The Open Assistant API provides a self-hosted, open-source framework for building and orchestrating AI assistants, compatible with the OpenAI API interface. It targets developers and organizations seeking greater control over their LLM applications, offering enhanced model flexibility, RAG capabilities, and tool integration beyond the official OpenAI offerings.

How It Works

This API acts as a proxy and orchestration layer, routing requests to various LLM providers via One API and integrating specialized components like the R2R RAG engine. Its architecture allows for local deployment, enabling developers to leverage custom tools, internet search, and multimodal features while maintaining data privacy and control.

Quick Start & Requirements

• Install and run via Docker Compose: docker compose up -d
• Requires Docker and Docker Compose.
• Configuration involves setting OPENAI_API_KEY (supports OneAPI) and optionally BING_SUBSCRIPTION_KEY in docker-compose.yml.
• R2R RAG engine integration is recommended for enhanced RAG capabilities.
• Official OpenAI Python library (openai) is used for interaction: pip install openai.
• API Base URL: http://127.0.0.1:8086/api/v1
• Documentation: http://127.0.0.1:8086/docs

Highlighted Details

• Supports multiple LLMs through One API integration.
• Integrates R2R RAG engine for advanced retrieval-augmented generation.
• Includes built-in internet search capabilities.
• Supports multimodal inputs and streaming output.
• Offers basic user isolation via token-based authentication.

Maintenance & Community

• Active community support via Slack and Discord channels.
• Relies on and acknowledges key projects: OpenOpenAI, One API, R2R, OpenAI-Python, LangChain, OpenGPTs, TaskingAI.
• Contribution guidelines are available.

Licensing & Compatibility

• MIT License.
• Compatible with commercial use and closed-source linking.

Limitations & Caveats

Code interpreter functionality is currently under development. While RAG is supported, the preliminary implementation may require configuration with a more robust engine like R2R for optimal performance.