mirror of
https://github.com/BerriAI/litellm.git
synced 2026-10-26 06:03:19 +00:00
With output_parse_pii enabled, Presidio's restore callback implements apply_guardrail, so on /v1/messages the proxy downgraded its native streaming hook to the unified scan-only path, which drops text rewrites. Clients got the raw <PERSON_1> placeholder back instead of the original value. Setting use_native_lifecycle_hooks keeps the restore callback on its own hook, matching every other guardrail that rewrites streamed content. That alone still corrupted a placeholder split across two SSE deltas, since each delta was unmasked in isolation. _unmask_sse_bytes_chunk now carries a pending buffer across chunks for the same stream, holding back only text that could still grow into a token, and flushes it with the truncation-fallback once the block actually ends (not on a keep-alive ping). Fixes #46087 |
||
|---|---|---|
| .. | ||
| _support | ||
| base_sdk_tests | ||
| benchmarks | ||
| code_coverage_tests | ||
| documentation_tests | ||
| e2e | ||
| e2e_harness | ||
| fixtures | ||
| harness_e2e | ||
| integration | ||
| rust-python-harness | ||
| test_litellm_rust | ||
| typing | ||
| unit | ||
| windows_tests | ||
| __init__.py | ||
| _fake_openai_endpoint_server.py | ||
| _flush_vcr_cache.py | ||
| _live_test_helpers.py | ||
| _master_key.py | ||
| _openai_record_replay_proxy.py | ||
| _process_helpers.py | ||
| _vcr_conftest_common.py | ||
| _vcr_redis_persister.py | ||
| _wait_helpers.py | ||
| _ws_vcr.py | ||
| AGENTS.md | ||
| capturing_transport.py | ||
| eval_swe_bench.py | ||
| fake_openai_endpoint.py | ||
| gettysburg.wav | ||
| large_text.py | ||
| openai_batch_completions.jsonl | ||
| pyrightconfig.json | ||
| README.MD | ||
| test_logging.conf | ||
| test_rust_python_harness.py | ||
| VCR_TESTING.md | ||
| white_100x100.png | ||
In total litellm runs 1000+ tests
[02/20/2025] Update:
To make it easier to contribute and map what behavior is tested,
we've started mapping the litellm directory in tests/unit
This folder can only run mock tests.