WFGY
WFGY is an open-source AI Troubleshooting Atlas for RAG, agents, and real-world AI workflows. Includes the 16-problem map, Global Debug Card, and WFGY 3.0. ⭐ Star to help more builders find this repo.
1.7k
Stars
+68
Stars/month
5
Releases (6m)
Star Growth
+10 (0.6%)
Overview
WFGY是一个开源AI故障排除地图集，专门为RAG（检索增强生成）、AI代理和实际AI工作流提供系统化的调试和问题诊断指导。该工具集包含16个问题图、全局调试卡和WFGY 3.0版本，帮助开发者快速识别和解决AI系统中的常见问题。作为一个结构化的诊断生态系统，WFGY将复杂的AI故障排除过程标准化，提供了从问题识别到解决方案的完整路径。它特别适合处理AI管道中的性能瓶颈、准确性问题和系统集成挑战，是AI开发者工具箱中不可缺少的调试伴侣。
Deep Analysis
Key Differentiator
The only open-source structured troubleshooting atlas specifically for AI/RAG/agent failures — route-first diagnosis instead of random patching
⚡ Capabilities
- • AI troubleshooting atlas for broken RAG/agent workflows
- • Route-first failure diagnosis methodology
- • 16-problem map for RAG debugging
- • Global debug card for image-first triage
- • Global fix map with cross-tool guardrails
- • TXT-based reasoning surface for frontier problems
- • Atlas router for automated failure routing
🔗 Integrations
Any LLM (TXT upload)RAG pipelinesAgent frameworks
✓ Best For
- ✓ Teams debugging broken RAG pipelines
- ✓ AI engineers diagnosing agent workflow failures
- ✓ Organizations wanting structured troubleshooting methodology
✗ Not Ideal For
- ✗ Building AI applications from scratch
- ✗ Teams wanting automated monitoring (use observability tools instead)
Languages
Markdown/TXT
Deployment
GitHub repositoryTXT pack upload to any LLM
Pricing Detail
Free: Fully open-source, MIT license
Paid: N/A
⚠ Known Limitations
- ⚠ Not a software library — methodology/knowledge base
- ⚠ Requires strong LLM to process TXT packs effectively
- ⚠ Academic/research-oriented presentation style
- ⚠ No programmatic API or SDK
Pros
- + 专门针对AI系统设计的故障排除框架，覆盖RAG、代理和工作流等核心场景
- + 开源项目拥有活跃社区支持，GitHub上已获得1684颗星的认可
- + 提供结构化的问题图和全局调试卡，将复杂的AI调试过程系统化和标准化
Cons
- - 专业性较强，需要一定的AI系统基础知识才能充分利用
- - 针对性工具，主要适用于AI相关问题，不适合通用软件调试
- - 文档和学习资料可能需要时间消化理解
Use Cases
- • RAG系统性能调优和准确性问题诊断，如检索质量差、答案不准确等问题排查
- • AI代理行为异常调试，包括决策逻辑错误、工具调用失败等问题定位
- • 复杂AI工作流故障排除，如多步骤管道中断、数据流问题和集成错误分析
Getting Started
1. 从GitHub克隆WFGY仓库到本地环境；2. 查看16个问题图了解常见AI故障类型和诊断路径；3. 根据遇到的具体问题使用全局调试卡进行系统化排查和解决