mirror of
https://github.com/BerriAI/litellm.git
synced 2026-11-20 10:02:54 +00:00
Every full catalog reload (the 30 second reload_mcp_servers_job, the reload after an admin write, add_deployment) wiped every server's cached upstream initialize instructions. Servers that need per-user auth cannot be prefetched, so gateway initialize returned no instructions for them until a user listed tools again. The reload already drops the entry of each server whose definition changed or was removed, so the global wipe only ever cleared unchanged servers. Remove it and the method that only it used. That wipe was also the only thing that expired two kinds of stale text, so fix both at the source. Each upstream session now records whether initialize finished and what it said, and an answer without instructions drops the cached entry. A per-server generation, bumped whenever a server's definition changes, is read before every upstream session, and a session that outlived an edit of its server no longer writes the old upstream's text back. Resolves LIT-9851 |
||
|---|---|---|
| .. | ||
| _support | ||
| audio_tests | ||
| 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.