scriptoriumadmin/litellm
1
0
Fork 0
mirror of https://github.com/BerriAI/litellm.git synced 2026-08-28 05:25:59 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 36
litellm/tests/test_litellm/proxy
History
Exact
tin-berri 2306816d40
fix(shadow_eval): refuse a judge model that also serves one of the arms it grades (#38589) 
A shadow eval whose judge_model is one of the router's tier models, the router's
default model, or a reverse job's baseline_model was accepted with no warning. An
LLM judge scores its own output higher than a rival's, so that tier's win rate
measures the judge instead of the models, and the job's whole budget buys a result
that has to be thrown away.

start_shadow_eval now rejects it with a 400 naming the colliding arm.

`judge_target` is the single answer to "where does a call to this name go for this
caller, and what answers it", and the resolvability gate, the collision gate and
the judge dispatch all read it. It has three outcomes and no others: the router
serves the name, the SDK serves it, or nothing does. Splitting that question is
what every bug here came from, so `router_resolves_model` and `answering_models`
are gone rather than joined by a third.

Two spellings of one model are one identity. A name is compared by what would
answer it, resolved through every channel `get_model_list` composes and then put
in the provider-qualified form litellm itself uses, so a judge given as `gpt-4o`
collides with a tier deployment serving `openai/gpt-4o`, and a judge given as
`openai/gpt-4o` collides with a deployment configured as bare `gpt-4o`. Both ends
are normalised because an admin writes them at different times.

Answering is also per-caller. The shadow and judge calls carry the shadowed key's
`user_api_key_team_id`, which is what the router selects deployments with, so the
endpoint derives the job's teams once from the keys it already looks up and every
check runs under them, and the judge dispatch picks its arm under the same team.
A team's public model name resolves to nothing for everyone else and a team's own
deployment resolves for nobody else, so a check that omits the team answers for a
caller who does not exist. A collision under any one team fails the job, because
every key's verdicts land in the same win rates.

Three sites were separately re-deriving "the provider models this name resolves
to", with unexplained divergence in whether they fell back to the literal name.
`Router.resolved_litellm_models` is now the one owner; the routing-plugin
candidate list and the stream-options check both delegate to it, and
`_deployment_litellm_model` is gone.

The router's arms come from `strategy_router_dependencies`, the same enumeration
the health check reads. Only the roles that serve are arms: a classifier or
embedding model picks the tier and never produces a response anyone judges. A
semantic auto-router keeps its routes in an opaque config blob, so only its
default model is enumerable and the guard is incomplete there by design, able to
miss a collision but never to invent one

The two regenerated artifacts carry `presidio_analyze_chunk_size_bytes` from
alters the spec; the sync gate runs on any PR touching litellm/proxy, so this one
has to carry the base's drift to go green
2026-08-27 18:44:44 -07:00
..
_experimental/mcp_server feat(mcp): let a resolved OAuth token target a custom upstream header (#38456) 2026-08-27 14:32:01 -07:00
a2a fix(a2a): reject malformed protocolVersion suffixes while keeping semver prereleases 2026-07-21 13:24:58 -07:00
agent_endpoints refactor(repositories): type prisma table access with one generic protocol 2026-08-25 12:14:17 +00:00
analytics_endpoints feat(ui): add error-code drilldown for failed requests on caching page 2026-08-24 15:45:45 -07:00
anthropic_endpoints refactor(repositories): type prisma table access with one generic protocol 2026-08-25 12:14:17 +00:00
auth feat(proxy): gate fallback model access enforcement behind enforce_fallback_model_access 2026-08-27 15:45:10 -07:00
batches_endpoints test: add regression coverage for twelve closed issues (#37974) 2026-08-22 22:24:05 +00:00
client test: drop the cwd-relative sys.path.insert calls from the test suite (#37802) 2026-08-22 09:25:58 -07:00
common_utils feat(proxy): opt-in budget rollover carrying overage into the next window (#38514) 2026-08-27 12:46:09 -07:00
config_resolvers feat(proxy): add SAML 2.0 SSO for the admin UI (#31429) 2026-07-24 12:51:28 -07:00
credential_endpoints test: drop the cwd-relative sys.path.insert calls from the test suite (#37802) 2026-08-22 09:25:58 -07:00
db fix: keep schema reconciliation from fighting a partitioned LiteLLM_SpendLogs (#38452) 2026-08-27 12:52:52 -07:00
discovery_endpoints test: drop the cwd-relative sys.path.insert calls from the test suite (#37802) 2026-08-22 09:25:58 -07:00
enterprise_billing feat(proxy): push-based OTLP billable-request metering for enterprise deployments (#31592) 2026-07-15 12:12:52 -07:00
experimental/mcp_server test: drop the cwd-relative sys.path.insert calls from the test suite (#37802) 2026-08-22 09:25:58 -07:00
fine_tuning_endpoints test: drop the cwd-relative sys.path.insert calls from the test suite (#37802) 2026-08-22 09:25:58 -07:00
google_endpoints test: drop the cwd-relative sys.path.insert calls from the test suite (#37802) 2026-08-22 09:25:58 -07:00
guardrails fix(guardrails): add fail-open mode to CrowdStrike AIDR guardrail (#38568) 2026-08-27 17:57:35 -07:00
health_endpoints feat(alerting): add native Microsoft Teams alerting destination (#38367) 2026-08-27 16:19:22 -07:00
hooks fix(interactions): poll queued background creates and drop the poll's deployment identity 2026-08-22 17:35:19 -07:00
image_endpoints test: drop the cwd-relative sys.path.insert calls from the test suite (#37802) 2026-08-22 09:25:58 -07:00
logging_endpoints
management_endpoints fix(shadow_eval): refuse a judge model that also serves one of the arms it grades (#38589) 2026-08-27 18:44:44 -07:00
management_helpers fix(key_management): allow /key/update to keep or shrink MCP server grants the key already holds (#38463) 2026-08-27 12:51:17 -07:00
memory fix(memory): return 404 when the memory row vanishes before delete 2026-08-25 16:20:17 -07:00
middleware feat(proxy): add Amazon Comprehend Medical passthrough provider 2026-08-17 15:44:06 -07:00
ocr_endpoints fix(ocr): validate body req_format in the proxy endpoint and run its tests in CI 2026-08-17 18:29:35 +00:00
openai_files_endpoint test: add regression coverage for twelve closed issues (#37974) 2026-08-22 22:24:05 +00:00
pass_through_endpoints test(pass_through): pin stream pricing tests to injected divergent rate cards 2026-08-26 16:54:05 -07:00
policy_engine test(callbacks): unwind the callbacks global the policy engine and realtime tests scaffold around (#37826) 2026-08-21 21:19:35 -07:00
prompts fix(prompts): propagate prompt deletes to every worker and pod 2026-08-26 18:01:08 -07:00
proxy_server fix(router): reject complexity-router settings written outside complexity_router_config (#38570) 2026-08-27 17:04:49 -07:00
public_endpoints test: drop the cwd-relative sys.path.insert calls from the test suite (#37802) 2026-08-22 09:25:58 -07:00
rag_endpoints feat(proxy): enforce vector-store upload security controls on /v1/rag/ingest 2026-08-24 12:41:15 -07:00
realtime_endpoints test: drop the cwd-relative sys.path.insert calls from the test suite (#37802) 2026-08-22 09:25:58 -07:00
rerank_endpoints fix(rerank): emit latency and cost headers on /rerank (#35419) 2026-08-25 15:54:25 -07:00
response_api_endpoints fix(completion_extras): forward reasoning_effort=max through the Responses API bridge 2026-08-25 09:40:23 -07:00
shutdown
spend_tracking feat(ui): session-level cache observability in request logs (#38442) 2026-08-27 17:10:01 -07:00
test_configs
types_utils fix(proxy): resolve router_settings.plugins dotted paths and load plugins from installed packages (#33644) 2026-07-17 11:23:18 -07:00
ui_crud_endpoints Merge remote-tracking branch 'origin/litellm_internal_staging' into litellm_decrease_anys_opus5_round2 2026-08-25 16:17:29 -07:00
utils fix(responses): run prompt hook before provider credential resolution in sync responses() 2026-08-26 15:08:23 -07:00
vector_store_endpoints refactor(repositories): type prisma table access with one generic protocol 2026-08-25 12:14:17 +00:00
vector_store_files_endpoints test: drop the cwd-relative sys.path.insert calls from the test suite (#37802) 2026-08-22 09:25:58 -07:00
video_endpoints fix(videos): forward uploaded source file on /v1/videos/edits to the provider 2026-08-24 15:34:21 -07:00
__init__.py
conftest.py test(proxy): keep a leaked llm_router out of the next test in the worker 2026-08-20 23:19:17 -07:00
test__types.py test(proxy): pin the request-validation contracts in proxy/_types.py 2026-08-25 23:29:26 -07:00
test_aiohttp_cleanup_closed.py
test_aiohttp_session_recovery.py
test_api_key_masking_in_errors.py
test_audio_speech_prometheus_hooks.py
test_batch_expiry.py test: drop the cwd-relative sys.path.insert calls from the test suite (#37802) 2026-08-22 09:25:58 -07:00
test_batch_metadata_none_fix.py test: drop the cwd-relative sys.path.insert calls from the test suite (#37802) 2026-08-22 09:25:58 -07:00
test_batch_retrieve_bedrock.py test: drop the cwd-relative sys.path.insert calls from the test suite (#37802) 2026-08-22 09:25:58 -07:00
test_batch_x_litellm_model_encoding.py test(proxy): give the x-litellm-model fallback test deterministic openai env creds 2026-08-17 12:42:57 -07:00
test_blocked_response_usage.py fix(responses_api): map bridged chat usage on guardrail-blocked replies 2026-08-14 17:04:27 -07:00
test_budget_reservation.py fix(router): forward the hold-back keepalive ping live and carry the withheld-output flag through the stream wrappers 2026-08-25 16:39:14 -07:00
test_caching_routes.py test: drop the cwd-relative sys.path.insert calls from the test suite (#37802) 2026-08-22 09:25:58 -07:00
test_chat_completion_metadata.py
test_claude_code_marketplace.py test: run the 30 test files stranded in the second mirror (#37595) 2026-08-20 10:59:43 -07:00
test_common_request_processing.py fix(logging): stop billing and logging response reads as LLM calls (#36890) 2026-08-26 18:34:17 -07:00
test_component_allowlists.py feat(proxy): authenticate to Azure Postgres with Microsoft Entra ID tokens 2026-08-20 11:50:16 -07:00
test_conftest.py test(proxy): stop monkeypatch.undo re-planting fixture-mocked prisma_client 2026-08-13 20:11:06 -07:00
test_cors_config.py
test_custom_proxy.py test: drop the cwd-relative sys.path.insert calls from the test suite (#37802) 2026-08-22 09:25:58 -07:00
test_dynamic_mcp_route.py fix(proxy): serve aggregate MCP endpoint on bare /mcp instead of 307-redirecting (#34845) 2026-08-14 17:04:32 -07:00
test_empty_model_list.py test: drop the cwd-relative sys.path.insert calls from the test suite (#37802) 2026-08-22 09:25:58 -07:00
test_enforce_user_param.py test: require a match= on broad pytest.raises, and drop duplicate parametrize cases (#37769) 2026-08-20 20:24:49 -07:00
test_fallback_management_endpoints.py
test_fastapi_offline_routes.py test: drop the cwd-relative sys.path.insert calls from the test suite (#37802) 2026-08-22 09:25:58 -07:00
test_filter_models_by_team_access_group.py test: drop the cwd-relative sys.path.insert calls from the test suite (#37802) 2026-08-22 09:25:58 -07:00
test_health_check_functions.py feat(health): opt-in model-group allowlist for background health checks and health-check routing (#38539) 2026-08-27 12:25:56 -07:00
test_health_check_max_tokens.py fix(proxy): derive auto-router health from its underlying models (#38174) 2026-08-26 16:41:54 -07:00
test_init_litellm_callbacks.py test: run the 30 test files stranded in the second mirror (#37595) 2026-08-20 10:59:43 -07:00
test_langfuse_passthrough_security.py
test_lazy_openapi_snapshot.py refactor(proxy): type the snapshot fragments and wrap a long test line 2026-08-26 14:39:51 -07:00
test_litellm_pre_call_utils.py feat(otel): support per-team/per-key service.name for OTel v2 destinations (#38532) 2026-08-27 11:50:41 -07:00
test_max_budget_env_var.py
test_mcp_asgi_response.py
test_model_based_routing_files_batches.py
test_model_deprecations_endpoint.py test: drop the cwd-relative sys.path.insert calls from the test suite (#37802) 2026-08-22 09:25:58 -07:00
test_model_dump_with_preserved_fields.py
test_model_id_header_propagation.py feat(proxy): add x-litellm-model-name response header with deployment model string (#33698) 2026-07-17 20:29:42 -07:00
test_model_info_default_limits.py
test_model_level_guardrails.py fix(proxy): apply key/team router_settings.model_group_alias (#35486) 2026-08-03 22:09:47 +00:00
test_model_list_healthy_only.py feat(proxy): hide unhealthy models from model listings, opt-in 2026-08-26 00:13:52 -07:00
test_modify_response_streaming_passthrough.py
test_openai_ws_passthrough_routes.py fix(proxy): close websocket cleanly when OpenAI credentials are missing 2026-08-16 14:40:37 -07:00
test_openapi_schema_validation.py
test_plugin_routes.py
test_pricing_field_strip.py test: drop the cwd-relative sys.path.insert calls from the test suite (#37802) 2026-08-22 09:25:58 -07:00
test_prisma_engine_watchdog.py test: run the 30 test files stranded in the second mirror (#37595) 2026-08-20 10:59:43 -07:00
test_prisma_migration.py fix(proxy): keep a failed prisma generate from failing the migration entrypoint (#37947) 2026-08-22 11:45:19 -07:00
test_prometheus_cleanup.py
test_provider_url_destination_guard.py test: drop the cwd-relative sys.path.insert calls from the test suite (#37802) 2026-08-22 09:25:58 -07:00
test_proxy_cli.py test: drop the cwd-relative sys.path.insert calls from the test suite (#37802) 2026-08-22 09:25:58 -07:00
test_proxy_logging_hook_detection.py test: enforce PT012 so a pytest.raises block cannot hide dead assertions (#37748) 2026-08-20 19:36:26 -07:00
test_proxy_server.py Merge remote-tracking branch 'origin/litellm_internal_staging' into litellm_fallback_access_group_check 2026-08-27 17:29:08 -07:00
test_proxy_types.py test: drop the cwd-relative sys.path.insert calls from the test suite (#37802) 2026-08-22 09:25:58 -07:00
test_proxy_utils.py fix(proxy): stop expected 4xx responses from saturating worker CPU on failure logging (#38102) 2026-08-25 12:44:16 -07:00
test_pyroscope.py
test_read_model_list.py
test_redis_auth_cache_flag.py test: add six ruff rules that catch tests which cannot fail (#37709) 2026-08-20 14:21:26 -07:00
test_response_model_sanitization.py test: drop the cwd-relative sys.path.insert calls from the test suite (#37802) 2026-08-22 09:25:58 -07:00
test_route_a2a_models.py test: drop the cwd-relative sys.path.insert calls from the test suite (#37802) 2026-08-22 09:25:58 -07:00
test_route_llm_request.py test: drop the cwd-relative sys.path.insert calls from the test suite (#37802) 2026-08-22 09:25:58 -07:00
test_sensitive_route_auth.py
test_shared_health_check.py fix(proxy): derive auto-router health from its underlying models (#38174) 2026-08-26 16:41:54 -07:00
test_spend_log_cleanup.py test: require a match= on broad pytest.raises, and drop duplicate parametrize cases (#37769) 2026-08-20 20:24:49 -07:00
test_swagger_chat_completions.py
test_team_member_update.py
test_team_org_move.py test: require a match= on broad pytest.raises, and drop duplicate parametrize cases (#37769) 2026-08-20 20:24:49 -07:00
test_tools_allowlist_enforcement.py fix(anthropic): close hybrid tool-name allowlist gap and keep native tools through guardrails 2026-08-26 20:37:11 -07:00
test_update_llm_router_resilience.py feat(complexity_router): custom classifier plugins via classifier_type 'custom' (#37249) 2026-08-18 14:09:19 -07:00