Omni is an AI Assistant and Search platform designed for the workplace, enabling employees to find information and complete tasks by connecting to various workplace applications. It offers a unified search experience, AI-powered assistance, and a self-hosted deployment model, ensuring data privacy and control for organizations.

How It Works

Omni centralizes its data management and search capabilities within PostgreSQL, leveraging ParadeDB for both BM25 full-text search and pgvector semantic search, eliminating the need for separate search or vector databases. Core backend services are built with Rust for performance-critical components like the indexer and connectors, Python for AI/LLM orchestration, and SvelteKit for the web frontend. Data source connectors operate as independent containers, facilitating diverse language use. The AI agent's code execution environment is a securely sandboxed Docker container with strict access controls, including Landlock filesystem restrictions and resource limits, ensuring safe analysis of connected data.

Quick Start & Requirements

Deployment is facilitated via Docker Compose for single-server setups or Terraform for scalable AWS/GCP deployments. Official documentation and deployment guides are available at https://docs.getomni.co. Specific hardware or software prerequisites beyond Docker and cloud provider access (for Terraform) are not detailed in the README.

Highlighted Details

• Unified Search: Integrates with Google Workspace, Slack, Confluence, Jira, and public websites, offering both BM25 and semantic search.
• AI Agent: Features a chat interface capable of tool use, document summarization, and secure, sandboxed code execution (Python/bash) for data analysis.
• Self-Hosted & Private: Runs entirely on user infrastructure, guaranteeing that no data leaves the organization's network.
• BYO LLM: Supports integration with various LLMs, including Anthropic, OpenAI, Gemini, and open-weight models via vLLM.
• Permission Inheritance: Automatically respects existing permissions from connected source systems.

Maintenance & Community

Community engagement is fostered through Discord (https://discord.gg/aNr2J3xD) and GitHub Discussions (https://github.com/getomnico/omni/discussions). Development guidelines are provided in the CONTRIBUTING.md file.

Licensing & Compatibility

The project is licensed under the Apache License 2.0, which permits commercial use and integration into proprietary applications.

Limitations & Caveats

The provided README does not explicitly detail specific limitations, unsupported platforms, or known bugs.