agentjido / req_llm / 1d67c3d1fc988dcca8f9161bb3737b48993e2052

49%

push

github

Fix: Translate max_tokens to max_completion_tokens for OpenAI reasoning models (#58)

* fix: translate max_tokens to max_completion_tokens for OpenAI reasoning models

- Extended parameter translation to cover all reasoning models (GPT-5, O4, Codex)
- Previously only O1/O3 models were handled, causing API errors for newer models
- Models with reasoning capability now properly use max_completion_tokens
- Temperature parameter still supported for GPT-5 models (only O-series drop it)
- Added comprehensive tests for GPT-5, O4, and Codex model parameter translation

Fixes #50

* Formatting

25 of 27 new or added lines in 1 file covered. (92.59%)

1745 of 2960 relevant lines covered (58.95%)

313.18 hits per line