Chat with your database, but keep the rows off the internet

What it does

DataLine is a local-first web tool that connects to databases and files—Postgres, Snowflake, MySQL, SQLite, CSV, Excel, and others—and lets you ask questions in plain English. It generates and executes SQL, renders charts and tables, and lets you tweak the generated queries before re-running them. The README claims it hides your actual data from the LLM by default while still using it to draft queries, and everything stores locally.

The interesting bit

The privacy architecture is the unusual part: it attempts to let an LLM write SQL for you without ever seeing the underlying rows, keeping execution and storage strictly local. That’s a harder engineering problem than the usual “upload your CSV to ChatGPT” flow, and the project treats it as the default rather than an afterthought.

Key highlights

• Runs fully offline with local storage; native binaries and a Docker image are provided.
• Connects to Postgres, MySQL, Snowflake, SQLite, CSV, Excel, SAS files, and more.
• Generates SQL and charts from natural language, with support for editing and re-running queries.
• Optional single-user basic auth for self-hosted Docker deployments.
• Open source, though the team is actively looking for maintainers.

Caveats

• The project is explicitly looking for maintainers, which suggests the bus factor is low.
• Excel imports are fragile: automatic header detection can misfire on sheets with logos or padding, and one failed sheet kills the entire import.
• Roadmap features like dashboards, knowledge-base RAG, and advanced chart types remain unchecked.

Verdict

Worth a spin for privacy-conscious developers or small teams who need ad-hoc querying without cloud lock-in. Look elsewhere if you need multi-user collaboration, mature BI dashboards, or bulletproof Excel ingestion.