AgentSwarms 1.6.0 – A complete source-available AI Data platform (Lakehouse + ETL + BI) built natively for multi-agent AI
retained text
Hi Team, We just released v1.6.0 of **AgentSwarms** (**Note**: It's not a lightweight Databricks!) Most agent platforms treat data as an afterthought—you build the AI, then duct-tape your warehouse and pipelines to it. AgentSwarms flips that. It is a complete, self-hostable data platform designed from the ground up to host and govern AI swarms. The data and agent layers share one semantic layer and permission model, an AI agent asked about revenue uses the exact same definition and access rules as your finance dashboard. **The Architecture:** * **The Data Layer:** A DuckDB/zstd Parquet lakehouse, visual Python ETL pipelines, and a BI workspace. * **The AI Layer:** A visual canvas to build multi-agent workflows and AI analysts that natively query, analyze, and cite your lakehouse data. **What's New in 1.5.0:** * **100% Self-Hostable:** Removed all external platform dependencies. Run the entire stack (lakehouse, app, edge functions) locally via one Docker command. * **Granular Governance:** You can now strictly authorize exactly what a lakehouse write *reads*. * **Native React SDK:** Embed analytics and swarms directly into your apps * **Vault Integration:** Model and database credentials can now live securely in HashiCorp Vault. You can run it entirely on-prem with local models like Ollama or plug in your cloud provider keys. I'd love for this community to try breaking it and share feedback.
document fetched 2026-10-26 00:28 · canonical https://reddit.com/r/selfhosted/comments/1xezri2/agentswarms_160_a_complete_sourceavailable_ai