h-codex

h-codex

A semantic code search tool for intelligent, cross-repo context retrieval.

✨ Features

• AST-Based Chunking: Intelligent code parsing using Abstract Syntax Trees for optimal chunk boundaries
• Embedding & Semantic Search: Using OpenAI's text-embedding-3-small model (support for voyage-code-3 planned)
• Vector Database: PostgreSQL with pgvector extension for efficient similarity search
• Multi-Language Support: TypeScript, JavaScript, and extensible for other languages
• Multi-Project Support: Index and search multiple projects
• MCP Integration: Seamlessly connects with AI coding assistants through Model Context Protocol

🚀 Demo

💻 Getting Started

h-codex can be integrated with AI assistants through the Model Context Protocol.

Example with Claude Desktop

Edit your claude_mcp_settings.json file:

{
  "mcpServers": {
    "h-codex": {
      "command": "npx",
      "args": ["@hpbyte/h-codex-mcp"],
      "env": {
        "LLM_API_KEY": "your_llm_api_key_here", 
        "LLM_BASE_URL": "your_llm_base_url_here (default is openai baseurl: https://api.openai.com/v1)",
        "DB_CONNECTION_STRING": "postgresql://postgres:password@localhost:5432/h-codex"
      }
    }
  }
}

🛠️ Development

Prerequisites

• Node.js (v18+)
• pnpm - Package manager
• Docker - For running PostgreSQL with pgvector
• OpenAI API key for embeddings

Getting Started

1. Clone the repository

   git clone https://github.com/hpbyte/h-codex.git
   cd h-codex
   

2. Set up environment variables

   cp packages/core/.env.example packages/core/.env
   

   Edit the .env file with your OpenAI API key and other configuration options.

3. Install dependencies

   pnpm install
   

4. Start PostgreSQL database

   cd dev && docker compose up -d
   

5. Set up the database

   pnpm run db:migrate
   

6. Start development server

   pnpm dev
   

🔧 Configuration Options

Environment Variable	Description	Default
LLM_API_KEY	LLM API key for embeddings	Required
LLM_BASE_URL	LLM Base url key for embeddings	https://api.openai.com/v1
EMBEDDING_MODEL	OpenAI model for embeddings	text-embedding-3-small
CHUNK_SIZE	Maximum chunk size in characters	1000
SEARCH_RESULTS_LIMIT	Max search results returned	10
SIMILARITY_THRESHOLD	Minimum similarity for results	0.5
DB_CONNECTION_STRING	PostgreSQL connection string	postgresql://postgres:password@localhost:5432/h-codex

🏗️ Architecture

graph TD
    subgraph "Core Package"
        subgraph "Ingestion Pipeline"
            Explorer["Explorer<br/>(file discovery)"]
            Chunker["Chunker<br/>(AST parsing & chunking)"]
            Embedder["Embedder<br/>(semantic embeddings)"]
            Indexer["Indexer<br/>(orchestration)"]

            Explorer --> Chunker
            Chunker --> Embedder
            Embedder --> Indexer
        end

        subgraph "Storage Layer"
            Repository["Repository"]
        end

        Indexer --> Repository
        Repository --> Database[(PostgreSQL Vector Database)]
    end

    subgraph "MCP Package"
        MCPServer["MCP Server"]
        CodeIndexTool["Code Index Tool"]
        CodeSearchTool["Code Search Tool"]

        MCPServer --> CodeIndexTool
        MCPServer --> CodeSearchTool
    end

    CodeIndexTool --> Indexer
    CodeSearchTool --> Repository

🗺️ Roadmap

• Support for additional embedding providers (Voyage AI)
• Enhanced language support with more tree-sitter parsers

📄 License

This project is licensed under the MIT License