Overview:
Onyx is an open-source AI platform that provides an application layer for large language models (LLMs). It enables LLMs with advanced capabilities including retrieval-augmented generation (RAG), web search, code execution, file creation, and deep research. Onyx is designed to be self-hosted by individuals or teams, with deployment options ranging from a lightweight Chat UI (Lite mode) for quick testing to a full-featured stack for serious users and larger organizations. It includes over 50 indexing-based connectors for integrating with external applications.
Core Features:
Agentic RAG: Hybrid indexing combined with AI agents for information retrieval to deliver search and answer quality.
Deep Research: Generates in-depth reports using a multi-step research workflow.
Custom Agents: Build AI Agents with unique instructions, knowledge, and actions.
Actions & MCP: Allows agents to interact with external applications, with flexible authentication options.
Code Execution: Executes code in a sandboxed environment for data analysis, graph rendering, or file modification.
Artifacts: Generates downloadable documents, graphics, and other file types.
Use Cases:
Developers building AI applications: Integrating LLMs with custom agents, web search, and code execution into existing workflows.
Teams deploying collaborative AI chatbots: Sharing chats and agents across an organization with SSO, RBAC, and query history.
Self-hosters evaluating LLM capabilities: Testing a lightweight Chat UI (Lite mode) or deploying a full RAG and indexing pipeline locally.
Researchers conducting multi-step analysis: Using the Deep Research feature to generate comprehensive reports from aggregated sources.
Why It Matters:
Onyx offers a self-hosted AI platform that supports both proprietary and self-hosted LLM providers, including Ollama, LiteLLM, vLLM, Anthropic, OpenAI, and Gemini. Its two-tier deployment model (Lite vs. Standard) lets users start with minimal resources (under 1GB memory) and scale up to a full RAG stack with vector indexes, job queues, and caching. The MIT-licensed Community Edition covers core Chat, RAG, Agents, and Actions features, while the Enterprise Edition adds SSO, RBAC, analytics, and whitelabeling for larger organizations.
