-
v0.4.5: reasoning models + multi-provider + REPL input hardening
frostbyte_neo 发布于
2026-05-13 12:51:46 +00:00 Headline:
- Thinking content blocks end-to-end (fixes #161) + multi-tool grouping (#186)
- reasoning_effort='xhigh' actually on the wire for GPT-5.5/o1/o3/o4
- DeepSeek V4 Pro + Xiaomi MiMo + Qwen 3.6 + Doubao in /setup
- Claude Code object-style hooks parser (#171)
- Default model: Claude Opus 4.7 + GPT-5.5
- REPL input: multi-line wrap / bracketed paste / CJK wide char fixes
- GitHub Actions CI workflow
Credits: @GO-player-hhy (#186), @Jxy-yxJ (#171), @GetIT-Sunday (#216 partial)
See CHANGELOG.md for full details.
下载附件