mirror of
https://github.com/BerriAI/litellm.git
synced 2026-11-07 07:58:09 +00:00
The pre_call header sweep claimed every policy pipeline step up front, so steps a pipeline never reached showed up in x-litellm-applied-guardrails and in spend logs. The sweep now leaves pipeline steps to the executor, which names each step when its hook runs, whatever it decides. Steps a stream runs after its headers go out (a pipeline the stream cannot hold, or one whose steps all stream live) are still named up front, as before. Resolves LIT-10096 |
||
|---|---|---|
| .. | ||
| __init__.py | ||
| test_attachment_registry.py | ||
| test_condition_evaluator.py | ||
| test_pipeline_executor.py | ||
| test_policy_engine_endpoints.py | ||
| test_policy_matcher.py | ||
| test_policy_registry.py | ||
| test_policy_resolver.py | ||
| test_policy_validator.py | ||
| test_policy_versioning.py | ||
| test_policy_versioning_e2e.py | ||
| test_response_retrieval.py | ||