AI API Server

A Node.js/Express.js backend server that proxies requests to Ollama and provides logging and monitoring capabilities.

Features

• Proxies requests to Ollama (running locally at http://localhost:11434)
• Supports both chat and generate endpoints
• Streaming responses for real-time AI interactions
• PostgreSQL database for logging prompts and errors
• API key authentication
• Modular, maintainable codebase

Prerequisites

• Node.js 14+ and npm
• PostgreSQL 12+
• Ollama running locally (http://localhost:11434)

Setup

1. Clone the repository
2. Install dependencies:

   npm install
   

3. Create a .env file with the following variables:

   PORT=5000
   NODE_ENV=development
   API_KEY=your_api_key_here
   DB_HOST=localhost
   DB_PORT=5432
   DB_NAME=ai_db
   DB_USER=your_db_user
   DB_PASSWORD=your_db_password
   

4. Create the database:

   createdb ai_db
   

5. Run the database migrations:

   psql -d ai_db -f src/database/schema.sql
   

6. Start the server:

   npm start
   

API Endpoints

Generate/Chat

• POST /api/generate
  • Supports both chat-style (messages array) and prompt-style requests
  • Requires API key in headers (api-key or Authorization: Bearer <key>)
  • Example request:

    {
      "model": "codellama:7b",
      "messages": [
        {"role": "user", "content": "Hello, how are you?"}
      ],
      "stream": true
    }
    

Logs

• GET /api/errors/prompts

  • Returns latest 100 prompts
  • Query parameter: limit (default: 100)
  • Requires API key

• GET /api/errors/logs

  • Returns latest 100 error logs
  • Query parameter: limit (default: 100)
  • Requires API key

Project Structure

src/
├── config/         # Configuration files
├── controllers/    # Request handlers
├── database/       # Database schema and migrations
├── middleware/     # Express middleware
├── models/         # Database models
├── network/        # External service communication
├── routes/         # API routes
└── server.js       # Application entry point

Error Handling

• All errors are logged to the database
• API errors return appropriate HTTP status codes
• Streaming errors are handled gracefully

Security

• API key authentication required for all endpoints
• Environment variables for sensitive data
• Input validation and sanitization
• Secure headers with Helmet (TODO)

Development

• Use npm run dev for development with auto-reload
• Use npm test to run tests (TODO)
• Follow the existing code style and structure

License

MIT