An open-source alternative to Claude (32k+ stars)!

In a world where large language models (LLMs) can access public knowledge, a new open-source alternative called Onyx aims to provide AI agents with contextual awareness of a team's knowledge. Onyx connects to applications and indexes knowledge from over 50 sources while safeguarding data sovereignty through self-hosted deployments. It offers advanced features like web search, sandboxes, skills, and more, creating a comprehensive representation of knowledge across all connected sources.

Compared to other search methods, Onyx provides reliable, low-latency, and low-cost context for queries, whether simple or complex. It supports major LLM providers, both self-hosted and proprietary, and allows for air-gapped, self-hosted deployment to ensure sensitive IP isn't leaked. Onyx can be deployed using Docker, Kubernetes, Helm/Terraform, and cloud providers, with guides available for various platforms.

Onyx offers two deployment options: standard for comprehensive features and lite for a lightweight AI chat UI with less resource consumption, ideal for testing or teams focusing solely on chat and agents. To try Onyx without deploying, users can visit Onyx Cloud. Onyx caters to teams of all sizes, from individuals to global enterprises, and welcomes contributions through a Contribution Guide.