About this project

Onyx is an open source AI platform that serves as an application layer for LLMs, offering a feature-rich chat interface that can be self-hosted. It supports all major LLM providers, including self-hosted options (Ollama, LiteLLM, vLLM) and proprietary services (Anthropic, OpenAI, Gemini). Key capabilities include: - Agentic RAG: Hybrid index combined with AI agents for information retrieval. - Deep Research: Multi-step research flow for generating in-depth reports. - Custom Agents: Build agents with custom instructions, knowledge, and actions. - Web Search: Integration with Serper, Google PSE, Brave, SearXNG, plus an in-house web crawler and Firecrawl/Exa support. - Artifacts: Generate downloadable documents, graphics, and files. - Actions & MCP: Connect agents to external applications with flexible authentication. - Code Execution: Run code in a sandbox for data analysis and file modification. - Voice Mode: Text-to-speech and speech-to-text interaction. - Image Generation: Create images from prompts. Deployment options include Onyx Lite (a lightweight Chat UI requiring under 1GB memory) and Standard Onyx (full feature set with vector/keyword indexing, background workers, model inference servers, Redis caching, and MinIO blob storage). Supported deployment methods cover Docker, Kubernetes, Helm/Terraform, and major cloud providers. Enterprise features include SSO (Google OAuth, OIDC, SAML), SCIM user provisioning, role-based access control, usage analytics, query history auditing, custom code execution for PII removal, and whitelabeling. Onyx offers a Community Edition under MIT license and an Enterprise Edition with additional features for larger organizations.