Dragon AI (dranai)

dranai (Dragon AI) is a comprehensive AI-powered application designed with a client-server architecture. It serves as a unified interface and proxy for various LLM providers, offering an Ollama-compatible API for seamless integration with existing tools.

Features

• Ollama-Compatible API: Implements key endpoints of the Ollama API, allowing it to work with clients that support Ollama.
  • /api/chat: Supports chat completions with streaming and non-streaming responses.
  • /api/tags: Lists available models from registered providers.
• LangChain Integration: Utilizes LangChain for robust LLM interactions and abstraction.
• Multimodal Support: Supports image inputs in chat requests (base64 encoded) for vision-capable models.
• Provider Management: Dynamically manages and routes requests to different LLM providers via a registry system.

Usage

Installation

npm install -g dranai

Running the Server

Start the server using the CLI command:

dranai
# or
npx dranai

The server listens on port 28700 by default.

Configuration

Configuration is handled via environment variables. You can define them in a .env file in your working directory.

| Variable | Description | Default | | :--- | :--- | :--- | | DRANAI_PORT | The port the server listens on. | 28700 | | DRANAI_OLLAMA_HOST | The URL of the backing Ollama instance. | http://localhost:11434 |

API Endpoints

The server exposes an Ollama-like API under /api/ollama/:id where :id is an identifier for the tenant or context (implementation dependent).

• Chat: POST /api/ollama/:id/api/chat
  • Standard Ollama chat request format.
  • Supports images (base64) for multimodal input.
• Tags: GET /api/ollama/:id/api/tags
  • Returns a list of available models from all registered providers.