Xyne: Open-source LLM-driven search engine for Google Workspace (github.com)

Xyne is an open-source, LLM-driven search-and-answer engine designed to index and query fragmented enterprise data across Google Workspace, Atlassian, Slack, GitHub and more, positioning itself as an alternative to Glean, Gemini and Microsoft Copilot. It builds a relationship graph over documents, emails, calendars, contacts, attachments, tickets and repos so teams can ask natural-language questions (e.g., about a customer, ticket, file or person) and receive up-to-date answers with provenance. The project emphasizes privacy and security—self-hostable (laptop, on-prem, cloud), permission-aware (live enforcement of app permissions), no telemetry and no training on customer data—making it attractive for security-sensitive orgs. Technically, Xyne is model-agnostic and integrates with any LLM or cloud provider (including local setups via ollama/DeepSeek), supports multi-threaded data ingestion, and exposes connectors and a Service Account workflow for deep Google Workspace integration (beyond Drive to include Gmail, Calendar and Contacts). It’s distributed via simple Docker Compose and has AWS EC2 deployment docs, inviting contribution. For the AI/ML community this matters because Xyne provides a production-ready retrieval, memory and tooling layer—an “OrgOS”—that augments both humans and LLM agents with contextual grounding, secure retrieval, and source-aware answers, enabling safer, customizable enterprise AI workflows.