What it does

This is a small Spring Boot service that scrapes unanswered questions from a Chinese knowledge-planet (知识星球) community, forwards them to the ChatGPT API, and posts the generated replies back automatically. It was built by a tech blogger to handle repetitive tech questions from his 400,000 followers without drowning in manual responses.

The interesting bit

The project is deliberately tiny—“麻雀虽小，五脏俱全” (small sparrow, fully equipped)—but walks through an unusually complete pipeline: DDD-layered architecture, scheduled jobs, Docker packaging, and even proxy handling for OpenAI’s China-blocked API. Each chapter has a matching Git branch, so you can check out the code at every step. The author also recorded free Bilibili videos line-by-lining the implementation.

Key highlights

• DDD architecture with domain services separating the scraper, ChatGPT client, and job orchestrator
• Full Docker deployment pipeline included, with branch 23_xfg_docker for the final stage
• Proxy support for api.openai.com via open.aiproxy.xyz, plus fallback to ChatGLM SDK
• 7 video chapters with matching code branches, from project init to multi-task configuration
• Apache 2.0 licensed, but explicitly bans resale by training bootcamp grifters

Caveats

• The README is heavily promotional for the author’s paid community and WeChat; technical documentation is sparse
• The “model training” section at the bottom is just a stub with Python/TensorFlow install notes—no actual training code is present
• Requires your own OpenAI API key and a 知识星球 account to function

Verdict

Worth a look if you’re a Java beginner who learns by following along with video courses and wants to see how DDD, scheduling, and Docker fit together in a real-ish project. Skip it if you need production-grade reliability or dislike tutorial projects wrapped in influencer marketing.