mmc/lingma-proxy-compose
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Release v1.4.4

Browse Source
This commit is contained in:
lutc5
2026-05-06 11:03:55 +08:00
parent a02fd51c19
commit a3a9c278f6
9 changed files with 542 additions and 42 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
CHANGELOG.md
View File

@@ -4,6 +4,14 @@
- Nothing yet.
## v1.4.4 - 2026-05-05
- Enabled real SSE streaming for OpenAI `/v1/chat/completions` and Anthropic `/v1/messages` requests that include tools.
- Added a tool-stream filter so normal text can stream immediately while prompt-emulated action blocks are buffered and emitted as proper `tool_calls` / `tool_use` events at the end.
- Added `LINGMA_AGGREGATE_TOOL_STREAM=1` as a compatibility switch to restore the previous aggregate output behavior for tool requests.
- Tightened tool-emulation instructions so conceptual chat and explanation requests do not trigger unnecessary terminal/tool calls.
- Added tests for hosted Anthropic web search handling, tool-stream filtering, and updated tool prompt guidance.
## v1.4.3 - 2026-04-30
- Added remote API timeout fallback with a configurable model order. The default order is Kimi-K2.6, MiniMax-M2.7, Qwen3-Coder, Qwen3.6-Plus, Qwen3-Max, and Qwen3-Thinking.