ThinkInAIXYZ/deepchat v1.0.3-beta.1 on GitHub

v1.0.3-beta.1 (2026-04-11)

Migrated model requests to the AI SDK runtime, improving prompt cache behavior, provider consistency, and streaming stability
Added NewAPI provider support and refined compatible endpoint configuration
Improved model management with more stable provider toggles and synchronized Ollama selectable model status
Added skill_view draft flow and automatic tool activation after skill previews to smooth skill setup
Enhanced Markdown and workspace link navigation, added sidebar panel toggle hotkeys, and fixed artifact viewer sizing in the side panel
将模型请求迁移到 AI SDK 运行时，进一步改善 Prompt Cache 表现、Provider 一致性与流式稳定性
新增 NewAPI Provider 支持，并完善兼容端点配置体验
改进模型管理，修复 Provider 模型开关稳定性并同步 Ollama 可选模型状态
新增 skill_view 草稿流，并在技能预览后自动激活工具，减少技能接入摩擦
优化 Markdown 与工作区链接跳转体验，新增侧栏面板切换快捷键，并修复侧边栏制品预览高度问题