agentjido / req_llm / 0f26fffa1224465c9a278b7a1a6204b6717c243d

49%

Build Type

push

github

Committed by web-flow

Commit Message

Fix OpenAI responses context replay and Anthropic structured output decode (#228)

* Add tests and new fixtures to illustrate the content append bug for OpenAI

* Fixed OpenAI Responses API assistant message re-encoding issue

Generated fixtures for 4 specific models - feel free to adjust.

REQ_LLM_FIXTURES_MODE=replay
REQ_LLM_MODELS="openai:gpt-5-mini,google:gemini-flash-latest,xai:grok-4-fast,anthropic:claude-haiku-4-5-20251001"
mix test --include coverage
test/coverage/{openai,google,xai,anthropic}/comprehensive_test.exs

* Fix for unrelated Anthropic coverage failure

Run Details

2 of 17 new or added lines in 2 files covered. (11.76%)

1 existing line in 1 file now uncovered.

3641 of 7438 relevant lines covered (48.95%)

57.34 hits per line