v0.4.0 - Model Selection, Skill Import & Agent Refinements / 模型选择、技能导入与智能体优化

Release v0.4.0 / 发布 v0.4.0

Added / 新增

• Agent Model Selection / 智能体模型选择: Per-session LLM provider & model selection in agent panel with localStorage persistence and input validation (PR #238–#240)
  智能体面板支持按会话选择 LLM 提供方与模型，支持本地持久化与输入校验
• Skill Import via Markdown / Markdown 技能导入: Import skills from .md files with allowed-tools validation, scope guard, and memoized parsing (PR #234, #235)
  支持通过 Markdown 文件导入技能，增加允许工具校验、作用域保护和缓存解析
• Rolling Work Mode / 滚动工作模式: Auto-scrolling work panel with configurable bottom threshold and useCallback optimization (PR #231, #233)
  自动滚动工作面板，可配置底部阈值，优化 useCallback
• GitHub Star History / GitHub Star 趋势图: Star history trend chart added to README (PR #7)
  README 新增 GitHub Star 历史趋势图
• Archive Process Code / 归档流程代码: Code archiving feature with history skip check improvements (PR #244, #245)
  新增代码归档功能，改进历史跳过检查

Changed / 变更

• Codebase Refactor / 代码重构: General codebase cleanup and restructuring (PR #236, #237)
  整体代码清理与结构优化
• README Updates / README 更新: Add research ideation, multi-tab preview, and Study Paper feature descriptions (PR #236)
  README 新增研究构思、多标签预览和论文研究功能说明

Fixed / 修复

• Agent/Agent Name Overlap / 智能体名称重叠: Fix agent session tab display where "agent" and "Agent" overlapped; remove shrink-0 for proper flex shrinking (PR #241–#243)
  修复智能体会话标签 "agent" 与 "Agent" 显示重叠，修复 flex 收缩
• Path Separator in History Check / 历史检查路径分隔符: Use path.sep in history skip check to avoid false matches on different OS (PR #245)
  历史跳过检查使用 path.sep 避免跨平台路径误匹配
• Skill MD Form / 技能 Markdown 表单: Validate allowed-tools list and add scope guard for skill import (PR #235)
  技能导入校验允许工具列表，增加作用域保护

---

Full Changelog / 完整变更日志: v0.3.0...v0.4.0

v0.3.0 - Domestic Models, Persistent Agents & Paper Discussion / 国产模型、持久化智能体与论文讨论

Release v0.3.0 / 发布 v0.3.0

Added / 新增

• Domestic Model Providers / 国产模型支持: SH-Lab model integration with configurable base URLs and env var overrides; dynamic isAIAvailable() derived from provider config (PR #211–#217, #219)
  新增上海实验室模型集成，支持可配置 base URL 与环境变量覆盖；动态 AI 可用性检测
• Paper Discussion Mode / 论文讨论模式: AI-powered paper discussion panel with multi-agent conversations and note management (PR #218, #226)
  AI 驱动的论文讨论面板，支持多智能体对话和笔记管理
• Persistent Agents & Auto-Continue / 持久化智能体与自动继续: Agent sessions persist across page navigation; configurable AGENT_MAX_STEPS with auto-continue support (PR #227–#228)
  智能体会话跨页面持久化；可配置最大步数与自动继续功能
• GitHub Pages Documentation / GitHub Pages 文档站: Project documentation site with institutional logos (PKU, TUM, NJU) (PR #207–#210)
  项目文档站上线，新增北大、慕尼黑工大、南大机构标识
• Daily Change Summary / 每日变更摘要: Automated daily summary workflow with Copilot CLI and Feishu notification
  自动每日变更摘要工作流，集成 Copilot CLI 与飞书通知
• K8s Deployment / K8s 部署: Kubernetes deployment support
  新增 Kubernetes 部署支持

Changed / 变更

• Agent Panel Branding / 智能体面板品牌: Renamed agent panel title from "Claude Code" to "Agent" (PR #2)
  智能体面板标题从 "Claude Code" 更名为 "Agent"
• Skill Catalog / 技能目录: Add isEnabled filter to skill catalog query; fix slug normalization with separate fallbackName/fallbackSlug (PR #170, #173)
  技能目录查询增加启用状态过滤；修复 slug 规范化逻辑
• Accessibility / 无障碍: Add aria-hidden and inert attributes to collapsed layout wrapper in minimal mode (PR #187)
  最小化模式下折叠布局增加无障碍属性

Fixed / 修复

• Agent State Preservation / 智能体状态保持: Fix AgentPanel state loss when switching to minimal mode (PR #184, #187)
  修复切换最小化模式时智能体面板状态丢失
• Feishu Bot Docs / 飞书机器人文档: Fix setup instructions to match actual workflow (PR #185, #188)
  修复飞书机器人配置文档与实际流程不匹配
• Repo Sync CI / 仓库同步 CI: Disable persist-credentials to use SYNC_TOKEN for private repo push
  禁用凭据持久化以使用 SYNC_TOKEN 推送私有仓库
• Claude Code Skill / Claude Code 技能: Move setup skill to .claude/commands/ and update README (PR #2)
  将 Claude Code 设置技能迁移至 .claude/commands/ 并更新 README

---

Full Changelog / 完整变更日志: v0.2.0...v0.3.0

v0.2.0 - Multi-Agent & Paper Study / 多智能体与论文研究

Release v0.2.0 / 发布 v0.2.0

Added / 新增

• Multi-Agent System / 多智能体系统: Multi-agent sessions with adjustable dialogue panels and agent display (PR #201, #203, #206)
  多智能体会话，支持可调节对话面板和智能体展示
• Paper Study Panel / 论文研究面板: Paper study with notes management, AI-powered roast review, AI search, and Semantic Scholar integration (PR #202)
  论文研究面板，支持笔记管理、AI 吐槽式审稿、AI 搜索及 Semantic Scholar 集成
• Style Theme System / 风格主题系统: Cartoon, cyberpunk-pixel, and retro-handheld visual themes (PR #199)
  新增卡通、赛博像素、复古掌机三种视觉主题
• Cluster & Paper Pages / 集群与论文页面: Cluster management pages and enhanced UI with particle effects (PR #190)
  新增集群管理页面，增强 UI 粒子特效
• Fuzzy Matching / 模糊匹配: Fuzzy search support for improved content discovery (PR #197)
  模糊搜索支持，提升内容检索体验
• Font Customization / 字体自定义: Configurable font type and size with shared font constants (PR #191)
  可配置字体类型与大小，抽取共享字体常量
• Agent Panel Cluster UI / 智能体面板集群 UI: Left-panel cluster layout for agent management (PR #192-#196)
  智能体管理左侧集群面板布局

Changed / 变更

• Rename: LabClaw → InnoClaw / 项目更名: Full project rename from LabClaw to InnoClaw across the entire codebase
  全项目范围从 LabClaw 更名为 InnoClaw
• Agent Panel Improvements / 智能体面板改进: Auto-wrap support, adjustable dialogue, UI beautification (PR #189, #200)
  智能体面板自动换行、可调节对话、UI 美化
• README & Docs / 文档改进: Improved setup instructions, default docs language set to Chinese (PR #205)
  改进安装说明，文档默认语言设为中文
• File Tree / 文件树: Support for directory deletion (PR #198)
  支持目录删除操作

Fixed / 修复

• HuggingFace Search / HuggingFace 搜索: Fix tests failing on empty keywords and API errors
  修复空关键词和 API 错误导致的测试失败
• Context Overflow / 上下文溢出: Improved detection counting tool payloads and sanitizing malformed tool_use
  改进上下文溢出检测，计算工具负载并清理格式错误的 tool_use
• CI/CD: Update Node.js to v22, fix lockfile sync, use npm install for cross-version compatibility
  升级 Node.js 至 v22，修复锁文件同步，使用 npm install 以兼容不同版本
• Lint Fixes / 代码检查修复: Resolve react-hooks, paper-study components, and i18n key lint errors
  修复 react-hooks、论文研究组件及 i18n 键的代码检查错误
• Repo Sync / 仓库同步: Add workflow to sync public repo to private repo (CI)
  新增公有仓库到私有仓库的同步工作流

---

Full Changelog / 完整变更日志: v0.1.0...v0.2.0

v0.1.0 - Initial Open-Source Release

LabClaw v0.1.0

First open-source release of LabClaw - a self-hostable AI research assistant inspired by Google NotebookLM.

Highlights

• Workspace & File Management: Map server-side folders as workspaces with file browsing, uploading, editing, and Git integration
• RAG-enhanced Chat: AI conversations grounded in document content with source citations
• Smart Note Generation: Auto-generate summaries, FAQs, briefings, timelines, daily and weekly reports
• Multi-LLM Support: OpenAI, Anthropic (Claude), and Google Gemini with configurable providers
• Agent System: AI agent with tool-calling capabilities
• 206 SCP Scientific Skills: Built-in skills covering drug discovery, genomics, protein engineering, and more
• Feishu (Lark) Bot: WebSocket-based bot integration with interactive card messages
• WeChat Enterprise Bot: Bot integration for team collaboration
• Internationalization: Full Chinese and English bilingual support
• Dark Mode: Theme toggle with system preference detection

Getting Started

git clone https://github.com/zjowowen/LabClaw.git
cd LabClaw
npm install
cp .env.example .env.local
# Edit .env.local with your configuration
npm run dev

For detailed instructions, see the documentation.

License

Apache License 2.0