert_eq!(provider.calls(), expected_calls);
-        let requests = seen.lock().unwrap();
-        assert_eq!(requests.len(), 1);
-        assert_eq!(
-            header(&requests[0], "authorization"),
-            Some(expected_authorization)
-        );
-    }
-
-    #[rstest]
-    #[case::missing_api_base(
-        false,
-        json!({}),
-        numbered_token,
-        |error: &Error| matches!(error, Error::Auth(litellm_auth::Error::MissingApiBase {
-            provider: "Azure AI",
-            environment_variable: "AZURE_AI_API_BASE",
-        })),
-        0
-    )]
-    #[case::unsupported_oidc_reference(
-        true,
-        json!({"azure_ad_token":"oidc/assertion","client_id":"client","tenant_id":"tenant"}),
-        numbered_token,
-        |error: &Error| matches!(error, Error::Auth(litellm_auth::Error::UnsupportedOidcReference)),
-        0
-    )]
-    #[case::empty_provider_token_ignores_static_token(
-        true,
-        json!({"azure_ad_token":"static-token"}),
-        |_| String::new(),
-        |error: &Error| matches!(error, Error::MissingAzureAiCredentials),
-        1
-    )]
-    #[tokio::test]
-    async fn credential_failures_send_no_provider_request(
-        #[case] with_api_base: bool,
-        #[case] optional_params: Value,
-        #[case] token: fn(usize) -> String,
-        #[case] expected: fn(&Error) -> bool,
-        #[case] expected_calls: usize,
-    ) {
-        let provider = CountingToken::new(token);
-        let (base, seen, server) = mock_server(vec![ocr_page()]).await;
-
-        let error = perform_ocr(azure_request(
-            &provider,
-            with_api_base.then_some(base.as_str()),
-            None,
-            Value::Null,
-            optional_params,
-        ))
-        .await
-        .unwrap_err();
-        server.abort();
-
-        assert!(expected(&error), "unexpected error: {error:?}");
-        assert_eq!(provider.calls(), expected_calls);
-        assert!(seen.lock().unwrap().is_empty());
-    }
-}

litellm-rust/crates/core/tests/azure_document_intelligence_ocr.rs

@@ -1,712 +0,0 @@
-use litellm_host::event::{CallEvent, MachineEvent};
-use litellm_llms::base_llm::ocr::{error::Error, settings::OcrSettings};
-use rstest::rstest;
-use serde_json::{Value, json};
-
-use super::{
-    test_support::{
-        MockResponse, mock_server, ocr_client, perform_ocr, perform_ocr_with, wire_request,
-    },
-    wire::{OcrWireRequest, decode_request},
-};
-use crate::ocr::route::LocalOcrHost;
-
-fn query_value(url: &str, key: &str) -> Option<String> {
-    url::Url::parse(url)
-        .unwrap()
-        .query_pairs()
-        .find_map(|(name, value)| (name == key).then(|| value.into_owned()))
-}
-
-#[tokio::test]
-async fn facade_maps_pages_features_and_url_document() {
-    let (base, seen, server) = mock_server(vec![MockResponse::json(json!({
-        "status":"succeeded",
-        "analyzeResult":{"pages":[]}
-    }))])
-    .await;
-    let mut request = wire_request(
-        "azure_ai/doc-intelligence/prebuilt-read",
-        &base,
-        json!({"pages":[2,0,0,1],"features":["keyValuePairs","languages"]}),
-    );
-    request.document =
-        serde_json::from_value::<litellm_llms::base_llm::ocr::transformation::OcrDocument>(json!({
-            "type":"document_url",
-            "document_url":"https://example.com/document.pdf"
-        }))
-        .unwrap()
-        .into();
-
-    perform_ocr(request).await.unwrap();
-    server.await.unwrap();
-    let request = &seen.lock().unwrap()[0];
-    let target = request.split_whitespace().nth(1).unwrap();
-    let url = format!("{base}{target}");
-    assert_eq!(query_value(&url, "pages").as_deref(), Some("1,2,3"));
-    assert_eq!(
-        query_value(&url, "features").as_deref(),
-        Some("keyValuePairs,languages")
-    );
-    let body: Value = serde_json::from_str(request.split_once("\r\n\r\n").unwrap().1).unwrap();
-    assert_eq!(
-        body,
-        json!({"urlSource":"https://example.com/document.pdf"})
-    );
-}
-
-#[rstest]
-#[case(json!({"pages":[true]}), Error::Pages("expected only integers or only strings".into()))]
-#[case(json!({"pages":[1,"2"]}), Error::Pages("expected only integers or only strings".into()))]
-#[case(json!({"pages":[-1]}), Error::Pages("negative page index".into()))]
-#[case(json!({"pages":"1&&features=bad"}), Error::Pages("invalid native page range".into()))]
-#[case(json!({"features":"languages&pages=1"}), Error::Features)]
-#[case(json!({"req_format":"azure"}), Error::RequestFormat)]
-#[tokio::test]
-async fn rejects_invalid_pages_features_and_format(
-    #[case] options: Value,
-    #[case] expected: Error,
-) {
-    let (base, seen, server) = mock_server(vec![MockResponse::json(json!({}))]).await;
-    let result = decode_request(OcrWireRequest {
-        model: "azure_ai/doc-intelligence/prebuilt-read".into(),
-        document: json!({"type":"document_url","document_url":"https://example.com/a.pdf"}),
-        api_key: Some(litellm_auth::SecretValue::new("key")),
-        api_base: Some(base),
-        custom_llm_provider: None,
-        extra_headers: None,
-        optional_params: options.as_object().unwrap().clone(),
-        input_sources: Default::default(),
-        timeout_seconds: Some(2.0),
-    });
-    let result = match result {
-        Ok(request) => perform_ocr(request).await,
-        Err(error) => Err(error),
-    };
-    server.abort();
-    let _ = server.await;
-    assert!(
-        seen.lock().unwrap().is_empty(),
-        "sent invalid options: {options}"
-    );
-    let error = result.unwrap_err();
-    assert_eq!(
-        std::mem::discriminant(&error),
-        std::mem::discriminant(&expected)
-    );
-    assert_eq!(error.http_status_code(), Some(400));
-    assert_eq!(error.to_string(), expected.to_string());
-}
-
-#[rstest]
-#[case(json!({}))]
-#[case(json!({"req_format":"litellm"}))]
-#[tokio::test]
-async fn missing_native_fields_keep_page_text_without_retaining_raw_response(
-    #[case] options: Value,
-) {
-    let operation = json!({
-        "status":"succeeded",
-        "analyzeResult":{"pages":[{"pageNumber":1,"lines":[{"content":"hello"}]}]}
-    });
-    let (base, seen, server) = mock_server(vec![MockResponse::json(operation)]).await;
-    let response = perform_ocr(wire_request(
-        "azure_ai/doc-intelligence/prebuilt-read",
-        &base,
-        options,
-    ))
-    .await
-    .unwrap();
-    server.await.unwrap();
-
-    assert_eq!(response.pages.len(), 1);
-    assert_eq!(response.pages[0].index, 0);
-    assert_eq!(response.pages[0].markdown, "hello");
-    assert_eq!(response.provider_native_response, None);
-    let serialized = response.into_json();
-    assert_eq!(serialized.get("content"), Some(&Value::Null));
-    assert_eq!(serialized.get("tables"), Some(&Value::Null));
-    assert_eq!(serialized.get("keyValuePairs"), Some(&Value::Null));
-    let requests = seen.lock().unwrap();
-    assert_eq!(requests.len(), 1);
-    let target = requests[0].split_whitespace().nth(1).unwrap();
-    let url = format!("{base}{target}");
-    for field in ["pages", "features", "req_format"] {
-        assert_eq!(query_value(&url, field), None);
-    }
-    let body: Value = serde_json::from_str(requests[0].split_once("\r\n\r\n").unwrap().1).unwrap();
-    assert_eq!(body, json!({"base64Source":"YWJj"}));
-}
-
-#[tokio::test]
-async fn inline_document_decodes_to_base64_source() {
-    let (base, seen, server) = mock_server(vec![MockResponse::json(json!({
-        "status":"succeeded"
-    }))])
-    .await;
-    let request = wire_request("azure_ai/doc-intelligence/prebuilt-read", &base, json!({}));
-
-    perform_ocr(request).await.unwrap();
-    server.await.unwrap();
-    let request = &seen.lock().unwrap()[0];
-    let body: Value = serde_json::from_str(request.split_once("\r\n\r\n").unwrap().1).unwrap();
-    assert_eq!(body, json!({"base64Source":"YWJj"}));
-}
-
-#[tokio::test]
-async fn immediate_response_normalizes_pages_and_preserves_native() {
-    let operation = json!({
-        "status":"succeeded",
-        "operationExtension":42,
-        "analyzeResult":{
-            "content":"A\n\nB",
-            "tables":[{"cells":[]}],
-            "keyValuePairs":[{"key":{"content":"A"}}],
-            "pages":[{
-                "pageNumber":"2",
-                "width":"8.5",
-                "height":11,
-                "unit":"inch",
-                "lines":[{"content":"A"},{"content":null},{"content":"B"}]
-            }]
-        }
-    });
-    let (base, _, server) = mock_server(vec![MockResponse::json(operation.clone())]).await;
-    let result = perform_ocr(wire_request(
-        "azure_ai/doc-intelligence/prebuilt-read",
-        &base,
-        json!({"req_format":"native"}),
-    ))
-    .await
-    .unwrap();
-    server.await.unwrap();
-
-    assert_eq!(result.pages[0].index, 1);
-    assert_eq!(result.pages[0].markdown, "A\n\nB");
-    assert_eq!(
-        serde_json::to_value(&result.pages[0].dimensions).unwrap(),
-        json!({"width":816,"height":1056,"dpi":96})
-    );
-    assert_eq!(result.usage_info.as_ref().unwrap().pages_processed, Some(1));
-    let serialized = result.clone().into_json();
-    assert_eq!(serialized["content"], "A\n\nB");
-    assert_eq!(serialized["tables"], json!([{"cells":[]}]));
-    assert_eq!(
-        serialized["keyValuePairs"],
-        json!([{"key":{"content":"A"}}])
-    );
-    assert!(serialized.get("key_value_pairs").is_none());
-    assert_eq!(
-        result.provider_native_response.map(Value::Object),
-        Some(operation)
-    );
-}
-
-#[tokio::test]
-async fn client_settings_choose_the_api_version_and_the_inch_to_pixel_dpi() {
-    let (base, seen, server) = mock_server(vec![MockResponse::json(json!({
-        "status":"succeeded",
-        "analyzeResult":{"pages":[{"pageNumber":1,"width":8.5,"height":11,"unit":"inch"}]}
-    }))])
-    .await;
-    let client = ocr_client().with_settings(OcrSettings {
-        document_intelligence_api_version: "2099-01-01".into(),
-        document_intelligence_dpi: 72,
-        ..OcrSettings::default()
-    });
-
-    let result = crate::ocr::client::perform(
-        &client,
-        wire_request("azure_ai/doc-intelligence/prebuilt-read", &base, json!({})),
-    )
-    .await
-    .unwrap();
-    server.await.unwrap();
-
-    let target = seen.lock().unwrap()[0]
-        .split_whitespace()
-        .nth(1)
-        .unwrap()
-        .to_string();
-    assert_eq!(
-        query_value(&format!("{base}{target}"), "api-version").as_deref(),
-        Some("2099-01-01")
-    );
-    assert_eq!(
-        serde_json::to_value(&result.pages[0].dimensions).unwrap(),
-        json!({"width":612,"height":792,"dpi":72})
-    );
-}
-
-#[tokio::test]
-async fn accepted_response_polls_to_success_with_only_credentials() {
-    let operation = json!({"status":"succeeded","analyzeResult":{"pages":[]}});
-    let (base, seen, server) = mock_server(vec![
-        MockResponse {
-            status: 202,
-            headers: vec![("Operation-Location", "{base}/operation".into())],
-            body: json!({}),
-        },
-        MockResponse {
-            status: 200,
-            headers: vec![("Retry-After", "0".into())],
-            body: json!({"status":"running"}),
-        },
-        MockResponse::json(operation.clone()),
-    ])
-    .await;
-    let mut request = wire_request(
-        "azure_ai/doc-intelligence/prebuilt-read",
-        &base,
-        json!({"req_format":"native"}),
-    );
-    request
-        .transport
-        .extra_headers
-        .push(("X-Trace".into(), "initial-only".into()));
-
-    let result = perform_ocr(request).await.unwrap();
-    server.await.unwrap();
-    assert_eq!(
-        result.provider_native_response.map(Value::Object),
-        Some(operation)
-    );
-    let requests = seen.lock().unwrap();
-    assert_eq!(requests.len(), 3);
-    assert!(requests[0].to_ascii_lowercase().contains("x-trace:"));
-    for poll in &requests[1..] {
-        assert!(!poll.to_ascii_lowercase().contains("x-trace:"));
-        assert!(
-            poll.to_ascii_lowercase()
-                .contains("ocp-apim-subscription-key: test-key")
-        );
-    }
-}
-
-#[tokio::test]
-async fn accepted_response_emits_response_received_before_polling() {
-    let (base, seen, server) = mock_server(vec![
-        MockResponse {
-            status: 202,
-            headers: vec![("Operation-Location", "{base}/operation".into())],
-            body: json!({"submitted": true}),
-        },
-        MockResponse::json(json!({"status":"succeeded"})),
-    ])
-    .await;
-    let request_count = seen.clone();
-    let host = LocalOcrHost::new(wire_request(
-        "azure_ai/doc-intelligence/prebuilt-read",
-        &base,
-        json!({}),
-    ))
-    .with_observer(move |event| {
-        let CallEvent::Machine(MachineEvent::ResponseReceived { raw }) = event else {
-            return;
-        };
-        match request_count.lock().unwrap().len() {
-            1 => assert_eq!(raw.body, r#"{"submitted":true}"#),
-            2 => assert!(raw.body.contains("succeeded")),
-            count => panic!("unexpected callback after {count} requests"),
-        }
-    });
-
-    perform_ocr_with(host).await.unwrap();
-    server.await.unwrap();
-    assert_eq!(seen.lock().unwrap().len(), 2);
-}
-
-#[tokio::test]
-async fn polling_forwards_bearer_credentials() {
-    let (base, seen, server) = mock_server(vec![
-        MockResponse {
-            status: 202,
-            headers: vec![("Operation-Location", "{base}/operation".into())],
-            body: json!({}),
-        },
-        MockResponse::json(json!({"status":"succeeded"})),
-    ])
-    .await;
-    let mut request = wire_request("azure_ai/doc-intelligence/prebuilt-read", &base, json!({}));
-    request.credentials.api_key = None;
-    request.transport.extra_headers = vec![("Authorization".into(), "Bearer token".into())];
-
-    perform_ocr(request).await.unwrap();
-    server.await.unwrap();
-    let requests = seen.lock().unwrap();
-    assert!(
-        requests[1]
-            .to_ascii_lowercase()
-            .contains("authorization: bearer token")
-    );
-}
-
-#[tokio::test]
-async fn polling_does_not_follow_redirects() {
-    let (base, seen, server) = mock_server(vec![
-        MockResponse {
-            status: 202,
-            headers: vec![("Operation-Location", "{base}/operation".into())],
-            body: json!({}),
-        },
-        MockResponse {
-            status: 302,
-            headers: vec![("Location", "{base}/redirected".into())],
-            body: json!({}),
-        },
-        MockResponse::json(json!({"status":"succeeded"})),
-    ])
-    .await;
-
-    let error = perform_ocr(wire_request(
-        "azure_ai/doc-intelligence/prebuilt-read",
-        &base,
-        json!({}),
-    ))
-    .await
-    .unwrap_err();
-
-    assert!(error.to_string().contains("status 302"), "{error}");
-    assert_eq!(seen.lock().unwrap().len(), 2);
-    server.abort();
-}
-
-#[tokio::test]
-async fn polling_rejects_terminal_failure() {
-    let (base, _, server) = mock_server(vec![
-        MockResponse {
-            status: 202,
-            headers: vec![("Operation-Location", "{base}/operation".into())],
-            body: json!({}),
-        },
-        MockResponse::json(json!({"status":"failed"})),
-    ])
-    .await;
-
-    let error = perform_ocr(wire_request(
-        "azure_ai/doc-intelligence/prebuilt-read",
-        &base,
-        json!({}),
-    ))
-    .await
-    .unwrap_err();
-    server.await.unwrap();
-    assert!(error.to_string().contains("status failed"));
-}
-
-#[tokio::test]
-async fn malformed_provider_pages_report_response_paths() {
-    for (analysis, path) in [
-        (json!({"pages":null}), "pages"),
-        (json!({"pages":[null]}), "pages[0]"),
-        (json!({"pages":[{"lines":null}]}), "lines"),
-        (json!({"pages":[{"width":"bad"}]}), "width"),
-    ] {
-        let (base, _, server) = mock_server(vec![MockResponse::json(json!({
-            "status":"succeeded",
-            "analyzeResult":analysis
-        }))])
-        .await;
-        let error = perform_ocr(wire_request(
-            "azure_ai/doc-intelligence/prebuilt-read",
-            &base,
-            json!({}),
-        ))
-        .await
-        .unwrap_err();
-        server.await.unwrap();
-        assert!(error.to_string().contains(path), "{error}");
-    }
-}
-
-#[tokio::test]
-async fn rejects_missing_invalid_and_cross_origin_operation_locations() {
-    for headers in [
-        Vec::new(),
-        vec![("Operation-Location", "/relative".into())],
-        vec![("Operation-Location", "http://example.com/operation".into())],
-        vec![(
-            "Operation-Location",
-            "http://user:password@127.0.0.1/operation".into(),
-        )],
-    ] {
-        let (base, _, server) = mock_server(vec![MockResponse {
-            status: 202,
-            headers,
-            body: json!({}),
-        }])
-        .await;
-        let error = perform_ocr(wire_request(
-            "azure_ai/doc-intelligence/prebuilt-read",
-            &base,
-            json!({}),
-        ))
-        .await
-        .unwrap_err();
-        server.await.unwrap();
-        assert!(error.to_string().contains("operation-location"));
-    }
-}
-
-#[tokio::test]
-async fn polling_deadline_bounds_retry_delay() {
-    let (base, _, server) = mock_server(vec![
-        MockResponse {
-            status: 202,
-            headers: vec![("Operation-Location", "{base}/operation".into())],
-            body: json!({}),
-        },
-        MockResponse {
-            status: 200,
-            headers: vec![("Retry-After", "9999".into())],
-            body: json!({"status":"notStarted"}),
-        },
-    ])
-    .await;
-    let request = wire_request("azure_ai/doc-intelligence/prebuilt-read", &base, json!({}));
-    let client = ocr_client().with_settings(OcrSettings {
-        poll_timeout: std::time::Duration::from_millis(100),
-        ..OcrSettings::default()
-    });
-
-    let error = tokio::time::timeout(
-        std::time::Duration::from_secs(1),
-        crate::ocr::client::perform(&client, request),
-    )
-    .await
-    .unwrap()
-    .unwrap_err();
-    server.await.unwrap();
-    assert!(error.to_string().contains("timed out"));
-}
-
-#[tokio::test]
-async fn model_id_is_encoded_and_dot_segments_are_rejected() {
-    let (base, seen, server) = mock_server(vec![MockResponse::json(json!({
-        "status":"succeeded"
-    }))])
-    .await;
-    perform_ocr(wire_request(
-        "azure_ai/doc-intelligence/a ?#é",
-        &base,
-        json!({}),
-    ))
-    .await
-    .unwrap();
-    server.await.unwrap();
-    assert!(seen.lock().unwrap()[0].contains("a%20%3F%23%C3%A9:analyze"));
-
-    for model in [
-        "azure_ai/doc-intelligence/.",
-        "azure_ai/doc-intelligence/..",
-    ] {
-        let error = perform_ocr(wire_request(model, "http://127.0.0.1:1", json!({})))
-            .await
-            .unwrap_err();
-        assert!(error.to_string().contains("dot segment"));
-    }
-}
-
-mod transformation {
-    use std::sync::{Arc, Mutex};
-
-    use litellm_host::event::{CallEvent, MachineEvent};
-    use litellm_llms::base_llm::ocr::transformation::OcrDocument;
-    use serde_json::{Value, json};
-
-    use super::*;
-    use crate::ocr::{
-        route::LocalOcrHost,
-        test_support::{MockResponse, mock_server, perform_ocr, perform_ocr_with, wire_request},
-    };
-
-    #[tokio::test]
-    async fn facade_maps_pages_features_and_url_document() {
-        let (base, seen, server) = mock_server(vec![MockResponse::json(json!({
-            "status":"succeeded",
-            "analyzeResult":{"pages":[]}
-        }))])
-        .await;
-        let mut request = wire_request(
-            "azure_ai/doc-intelligence/prebuilt-read",
-            &base,
-            json!({"pages":[2,0,0,1],"features":["keyValuePairs","languages"], "future_option": {"nested":null}, "extra_body":{"provider_option":false}}),
-        );
-        request.document = serde_json::from_value::<OcrDocument>(json!({
-            "type":"document_url",
-            "document_url":"https://example.com/document.pdf"
-        }))
-        .unwrap()
-        .into();
-
-        perform_ocr(request).await.unwrap();
-        server.await.unwrap();
-        let request = &seen.lock().unwrap()[0];
-        let target = request.split_whitespace().nth(1).unwrap();
-        let url = format!("{base}{target}");
-        assert_eq!(query_value(&url, "pages").as_deref(), Some("1,2,3"));
-        assert_eq!(
-            query_value(&url, "features").as_deref(),
-            Some("keyValuePairs,languages")
-        );
-        let body: Value = serde_json::from_str(request.split_once("\r\n\r\n").unwrap().1).unwrap();
-        assert_eq!(
-            body,
-            json!({"urlSource":"https://example.com/document.pdf", "future_option":{"nested":null}, "provider_option":false})
-        );
-    }
-
-    #[tokio::test]
-    async fn rejects_invalid_pages_features_and_format() {
-        for options in [
-            json!({"pages":[true]}),
-            json!({"pages":[1,"2"]}),
-            json!({"pages":[-1]}),
-            json!({"pages":"1&&features=bad"}),
-            json!({"features":"languages&pages=1"}),
-            json!({"req_format":"azure"}),
-        ] {
-            let request = wire_request(
-                "azure_ai/doc-intelligence/prebuilt-read",
-                "http://127.0.0.1:1",
-                options.clone(),
-            );
-            let rejected = perform_ocr(request).await.is_err();
-            assert!(rejected, "accepted {options}");
-        }
-    }
-
-    #[tokio::test]
-    async fn immediate_response_normalizes_pages_and_preserves_native() {
-        let operation = json!({
-            "status":"succeeded",
-            "operationExtension":42,
-            "analyzeResult":{
-                "content":"A\n\nB",
-                "tables":[{"cells":[]}],
-                "keyValuePairs":[{"key":{"content":"A"}}],
-                "pages":[{
-                    "pageNumber":"2",
-                    "width":"8.5",
-                    "height":11,
-                    "unit":"inch",
-                    "lines":[{"content":"A"},{"content":null},{"content":"B"}]
-                }]
-            }
-        });
-        let (base, _, server) = mock_server(vec![MockResponse::json(operation.clone())]).await;
-        let result = perform_ocr(wire_request(
-            "azure_ai/doc-intelligence/prebuilt-read",
-            &base,
-            json!({"req_format":"native"}),
-        ))
-        .await
-        .unwrap();
-        server.await.unwrap();
-
-        assert_eq!(result.pages[0].index, 1);
-        assert_eq!(result.pages[0].markdown, "A\n\nB");
-        assert_eq!(
-            serde_json::to_value(&result.pages[0].dimensions).unwrap(),
-            json!({"width":816,"height":1056,"dpi":96})
-        );
-        assert_eq!(result.usage_info.as_ref().unwrap().pages_processed, Some(1));
-        let serialized = result.clone().into_json();
-        assert_eq!(serialized["content"], "A\n\nB");
-        assert_eq!(serialized["tables"], json!([{"cells":[]}]));
-        assert_eq!(
-            serialized["keyValuePairs"],
-            json!([{"key":{"content":"A"}}])
-        );
-        assert!(serialized.get("key_value_pairs").is_none());
-        assert_eq!(
-            result.provider_native_response.as_ref(),
-            operation.as_object()
-        );
-    }
-
-    #[tokio::test]
-    async fn accepted_response_polls_to_success_with_only_credentials() {
-        let operation = json!({"status":"succeeded","analyzeResult":{"pages":[]}});
-        let (base, seen, server) = mock_server(vec![
-            MockResponse {
-                status: 202,
-                headers: vec![("Operation-Location", "{base}/operation".into())],
-                body: json!({}),
-            },
-            MockResponse {
-                status: 200,
-                headers: vec![("Retry-After", "0".into())],
-                body: json!({"status":"running"}),
-            },
-            MockResponse::json(operation.clone()),
-        ])
-        .await;
-        let mut request = wire_request(
-            "azure_ai/doc-intelligence/prebuilt-read",
-            &base,
-            json!({"req_format":"native"}),
-        );
-        request
-            .transport
-            .extra_headers
-            .push(("X-Trace".into(), "initial-only".into()));
-
-        let result = perform_ocr(request).await.unwrap();
-        server.await.unwrap();
-        assert_eq!(
-            result.provider_native_response.as_ref(),
-            operation.as_object()
-        );
-        let requests = seen.lock().unwrap();
-        assert_eq!(requests.len(), 3);
-        assert!(requests[0].to_ascii_lowercase().contains("x-trace:"));
-        for poll in &requests[1..] {
-            assert!(!poll.to_ascii_lowercase().contains("x-trace:"));
-            assert!(
-                poll.to_ascii_lowercase()
-                    .contains("ocp-apim-subscription-key: test-key")
-            );
-        }
-    }
-
-    #[tokio::test]
-    async fn accepted_response_emits_response_received_for_submission_and_completed_poll() {
-        let (base, seen, server) = mock_server(vec![
-            MockResponse {
-                status: 202,
-                headers: vec![("Operation-Location", "{base}/operation".into())],
-                body: json!({"submitted": true}),
-            },
-            MockResponse::json(json!({"status":"succeeded"})),
-        ])
-        .await;
-        let responses_received = Arc::new(Mutex::new(Vec::new()));
-        let request_count = seen.clone();
-        let observed = responses_received.clone();
-        let host = LocalOcrHost::new(wire_request(
-            "azure_ai/doc-intelligence/prebuilt-read",
-            &base,
-            json!({}),
-        ))
-        .with_observer(move |event| {
-            if let CallEvent::Machine(MachineEvent::ResponseReceived { raw }) = event {
-                observed
-                    .lock()
-                    .unwrap()
-                    .push((request_count.lock().unwrap().len(), raw.body.clone()));
-            }
-        });
-
-        perform_ocr_with(host).await.unwrap();
-        server.await.unwrap();
-        assert_eq!(seen.lock().unwrap().len(), 2);
-        assert_eq!(
-            *responses_received.lock().unwrap(),
-            [
-                (1, r#"{"submitted":true}"#.to_string()),
-                (2, r#"{"status":"succeeded"}"#.to_string()),
-            ]
-        );
-    }
-}

litellm-rust/crates/core/tests/chat_completions.rs

@@ -0,0 +1,320 @@
+use std::time::Duration;
+
+use litellm_core::chat_completions::{
+    Error, chat_completions, chat_completions_decline_reason, types::ChatCompletionsRequest,
+};
+use litellm_http::transport::Error as TransportError;
+use rstest::{fixture, rstest};
+use serde_json::{Map, Value, json};
+use wiremock::ResponseTemplate;
+
+mod support;
+use support::*;
+
+const ANTHROPIC_MESSAGE: &str = r#"{"id":"msg_1","type":"message","role":"assistant","model":"claude-sonnet-4-5-20260101","content":[{"type":"text","text":"hello"}],"stop_reason":"end_turn","stop_sequence":null,"usage":{"input_tokens":11,"output_tokens":4}}"#;
+
+fn object(value: Value) -> Map<String, Value> {
+    let Value::Object(map) = value else {
+        panic!("expected a json object, got {value}");
+    };
+    map
+}
+
+fn anthropic_response(body: &str) -> ResponseTemplate {
+    ResponseTemplate::new(200).set_body_raw(body, "application/json")
+}
+
+fn hi() -> Value {
+    json!([{"role": "user", "content": "hi"}])
+}
+
+#[fixture]
+fn request() -> ChatCompletionsRequest<'static> {
+    ChatCompletionsRequest {
+        model: "anthropic/claude-sonnet-4-5",
+        messages: hi(),
+        optional_params: object(json!({"max_tokens": 16})),
+        api_key: Some("sk-test"),
+        api_base: None,
+        custom_llm_provider: None,
+        extra_headers: None,
+        timeout: Some(Duration::from_secs(10)),
+    }
+}
+
+#[rstest]
+#[tokio::test]
+async fn anthropic_round_trip_translates_the_conversation_and_normalizes_the_response(
+    request: ChatCompletionsRequest<'static>,
+) {
+    let upstream = upstream([anthropic_response(ANTHROPIC_MESSAGE)]).await;
+    let base = upstream.uri();
+
+    let response = chat_completions(ChatCompletionsRequest {
+        messages: json!([
+            {"role": "system", "content": "be terse"},
+            {"role": "user", "content": "hi"}
+        ]),
+        api_base: Some(&base),
+        ..request
+    })
+    .await
+    .expect("call succeeds");
+
+    let sent = only_request(&upstream).await;
+    assert_eq!(sent.url.path(), "/v1/messages");
+    assert_eq!(sent.header_values("x-api-key"), ["sk-test"]);
+    let body = sent.json();
+    assert_eq!(body["model"], "claude-sonnet-4-5");
+    assert_eq!(
+        body["messages"],
+        json!([{"role": "user", "content": [{"type": "text", "text": "hi"}]}])
+    );
+    assert_eq!(
+        body["system"],
+        json!([{"type": "text", "text": "be terse"}])
+    );
+    assert_eq!(body["max_tokens"], 16);
+    assert_eq!(
+        response.choices[0].message.content.as_deref(),
+        Some("hello")
+    );
+    assert_eq!(response.usage.total_tokens, 15);
+}
+
+#[rstest]
+#[tokio::test]
+async fn the_deployment_key_replaces_a_caller_supplied_x_api_key(
+    request: ChatCompletionsRequest<'static>,
+) {
+    let upstream = upstream([anthropic_response(ANTHROPIC_MESSAGE)]).await;
+    let base = upstream.uri();
+
+    chat_completions(ChatCompletionsRequest {
+        api_base: Some(&base),
+        extra_headers: Some(object(
+            json!({"x-api-key": "caller-key", "x-trace": "kept"}),
+        )),
+        ..request
+    })
+    .await
+    .expect("call succeeds");
+
+    let sent = only_request(&upstream).await;
+    assert_eq!(sent.header_values("x-api-key"), ["sk-test"]);
+    assert_eq!(sent.header("x-trace"), Some("kept"));
+}
+
+#[rstest]
+#[tokio::test]
+async fn bedrock_round_trip_is_signed_and_normalized(request: ChatCompletionsRequest<'static>) {
+    let upstream = upstream([json_response(json!({
+        "output": {"message": {"role": "assistant", "content": [{"text": "hello"}]}},
+        "stopReason": "end_turn",
+        "usage": {"inputTokens": 11, "outputTokens": 4, "totalTokens": 15}
+    }))])
+    .await;
+    let base = upstream.uri();
+
+    let response = chat_completions(ChatCompletionsRequest {
+        model: "bedrock/anthropic.claude-sonnet-4-5",
+        optional_params: object(json!({
+            "aws_access_key_id": "access-key",
+            "aws_secret_access_key": "secret-key",
+            "aws_region_name": "eu-west-1"
+        })),
+        api_key: None,
+        api_base: Some(&base),
+        ..request
+    })
+    .await
+    .expect("call succeeds");
+
+    let sent = only_request(&upstream).await;
+    assert_eq!(
+        sent.url.path(),
+        "/model/anthropic.claude-sonnet-4-5/converse"
+    );
+    let authorization = sent.header("authorization").expect("request is signed");
+    assert!(
+        authorization.contains("/eu-west-1/bedrock/aws4_request"),
+        "{authorization}"
+    );
+    assert_eq!(
+        sent.json()["messages"],
+        json!([{"role": "user", "content": [{"text": "hi"}]}])
+    );
+    assert_eq!(
+        response.choices[0].message.content.as_deref(),
+        Some("hello")
+    );
+    assert_eq!(response.usage.total_tokens, 15);
+}
+
+/// The provider already answered and billed these, so the host must not retry them on
+/// its own path: they surface as `InvalidResponse`, never as a pre-send decline.
+#[rstest]
+#[case::missing_usage(
+    r#"{"model":"m","content":[{"type":"text","text":"hi"}],"stop_reason":"end_turn"}"#
+)]
+#[case::tool_use_block(r#"{"model":"m","content":[{"type":"tool_use","id":"t","name":"f","input":{}}],"stop_reason":"tool_use","usage":{"input_tokens":1,"output_tokens":1}}"#)]
+#[case::not_json("not json")]
+#[tokio::test]
+async fn a_response_it_cannot_normalize_is_reported_as_already_sent(
+    request: ChatCompletionsRequest<'static>,
+    #[case] body: &str,
+) {
+    let upstream = upstream([anthropic_response(body)]).await;
+    let base = upstream.uri();
+
+    let error = chat_completions(ChatCompletionsRequest {
+        api_base: Some(&base),
+        ..request
+    })
+    .await
+    .expect_err("response cannot be normalized");
+
+    assert!(matches!(error, Error::InvalidResponse(_)), "{error:?}");
+}
+
+#[rstest]
+#[case::rate_limited(429)]
+#[case::server_error(500)]
+#[tokio::test]
+async fn an_upstream_error_status_keeps_its_code_and_body(
+    request: ChatCompletionsRequest<'static>,
+    #[case] status: u16,
+) {
+    let upstream = upstream([ResponseTemplate::new(status).set_body_string("slow down")]).await;
+    let base = upstream.uri();
+
+    let error = chat_completions(ChatCompletionsRequest {
+        api_base: Some(&base),
+        ..request
+    })
+    .await
+    .expect_err("upstream rejects");
+
+    assert_eq!(
+        error,
+        Error::Transport(TransportError::Http {
+            status,
+            body: "slow down".into()
+        })
+    );
+}
+
+/// Nothing was sent, so nothing was billed and the host can still serve the request.
+#[rstest]
+#[tokio::test]
+async fn a_connection_that_is_never_established_declines_instead_of_failing(
+    request: ChatCompletionsRequest<'static>,
+) {
+    let error = chat_completions(ChatCompletionsRequest {
+        api_base: Some(UNREACHABLE_BASE),
+        ..request
+    })
+    .await
+    .expect_err("nothing is listening");
+
+    assert!(
+        matches!(error, Error::Transport(TransportError::Connect(_))),
+        "{error:?}"
+    );
+}
+
+#[rstest]
+#[tokio::test]
+async fn a_timeout_after_sending_is_not_a_pre_send_decline(
+    request: ChatCompletionsRequest<'static>,
+) {
+    let upstream =
+        upstream([anthropic_response(ANTHROPIC_MESSAGE).set_delay(Duration::from_secs(5))]).await;
+    let base = upstream.uri();
+
+    let error = chat_completions(ChatCompletionsRequest {
+        api_base: Some(&base),
+        timeout: Some(Duration::from_millis(100)),
+        ..request
+    })
+    .await
+    .expect_err("the call times out");
+
+    assert!(
+        matches!(error, Error::Transport(TransportError::Network(_))),
+        "{error:?}"
+    );
+}
+
+#[rstest]
+#[case::accepted("anthropic/claude-sonnet-4-5", None, hi(), json!({"max_tokens": 16}), None)]
+#[case::accepted_bedrock("bedrock/anthropic.claude-sonnet-4-5", None, hi(), json!({}), None)]
+#[case::unknown_provider(
+    "gpt-4o",
+    Some("openai"),
+    hi(),
+    json!({}),
+    Some("provider is not on the rust chat completions path")
+)]
+#[case::unreadable_messages(
+    "anthropic/claude-sonnet-4-5",
+    None,
+    json!("hi"),
+    json!({}),
+    Some("unreadable message list")
+)]
+#[case::empty_messages("anthropic/claude-sonnet-4-5", None, json!([]), json!({}), Some("empty message list"))]
+#[case::streaming(
+    "anthropic/claude-sonnet-4-5",
+    None,
+    hi(),
+    json!({"stream": true}),
+    Some("streaming")
+)]
+#[case::unrecognized_param(
+    "anthropic/claude-sonnet-4-5",
+    None,
+    hi(),
+    json!({"not_a_param": 1}),
+    Some("unrecognized request parameter")
+)]
+#[case::opens_on_assistant_turn(
+    "anthropic/claude-sonnet-4-5",
+    None,
+    json!([{"role": "assistant", "content": "hi"}]),
+    json!({}),
+    Some("conversation does not open on a user turn")
+)]
+fn decline_reason_names_why_the_core_would_not_serve_the_request(
+    #[case] model: &str,
+    #[case] provider: Option<&str>,
+    #[case] messages: Value,
+    #[case] params: Value,
+    #[case] reason: Option<&str>,
+) {
+    assert_eq!(
+        chat_completions_decline_reason(model, provider, messages, &object(params)),
+        reason
+    );
+}
+
+/// A request the decline check accepts must not be declined by the call itself.
+#[rstest]
+#[tokio::test]
+async fn a_declined_request_fails_the_call_before_sending(
+    request: ChatCompletionsRequest<'static>,
+) {
+    let upstream = upstream([anthropic_response(ANTHROPIC_MESSAGE)]).await;
+    let base = upstream.uri();
+
+    let error = chat_completions(ChatCompletionsRequest {
+        optional_params: object(json!({"stream": true})),
+        api_base: Some(&base),
+        ..request
+    })
+    .await
+    .expect_err("streaming is declined");
+
+    assert_eq!(error, Error::Unsupported("streaming"));
+    assert!(received(&upstream).await.is_empty());
+}