wangwei 05419db1f9 fix: support max_completion_tokens for newer models (gpt-5.x) in connectivity test ...

Newer OpenAI models (gpt-5.4 etc.) reject max_tokens and require
max_completion_tokens. Try max_completion_tokens first, fall back to
max_tokens for older models / compatible APIs.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

2026-06-23 14:51:28 +08:00

..

api

fix: support max_completion_tokens for newer models (gpt-5.x) in connectivity test

2026-06-23 14:51:28 +08:00

services

feat: add InlineScorer service with LLM client caching

2026-06-22 15:03:43 +08:00

static

fix: restore LLM profile test connectivity buttons (lost from git)

2026-06-23 13:58:43 +08:00

__init__.py

Add RAGAS evaluation web console (FastAPI + vanilla JS)

2026-06-15 15:53:57 +08:00

models.py

fix: change ScoreRequest json_schema_extra from examples list to example dict

2026-06-23 10:03:46 +08:00

server.py

feat: add detailed logging to all API routes and global access log middleware

2026-06-23 10:35:00 +08:00