ci: run provider replay harness in CircleCI

.circleci/config.yml

@@ -2915,6 +2915,25 @@ jobs:
               exit 1
             fi
 
+  provider_replay_harness:
+    docker:
+      - *python312_image
+    working_directory: ~/project
+    resource_class: medium
+    steps:
+      - setup_litellm_test_deps
+      - run:
+          name: Test provider replay harness
+          command: |
+            mkdir -p test-results/provider-replay-harness
+            uv run --no-sync pytest -q --noconftest -o addopts= -o pythonpath=tests/e2e -p no:rerunfailures \
+              --junitxml=test-results/provider-replay-harness/junit.xml \
+              tests/e2e/test_provider_edge.py tests/e2e/test_fixture_bundle.py \
+              tests/e2e/test_fixture_canonical.py tests/e2e/test_fixture_mode.py \
+              tests/code_coverage_tests/test_provider_replay_harness.py
+      - store_test_results:
+          path: test-results/provider-replay-harness
+
   integration_contracts:
     parameters:
       suite:
@@ -2967,6 +2986,8 @@ workflows:
               only:
                 - main
                 - /litellm_.*/
+      - provider_replay_harness:
+          filters: *main_branches
       - base_sdk_install:
           filters: *main_branches
       - local_testing_part1:

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

@@ -83,14 +83,6 @@ 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: test_provider_replay_harness
-        run: |
-          pwd
-          uv run --no-sync pytest -q --noconftest -o addopts= -o pythonpath=tests/e2e -p no:rerunfailures \
-            tests/e2e/test_provider_edge.py tests/e2e/test_fixture_bundle.py \
-            tests/e2e/test_fixture_canonical.py tests/e2e/test_fixture_mode.py \
-            tests/code_coverage_tests/test_provider_replay_harness.py
-
       - name: router_code_coverage
         run: uv run --no-sync python ./tests/code_coverage_tests/router_code_coverage.py
 

tests/e2e/CONTRIBUTING.md

@@ -247,4 +247,4 @@ The semantic header set is `content-type`, `accept`, `anthropic-version`, `anthr
 
 Excluded transport and telemetry headers are `host`, `content-length`, `connection`, `accept-encoding`, `user-agent`, `traceparent`, `tracestate`, `x-request-id`, `x-client-request-id` and `x-stainless-*`. Inbound transfer-encoding is unsupported; send JSON with content-length framing. The destination represents host identity and the relay carries original body bytes. Replay does not verify credentials, SDK timeout/retry behavior, transport performance, model availability or stateful remote IDs. Live relay uses original request bytes and header values, never the stored identity
 
-Strict replay harness regression tests live in `tests/code_coverage_tests/test_provider_replay_harness.py`. The `test_provider_replay_harness` code-quality step runs them alongside the existing legacy harness files with `--noconftest -o pythonpath=tests/e2e`; they need only synthetic HTTP providers and temporary fixture storage
+Strict replay harness regression tests live in `tests/code_coverage_tests/test_provider_replay_harness.py`. The CircleCI `provider_replay_harness` job runs them alongside the existing legacy harness files with `--noconftest -o pythonpath=tests/e2e`; they need only synthetic HTTP providers and temporary fixture storage