WrenAI
Canner/WrenAI · Python · Apache-2.0
Ask your database questions in plain English; it clarifies, then writes the SQL.
WrenAI sits on top of your warehouse, builds a semantic model of the schema, asks follow-up questions to disambiguate, and returns SQL plus a chart you can trust.
Score
Setup at a glance
docker compose up -d Deep dive
WrenAI is an open-source GenBI stack: ask a question in natural language and get a governed answer plus a deployable dashboard — not just a SQL snippet.
Correctness is the design centre — schema-aware retrieval, MDL (semantic-layer) planning and dry-run validation run before anything executes, so the system can correct itself instead of confidently returning wrong numbers. Business meaning lives in Git: metric definitions, enums, approved joins and prompts are versioned like code.
It works through the agents you already run — Claude Code, Cursor, Cline, Codex, MCP — and sits on your existing warehouse and transformation stack rather than demanding a new one.