-
v0.3.46
Release / Desktop / macOS ARM64 (push) Has been cancelledRelease / Desktop / Windows ARM64 (push) Has been cancelledRelease / Desktop / macOS x86_64 (push) Has been cancelledRelease / Desktop / Windows x86_64 (push) Has been cancelledRelease / Desktop / Linux x86_64 (push) Has been cancelledRelease / CLI / aarch64-apple-darwin (push) Has been cancelledRelease / CLI / x86_64-apple-darwin (push) Has been cancelledRelease / CLI / aarch64-unknown-linux-gnu (push) Has been cancelledRelease / CLI / x86_64-unknown-linux-gnu (push) Has been cancelledRelease / CLI / aarch64-pc-windows-msvc (push) Has been cancelledRelease / CLI / x86_64-pc-windows-msvc (push) Has been cancelledRelease / Docker Image (push) Has been cancelled发布于
2026-03-11 00:25:29 +00:00 Community PR batch — 11 PRs implemented, 19 files changed:
- Fix #438: IME composition guard for CJK input (@pandego)
- Fix #433: Default User-Agent header on all LLM drivers (@ozekimasaki)
- Fix #417/#392: Default token quota unlimited instead of 1M (@f-liva, @cryptonahue)
- Fix #410: Safe string slicing in channel_bridge + shutdown race fix (@hobostay)
- Fix #413/#275: Embedding driver respects provider_urls config + smart model defaults (@castorinop, @woodcoal)
- Fix #464: Wizard uses multi-line TOML strings for system_prompt (@citadelgrad)
- Fix #419/#480/#439: Added Z.AI, Kimi for Code, Moonshot models + fixed base URL (@shipdocs, @skeltavik, @modship)
- Fix: XML-attribute tool call recovery (Pattern 9) for Groq/Llama models
- 187 models, 40+ providers, 2040+ tests, 0 clippy warnings
下载附件