AI 应用开发 · 智能体系统 · 全栈工程
AI Application Engineering · Agent Systems · Full-Stack Development
让 AI 从想法走向可用的产品。
Turning AI ideas into usable products.
我是一名 AI 应用开发者，关注 Agent、RAG、工作流编排与全栈产品开发。喜欢把模型、工具和数据连接起来，从后端运行时到前端交互，构建可追踪、可调试、易于使用的 AI 应用。
I'm an AI application developer focused on agents, retrieval-augmented generation (RAG), workflow orchestration, and full-stack products. I connect models, tools, and data, building AI applications with traceable execution, practical debugging, and thoughtful interfaces—from backend runtimes to the user experience.
AI Agent 执行沙箱 · An execution sandbox for AI agents
围绕 Agent 执行过程中的权限策略、执行记录与回放调试构建，提供 MCP 接入、Python SDK 和可视化控制台。
A runtime sandbox built around permission policies, execution traces, and replay debugging, with MCP integration, a Python SDK, and a visual dashboard.
Go Gin React TypeScript MCP OpenTelemetry SQLite
电商数据分析与对话助手 · E-commerce analytics and a conversational assistant
我的实现位于这个 fork 仓库，涵盖 Keepa 数据处理、商品资格与 ROI 分析、多轮 AI 对话，以及后台数据刷新。
My implementation in this fork covers Keepa data processing, product eligibility and ROI analysis, multi-turn AI conversations, and background data refresh.
Python LangGraph Vue 3 PostgreSQL Redis Celery Docker
| 方向 · Area | 技术与实践 · Technologies & practices |
|---|---|
| AI 应用 · AI applications | Agent runtimes, RAG, tool calling, MCP, workflow orchestration |
| 后端 · Backend | Python, Go, FastAPI, Gin |
| 前端与桌面端 · Frontend & desktop | TypeScript, React, Vue, Next.js, Electron |
| 数据与检索 · Data & retrieval | PostgreSQL, Elasticsearch, Redis, SQLite, hybrid search |
| 工程与可观测性 · Engineering & observability | Docker, GitHub Actions, OpenTelemetry, Langfuse |
- 可控的 Agent 执行：工具权限、沙箱边界与人工确认。
Controlled agent execution: tool permissions, sandbox boundaries, and human approval. - 可追溯的检索与生成：把回答、来源和执行过程连接起来。
Traceable retrieval and generation: connecting answers with sources and execution history. - 完整的产品体验：让后端能力通过清晰的界面进入实际使用流程。
End-to-end product experience: making backend capabilities useful through clear interfaces.
欢迎通过公开项目的 Issues 交流技术问题与改进建议。
Questions and suggestions are welcome through the issues on my public projects.