bytedance / deer-flow

项目摘要

DeerFlow is an open-source super agent harness by ByteDance that orchestrates sub-agents, memory, sandboxes, tools, and skills to handle long-horizon tasks—from deep research and coding to content creation—that can take minutes to hours.

为什么值得关注

It rocketed to #1 on GitHub Trending with 83K+ stars and 289 contributors, representing a ground-up 2.0 rewrite that introduces a comprehensive multi-agent orchestration framework with sandboxed execution, persistent memory, and an extensible skill system.

优势

• Comprehensive orchestration of sub-agents, memory, sandboxes, and skills for long-horizon tasks
• Massive community traction with 83K+ stars and 289 contributors
• Active development with a ground-up 2.0 rewrite and recent GitHub Trending #1 spot
• Multilingual documentation and official website with real demos
• Integration with ByteDance's InfoQuest search/crawling toolset and LLM Space sister project

局限性

• Version 2.0 is a ground-up rewrite sharing no code with v1, suggesting potential instability and migration challenges
• Only 1 release in the past 6 months and no Docker support, indicating maturing packaging
• No examples directory despite the project's complexity
• README leans heavily toward marketing with promotional coding plan links

使用场景

• Deep research and multi-source information synthesis
• Automated coding and software development workflows
• Long-horizon content creation and podcast generation
• Complex multi-step task automation requiring memory and tool orchestration