Onyx is an open-source Gen-AI platform designed to connect with an organization's internal knowledge base, enabling team-wide AI-powered search and chat. It targets businesses seeking to leverage their proprietary data with LLMs, offering a unified interface for accessing and interacting with information across various applications.

How It Works

Onyx integrates with over 40 enterprise connectors (e.g., Google Drive, Slack, Salesforce) to index and synchronize data, maintaining access controls. It utilizes custom deep learning models for indexing and inference, aiming to improve retrieval and learning from user feedback. Users can build custom AI agents with specific prompts and knowledge, and the platform supports plugging into any LLM of choice.

Quick Start & Requirements

• Install/Run: docker compose up for local deployment. Onyx Cloud offers a free trial.
• Prerequisites: Docker.
• Resources: Local deployment is possible on a laptop; scalable options for Kubernetes are available.
• Docs: https://docs.onyx.app/quickstart

Highlighted Details

• Supports over 40 connectors including Google Drive, Slack, Confluence, Salesforce, and Microsoft SharePoint.
• Offers flexible security features like SSO (OIDC/SAML/OAuth2) and RBAC.
• Enables creation of custom AI agents with unique prompts and actions.
• Scalable deployment options tested for tens of thousands of users and millions of documents.

Maintenance & Community

The project is actively developed, with a roadmap including new retrieval methods (StructRAG, LightGraphRAG), personalized search, and expert suggestion features. Community engagement details are not explicitly provided in the README.

Licensing & Compatibility

Onyx Community Edition (CE) is licensed under the MIT Expat license, permitting commercial use and linking. An Enterprise Edition (EE) with additional features is available for larger organizations, requiring contact for self-hosting or using Onyx Cloud.

Limitations & Caveats

The README does not detail specific limitations of the Community Edition or mention any alpha/beta status. Information on specific LLM compatibility beyond "any LLM of your choice" is not provided.