NexusTodo (VibeCoding Edition)

An end-to-end, AI-assisted task system built through VibeCoding: a Go backend, a Python Agently-based agent service, and a streaming web client.

Chinese README: README_cn.md

Background / Context

This project was completed end-to-end via VibeCoding. The Golang backend was implemented by TRAE, while the remaining parts (agent service, client, and documentation) were built with VSCode + Codex. Parts of the setup were demonstrated live in a 51CTO course hosted by Maplemx. The overall build-and-optimization cycle took about six hours, with roughly four hours covered in the live session.

What’s Inside

• backend (Go): task API, persistence, device registration.
• auto_agent (Python): ReAct loop, structured outputs, SSE streaming, task tool calls.
• client (Web): chat UI with streaming bubbles and task cards.

Quick Start (2 Services + 1 Client)

1. Backend (default 8080)

   cd backend
   go run main.go

2. Auto Agent (default 15590)

   cd ..
   python -m uvicorn auto_agent.app:app --host 0.0.0.0 --port 15590

3. Client

   cd client
   npm install
   npm run dev

Full instructions: PROJECT_OVERVIEW.md

Agently in This Project

The intelligent module is built on the Agently AI application framework.

• GitHub: https://github.com/AgentEra/Agently
• Official sites: https://Agently.tech (EN), https://Agently.cn (CN)

Agently provides the core primitives used here:

• Contract-first structured output (output() + ensure_keys).
• Tool planning without vendor lock-in.
• Streaming-friendly orchestration and ReAct-style loops.

Documentation Map

• Project overview and startup: PROJECT_OVERVIEW.md
• Backend API and design: backend/docs/api-documentation.md, backend/docs/backend-design.md
• Agent design and APIs: auto_agent/docs/dev_design.md, auto_agent/docs/api-documentation.md
• Chat/LLM ops spec: auto_agent/docs/spec-dd-llm-chat-ops.md
• Client usage: client/docs/client-documentation.md
• Test reports: auto_agent/docs/test-report.md, backend/docs/test-report.md
• Launch / promo article: docs/wechat-article.md, docs/wechat-article_cn.md

VibeCoding Process & Quality Loop

A detailed write-up of guidance, self-checks, scenario design, and iterative optimization:

• English: docs/vibecoding-process.md
• Chinese: docs/vibecoding-process_cn.md

Agently + VibeCoding Learnings

We summarize the effectiveness of Agently-driven VibeCoding and the self-bootstrapping loop for production-grade agent systems:

• Deterministic, schema-first outputs to reduce ambiguity.
• Real-API integration scenarios as the primary validation gate.
• Prompt + parser updates as regression fixes, captured in docs.

Community & Contribution (Agently)

• Documentation: https://Agently.tech/docs
• GitHub: https://github.com/AgentEra/Agently
• Examples (Agently repo):
  • https://github.com/AgentEra/Agently/tree/main/examples/step_by_step/
  • https://github.com/AgentEra/Agently/tree/main/examples/step_by_step/13-auto_loop_fastapi/
• Discussions: https://github.com/AgentEra/Agently/discussions
• Issues & contributions: https://github.com/AgentEra/Agently/issues

Join the Agently WeChat Group

You can find the official WeChat group entry in either place below:

• Agently official site (https://Agently.tech): navigate to the WeChat Group / Join Us entry.
• Agently GitHub homepage (https://github.com/AgentEra/Agently): open the "WeChat Group (Join Us)" section.

Current application form from the GitHub README:

• https://doc.weixin.qq.com/forms/AIoA8gcHAFMAScAhgZQABIlW6tV3l7QQf