feat(providers): add LLM sampling controls - #427
Conversation
rng1995
left a comment
There was a problem hiding this comment.
[SkillSpector Review]\n\nApproved. Sampling controls are validated centrally and mapped appropriately across OpenAI-compatible, Azure, Anthropic, Bedrock, and proxy providers, with unsupported seed parameters omitted where required. Current required checks pass.
6148d41 to
ccae0e7
Compare
Signed-off-by: Deepak Jain <deepujain@gmail.com>
Head branch was pushed to by a user without write access
bb495fe to
705bc1d
Compare
|
Resolved the conflict with current |
Summary
SKILLSPECTOR_TEMPERATUREsampling control for hosted providersSKILLSPECTOR_SEEDforwarding for OpenAI-compatible and Azure OpenAI providersThis addresses the sampling-control portion of #334. Provider/model attribution in generated reports is intentionally unchanged because that requires a separate report-schema decision.
Validation
pytest tests/unit/test_providers.py tests/unit/test_anthropic_proxy_provider.py tests/unit/test_new_providers.py tests/unit/test_bedrock_provider.py tests/unit/test_constants.py -q— 176 passed, 9 skippedpytest -m 'not integration and not provider' tests/ -q— 2804 passed, 13 skipped, 38 deselected, 4 xfailedruff check src testsruff format --check src testsgit diff --checkRisk
0..1range; seed support remains provider/model dependent