oh-my-github

The VC Analyst for Open Source Code

🚧 Project Status: Work In Progress (MVP Phase) 🚧

oh-my-github is an AI-powered system that helps developers discover high-quality open source projects worth contributing to. It acts like a Venture Capital analyst, going beyond simple star counts to analyze project activity, contribution friendliness, and code quality.

🎯 Key Goals

• Find "Alpha": Discover early-stage projects (100-5k stars) with strong growth momentum.
• Analyze Friendliness: Evaluate PR merge rates, issue response times, and maintainer engagement.
• Deep Audits: AI-driven analysis of architecture, test coverage, and onboarding difficulty.
• Save Time: Transform "mass browsing" into "precise matching".

🛠 Tech Stack

• Frontend: Next.js 15 (App Router), Bun, Vercel AI SDK, TailwindCSS + Shadcn/ui
• Backend: Next.js API Routes (Edge Runtime), LangGraph.js
• AI: DeepSeek V3
• Infrastructure: Vercel, Upstash Redis

🤖 Agent Coordinator

The Agent Coordinator is a conversational AI system that orchestrates specialized agents to handle different tasks:

Architecture

User Message → Coordinator
                   ↓
           [Intent Classifier] (DeepSeek V3)
                   ↓
         ┌─────────┼──────────┐
         ↓         ↓          ↓
   [Search Team] [Auditor] [Comparator] ...
                   ↓
           [Synthesizer]
                   ↓
        SSE Stream → Frontend

API Endpoint: /api/chat

POST Request:

{
  "message": "find React state management libraries",
  "conversationId": "optional-uuid",
  "history": []
}

Response: Server-Sent Events (SSE) stream with the following event types:

• conversation_created - New conversation initialized
• log - Agent thinking steps and progress updates
• text - Incremental text chunks of the final summary
• data - Structured data (repository lists, comparisons, etc.)
• done - Request completed with execution stats
• error - Error occurred during processing

Example SSE Events:

data: {"type":"conversation_created","conversationId":"abc-123"}

data: {"type":"log","message":"Understanding your request...","agent":"coordinator"}

data: {"type":"text","content":"I found **3 repositories** that match your search..."}

data: {"type":"data","structuredData":{"type":"repo_list","items":[...]}}

data: {"type":"done","stats":{"executionTime":8500}}

Intent Classification

The coordinator uses DeepSeek V3 to classify user messages into 5 intent types:

• search - Find repositories matching criteria
• analyze - Deep analysis of a specific repository
• compare - Side-by-side comparison of multiple repos
• chat - Conversational responses (acknowledgments, help)
• clarify - Request clarification when intent is ambiguous

Confidence Threshold: 0.7 (below this, routes to clarify)

Conversation Management

• Storage: In-memory (Map-based)
• TTL: 1 hour of inactivity
• History: Last 20 messages preserved
• Context: Last 3 messages used for intent classification

Features

• Multi-turn Conversations: Maintain context across multiple queries
• Streaming Responses: Real-time feedback via SSE
• Rate Limiting: 100 requests/hour per IP
• Context Compression: LLM-based summarization for large content
• Request Validation: Zod schema validation for all inputs
• Error Handling: Graceful degradation with fallback responses

Configuration

Environment variables:

# LLM Configuration (inherited from existing setup)
DEEPSEEK_V3_API_KEY=sk-your_key_here

Development

Run tests:

bun test
# or
npm test

Check test coverage:

bun run test:coverage

🚀 Getting Started

Prerequisites

• Node.js 18+ or Bun 1.0+
• GitHub Personal Access Token (create one here)
• DeepSeek V3 API Key (get one here)

Installation

1. Clone the repository

   git clone https://github.com/yourusername/oh-my-github.git
   cd oh-my-github

2. Install dependencies

   bun install
   # or
   npm install

3. Set up environment variables

   cp .env.local.example .env.local

   Edit .env.local and add your tokens:

   GITHUB_TOKEN=ghp_your_token_here
   DEEPSEEK_V3_API_KEY=sk-your_key_here

4. Run the development server

   bun run dev
   # or
   npm run dev

5. Open http://localhost:3000

Usage

1. Enter a repository name in the format owner/name (e.g., facebook/react)
2. Click "Analyze" or try the example button
3. Wait 20-30 seconds for the AI analysis to complete
4. View the detailed analysis report

Available Scripts

• bun run dev - Start development server
• bun run build - Build for production
• bun run start - Start production server
• bun run lint - Run ESLint

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.