mirror of
https://github.com/BerriAI/litellm.git
synced 2026-10-29 06:33:28 +00:00
7d31f0ac36
`provider: nadir` routes the auto router on Nadir's /v1/bucket verdict, beside jev, laya, bespoke and databricks. Nadir is not a System One server, so the classifier client translates both ways. Out: Nadir's classifier is trained on raw conversations, so it gets role-tagged messages, not the LLM-judge prompt the other providers read. They carry the same caller system prompt and the same budgeted prior turns, so classifier_context_window_size and its siblings bound what leaves the gateway exactly as before; with a window of 0 only the current ask is sent. The messages ride on an excluded field, so no System One request ever carries them. In: the verdict's `routing_tier` (falling back to `bucket`) becomes the `tier` choice answer, with canonical tier labels the router already resolves. A verdict naming no usable tier, or a non-2xx, falls back like any other classifier failure. NADIR_API_KEY is only sent to NADIR_API_BASE or the hosted API, the same scoping as the nadir chat provider. With no key the call is anonymous, so an api_base without an api_key is allowed for this provider. `model` defaults to `bucket` and only labels logs; `instructions` is rejected because Nadir classifies with its own model. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> |
||
|---|---|---|
| .. | ||
| _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.