v1.6.0

🌐 English Version

Release v1.6.0

🚀 New Features & Enhancements

• Agentic Insight v2 Enhancement: Architecture and performance optimization for the deep research system. (#886)
• Knowledge Search Support: Added Sirchmunk-based knowledge search module, enabling automatic retrieval over local codebases and documents during agent conversations. (#882)
• Multimodal Model Support: Added multimodal capabilities for LLM agents, supporting image/video input via OpenAI-compatible API. (#879)
• Context Compression Support: Token usage monitoring and overflow detection, with automatic context compression via pruning historical tool outputs and LLM summarization. (#885)
• Auto Skills Logging: Added configurable log level settings for the Auto Skills module. (#883)

🛡️ Security

• CVE-2026-2256 Fix: Fixed a Shell tool regex denylist bypass leading to command injection. Removed the vulnerable Shell tool and migrated Code Genesis project code execution to a sandbox environment. (#880)
• Added SECURITY.md with vulnerability reporting guidelines. (#877)

🛠 Bug Fixes & Improvements

• Fixed WebUI Windows compatibility issues. (#863)
• Fixed MCP UI resources typo. (#866)

🧪 Testing

• Added comprehensive unit tests for Skill DAG upstream-downstream task analysis. (#867)

📝 Documentation

• Added Code Genesis documentation in both Chinese and English. (#865)
• Updated video generation documentation. (#864)
• Updated README. (#874)

Full Changelog: v1.6.0rc1...v1.6.0

---

🇨🇳 中文版本

版本更新 v1.6.0

🚀 新功能与增强

• Agentic Insight v2 增强：深度研究系统架构与性能优化。(#886)
• 知识库搜索支持：新增基于 Sirchmunk 的知识搜索模块，支持在 Agent 对话中自动检索本地代码库与文档。(#882)
• 多模态模型支持：为 LLM Agent 新增多模态能力，支持通过 OpenAI 兼容 API 进行图片/视频输入。(#879)
• 上下文压缩支持：支持 Token 用量监控与溢出检测，通过裁剪历史工具输出和 LLM 摘要自动压缩对话上下文。(#885)
• Auto Skills 日志配置：为 Auto Skills 模块新增可配置的日志级别设置。(#883)

🛡️ 安全修复

• CVE-2026-2256 修复：修复了 Shell 工具正则表达式黑名单绕过导致的命令注入漏洞，移除了存在风险的 Shell 工具，并迁移 Code Genesis 项目代码运行过程到沙箱环境。(#880)
• 新增 SECURITY.md 安全漏洞报告指南。(#877)

🛠 修复与改进

• 修复了 WebUI 的 Windows 兼容性相关问题。(#863)
• 修复了 MCP UI 资源拼写错误。(#866)

🧪 测试

• 新增 Skill DAG 上下游任务分析的完整单元测试。(#867)

📝 文档更新

• 新增 Code Genesis 中英文文档。(#865)
• 更新视频生成文档。(#864)
• 更新 README。(#874)

完整变更日志: v1.6.0rc1...v1.6.0

v1.6.0rc1

🌐 English Version

Release v1.6.0rc1

🚀 New Features & Enhancements

• DeepResearch Refactor: Significant refactoring of the DeepResearch module for better performance and structure. ([#858](#858))

• Video Generation Improvements:

  • Added comprehensive video documentation and demos in both Chinese and English. ([#842](#842), [#854](#854), [#856](#856))
  • Fixed Chinese character rendering issues in video generation. ([#855](#855))
  • Optimized video generation by excluding edit_file tool where unnecessary. ([#861](#861))

• Robustness: Updated the README with detailed sections on output formats and improved retry logic. ([#857](#857))

🛠 Bug Fixes & Refactors

• API & Connectivity: Fixed API endpoint issues to ensure stable service connections. ([#850](#850))
• Code Quality: Performed linting, fixed broken links, and added debug logs for better traceability. ([#853](#853), [#859](#859))

📝 Documentation

• Major Updates: Merged Release 1.6 changes and performed multiple rounds of README enhancements across all modules. ([#851](#851), [#852](#852), [#862](#862))

Full Changelog: [v1.6.0rc0...v1.6.0rc1](v1.6.0rc0...v1.6.0rc1)

---

🇨🇳 中文版本

版本更新 v1.6.0rc1

🚀 新功能与增强

• DeepResearch 重构：对 DeepResearch 模块进行了深度重构，优化了核心架构。([#858](#858))

• 视频生成功能优化：

  • 新增了视频相关的 README 文档及中英文演示示例。([#842](#842), [#854](#854), [#856](#856))
  • 修复了视频生成过程中的中文字符显示问题。([#855](#855))
  • 在视频生成流程中优化了工具调用，排除了不必要的 edit_file 操作。([#861](#861))

• 稳定性提升：在文档中新增了关于输出格式说明及重试逻辑的详细章节。([#857](#857))

🛠 修复与重构

• API 修复：修复了 API 终端连接问题，提升了服务调用的稳定性。([#850](#850))
• 代码质量：完成了代码 Lint 检查，修复了失效链接，并新增了调试日志以方便问题排查。([#853](#853), [#859](#859))

📝 文档更新

• 文档维护：合并了 Release 1.6 的相关改动，并对项目 README 进行了多轮润色与内容补充。([#851](#851), [#852](#852), [#862](#862))

完整变更日志: [v1.6.0rc0...v1.6.0rc1](v1.6.0rc0...v1.6.0rc1)

v1.6.0rc0

🚀 Release Notes - v1.6.0rc0

The v1.6.0rc0 version introduces significant enhancements to the WebUI, multi-modal capabilities (video and animation), and a major refactor of the Skills system.

🌟 Highlights

• WebUI: A new, simple WebUI has been added, featuring "Code Genesis," video support, and enhanced project management.
• Skills System Refactor: Multi-skills DAGs analysis and execution; New skill pipeline.
• Multi-modal Enhancements: Introduction of a new animation engine and major refactors to video generation and subtitle processing.

---

📝 What's Changed

🎨 WebUI & User Experience

• Code Genesis: Multiple iterations and enhancements to the "Code Genesis" feature within the WebUI by @vinci-grape ([#837](#837), [#841](#841), [#843](#843), [#844](#844), [#847](#847)).
• Video Integration: Added video support to the WebUI by @suluyana ([#836](#836)).
• MCP Resources: Supported MCP UI resources by @suluyana ([#818](#818)).
• Infrastructure: Added a simple WebUI framework, fixed build issues, and optimized for conda environments ([#827](#827), [#831](#831), [#830](#830), [#840](#840)).

🧠 Agent Skills & Core Features

• Skills Refactor: Comprehensive refactor of the Skills system by @wangxingjun778 ([#845](#845)).
• Hub Integration: Added support for downloading skills directly from ModelScope Hub via repo_id by @wangxingjun778 ([#846](#846)).
• Financial Research: Enhanced financial research capabilities by @alcholiclg ([#809](#809)).
• CLI Management: Added support for running projects via CLI by @suluyana ([#838](#838)).
• Usage Tracking: Added feature to track current resource usage by @suluyana ([#821](#821)).

🎬 Video, Audio & Animation

• Animation Engine: Introduced a new animation engine by @xiaoxianhjy ([#825](#825)).
• Video Generation: Refactored the video generation pipeline by @tastelikefeet ([#826](#826)).
• Subtitles & TTS: Supported multiple subtitles per segment and extended TTS voice options; fixed subtitle merging bugs ([#802](#802), [#828](#828)).

🛠 Bug Fixes & Internal Improvements

• Output Optimization: Removed duplicate output for agent run tasks ([#822](#822)).
• Memory Management: Memory-related optimizations and fixes ([#820](#820)).
• Maintenance: Code linting, documentation updates (README), and internal comments added ([#824](#824), [#829](#829), [#848](#848), [#849](#849)).

---

🆕 New Contributors

• @yaqiangsun made their first contribution in [#822](#822)!

Full Changelog: [v1.5.2...v1.6.0rc0](v1.5.2...v1.6.0rc0)

v1.5.2

Release Notes

📚 Documentation Improvements

• Fixed broken links and formatting in Chinese documentation
  • Corrected doc links in index.rst and Projects/fin-research reference (#790, #791, #792)
  • Updated agent skills documentation (#793)
  • Improved English documentation and site configuration (#794, #795)
  • Added related links to RinResearch README (#796)
    — by @wangxingjun778

🛠️ Bug Fixes

• Code Generation & Parsing

  • Fixed import parser bugs, including Vue file support and path alias (@/) resolution via tsconfig.json/vite.config.js (#801, #811)
  • Improved robustness of code extraction using <result> tags instead of markdown code blocks (#801)
  • Fixed concurrency issues with a new file locking mechanism (#801)
  • Refined stop-sequence handling to prevent KeyError and improve LLM interaction (#799, #812)
  • Fixed filesystem tool errors (#816)

• Workflow & Tooling

  • Resolved import error in workflow example (#806) — by @xucailiang

— Primarily by @tastelikefeet

🤗 Contributors

Special thanks to our contributors for this release:
@wangxingjun778, @tastelikefeet, and @xucailiang.

New Contributors

• @xucailiang made their first contribution in #806

Full Changelog: v1.5.1...v1.5.2

v1.5.1

🔥 Highlights

• Add Financial Research Application based on Gradio, refer to: FinResearch DEMO
• Improve & Enhance code tool calling stability for FinResearch
• Support skill_id on ModelScope Hub for Agent Skills, refer to SKILL_README

🚀 What's Changed

Financial Research Application by @alcholiclg

• #781 Enhance code tool stability when loading Docker images
• #789 feat: improve README/code tool integration
• #787 feat/fin app: financial application skill enhancements
• #782 Support page_range, max_page_num, and max_file_size to control document ingestion

🎥 Video & Multi-modal Generation

• Enhance video generation pipeline with MLLM-based QA by @tastelikefeet
  • #784 Add multi-modal check for video generation (detect/layout fix Manim frames)
  • #786 Support T2SV images in generation pipeline

🧠 Agent Core & Memory

• Refine agent memory and planning by @suluyana
  • #758 feat(anthropic): enable thinking step for Anthropic models
  • #761 feat/refine memory: improve long/short-term memory handling

🔌 Skill & Tooling Improvements

• Expand skill & tool capabilities by @alcholiclg & @wangxingjun778
  • #785 feat: support agent skill IDs on ModelScope Hub
  • #788 feat: support subfolder paths for skill_id

📄 Documentation & Readme

• Update README language & fix display issues by @wangxingjun778
  • #779 Fix video example rendering (replace <video> tags with external links)
  • #780 Fix roadmap section
  • #783 Fix/update README (Chinese/English)

Full Changelog: v1.5.0...v1.5.1

v1.5.0

🚀 MS-Agent v1.5.0 Release Notes

✨ New Features

• FinResearch--Financial Research Multi-Agent Workflow
  Introduced a dedicated multi-agent system for financial research tasks, supporting data collection from Akshare and Baostock. As well as leveraging Web Search to retrieve online public sentiment data, enabling integrated analysis of structured and unstructured financial data, and ultimately generating rich, multimodal financial analysis reports with both text and visualizations . Refer to: FinResearch
  (#766, #774, #776)

📚 Documentation Improvements

• Modernized Documentation Theme & Structure
  Migrated documentation to Sphinx + sphinxawesome_theme, with full support for English and Chinese versions. Improved navigation, maintainability, and build workflow.
  (#773, #775)

• Updated Release News & Skill Documentation
  Added release announcement for v1.5.0, enhanced FinResearch README with usage examples, and updated skill documentation for contributors.
  (#772, #774, #775)

🔧 Fixes & Refinements

• DeepResearch Logic Refinement
  Fixed and refined behaviors in the DeepResearch agent to improve robustness and output quality.
  (#742)

• Resource Limit Standardization
  Unified code_executor default resource limits across configurations:
  cpu_limit = 1.0, memory_limit = "1g" for better resource management.
  (#774)

• Import Clean-up & Lazy Loading
  Refactored financial tool imports to be lazy and conditional; internal types (e.g., FinancialDataFetcher) no longer exposed in tools/__init__.py.
  (#776)

What's Changed

• Fix/refine deepresearch by @alcholiclg in #742
• Update skill readme for skill seekers by @wangxingjun778 in #772
• Feat/financial research by @alcholiclg in #766
• Feat/refactor doc theme by @wangxingjun778 in #773
• Fix/update fin readme by @alcholiclg in #774
• Update docs by @wangxingjun778 in #775
• fix imports by @alcholiclg in #776

Full Changelog: v1.4.0...v1.5.0

v1.4.0

✅ Highlights:

• ✨ Added Agent Skills support, fully compatible with the Anthropic Skills protocol
• 📘 Add MS-Agent documentation
• 🛡️ Add Sandbox using ms-enclave

---

✨ Features

• Agent Skills System

  • Add agent skills framework, fully compatible with Anthropic Skills protocol
    (#760) — @wangxingjun778
  • Add documentation for agent skills
    (#768) — @wangxingjun778

• Code Tool

  • Introduce code tool for enhanced code generation & execution
    (#756) — @alcholiclg

• Video Generation

  • Add GenerateVideo feature for multimodal content creation
    (#759) — @xiaoxianhjy

• DeepResearch Enhancement

  • Add DeepResearch documentation
    (#752) — @alcholiclg
  • Refine docs & add English version
    (#767) — @wangxingjun778

• Infrastructure

  • Add workflow imports for better composability
    (#743) — @wangxingjun778

---

🐞 Bug Fixes

• Fix continue generation logic (LLM streaming & artifact handling)
  (#736) — @tastelikefeet
• Fix mem0 integration & code prompt
  (#741) — @tastelikefeet
• Fix Claude 4.5 compatibility errors
  (#750) — @tastelikefeet
• Fix general agent bugs
  (#751) — @tastelikefeet
• Fix memory-related bugs
  (#753) — @vinci-grape
• Fix bugs in Anthropic API usage
  (#757) — @suluyana
• Fix environb import on Windows
  (#762) — @wangxingjun778

---

📘 Documentation

• Fix & refine deep_research README
  (#739, #740) — @alcholiclg
• Update docs index & fix formatting
  (#769, #770) — @wangxingjun778

---

🧪 Testing & CI

• Fix CI test case failures
  (#745) — @suluyana
• Fix flaky test cases (remove time.sleep, add server warm-up)
  (#747) — @suluyana

---

🧹 Maintenance & Refactor

• Add docs, refactor code & fix bugs in bulk
  (#744) — @tastelikefeet
• Refactor core agent codebase for better modularity
  (#749) — @tastelikefeet
• Set default version for main branch
  (#763) — @wangxingjun778
• Apply linting fixes (pre-commit compliance)
  (#764) — @tastelikefeet

---

🙌 Contributors

Thanks to all contributors for this release:
@tastelikefeet, @alcholiclg, @wangxingjun778, @suluyana, @vinci-grape, @xiaoxianhjy

v1.3.0

What's Changed

✨ Features

• Feat: integrate Mem0 memory management into code_scratch by @vinci-grape in #730
• Feat: anthropic llm by @suluyana in #722
• Feat: general research pipeline by @alcholiclg in #731
• Add memory module by @suluyana in #721
• Add: Ray support for accelerating document information extraction by @alcholiclg in #729

🛠 Fixes & Improvements

• Fix: deep research workflow by @wangxingjun778 in #720
• Fix: MCP type matching by @suluyana in #727
• Fix: some compatible bugs by @tastelikefeet in #732

🚀 Enhancements

• flexible MCPClient initialization by @suluyana in #705
• update: runner by @suluyana in #728

📄 Documentation

• update README for news by @wangxingjun778 in #718
• update news by @wangxingjun778 in #719
• update doc research README by @wangxingjun778 in #734
• update: README for agent chat by @wangxingjun778 in #737
• update readme for code scratch by @vinci-grape in #735
• update readme for memory by @suluyana in #733

New Contributors

• @vinci-grape made their first contribution in #730

Full Changelog: v1.2.0...v1.3.0

v1.2.0

English Version

Key Features

• DocResearch workflow now supports txt file processing, #706
• DocResearch adds functionality to push to GitHub, ModelScope, and HuggingFace, #708, #709, #712
• DocResearch adds Markdown report export feature, supporting export to PDF, PPTX, DOCX, and HTML formats, #713

What's Changed

• Support for txt processing and file preprocessing by @alcholiclg in #706
• Fix/table mismatch by @alcholiclg in #703
• Feat/push to GitHub by @wangxingjun778 in #708
• Feat/push to modelscope hub by @wangxingjun778 in #709
• Update doc research readme by @wangxingjun778 in #710
• feat: Add push to huggingface by @wangxingjun778 in #712
• fix: mcp client raise error by @suluyana in #711
• Feat/add markdown converter by @wangxingjun778 in #713
• Fix/push to ms by @wangxingjun778 in #714
• fix input_callback by @tastelikefeet in #716
• Fix newspaper run case and update comments by @wangxingjun778 in #715
• Update doc research readme by @wangxingjun778 in #717

中文版本

关键特性

• DocResearch workflow支持 txt 文件处理， #706
• DocResearch新增推送至 GitHub、ModelScope、HuggingFace功能，#708, #709 , #712
• DocResearch新增 Markdown报告导出功能，支持导出为PDF、PPTX、DOCX、HTML格式，#713

变更内容

• 支持 txt 文件处理及文件预处理，由 @alcholiclg 在 #706 中贡献
• 修复DocResearch workflow中表格不匹配问题，由 @alcholiclg 在 #703 中贡献
• 新增推送至 GitHub 功能，由 @wangxingjun778 在 #708 中贡献
• 新增推送至 ModelScope Hub 功能，由 @wangxingjun778 在 #709 中贡献
• 更新DocResearch README，由 @wangxingjun778 在 #710 中贡献
• feat: 新增推送至 Hugging Face 功能，由 @wangxingjun778 在 #712 中贡献
• 修复：MCP 客户端报错问题，由 @suluyana 在 #711 中贡献
• 新增 Markdown 转换器，由 @wangxingjun778 在 #713 中贡献
• 修复推送至 ModelScope 的问题，由 @wangxingjun778 在 #714 中贡献
• 修复 input_callback 问题，由 @tastelikefeet 在 #716 中贡献
• 修复cli.sh运行示例并更新注释，由 @wangxingjun778 在 #715 中贡献
• 更新DocResearch README，由 @wangxingjun778 在 #717 中贡献

Full Changelog: v1.1.2...v1.2.0

v1.1.2

What's Changed

• Update readme for doc research by @wangxingjun778 in #700
• update readme by @wangxingjun778 in #702
• refine mcp connect&call error by @suluyana in #704

Full Changelog: v1.1.1...v1.1.2