Summary

DataFoundry is an enterprise AI workbench for trustworthy, auditable data analysis. It mitigates risks of AI querying production data via a unified, governed, self-hosted workspace, enabling controllable, verifiable interactive analytics for enhanced accuracy.

How It Works

This workbench embeds AI agents in a semantic, policy-aware, evidence-preserving task system. It prioritizes enterprise needs through schema-first analysis via its Data Gateway, ensuring safety with read-only queries, credential isolation, and audit trails. Complex analyses are decomposed and converged into materialized outputs, with steps persisted for replay. This focus on verifiable data workflows distinguishes it from code-centric or Q&A AI tools.

Quick Start & Requirements

Installation: clone repo, npm install, configure environment variables for LLM providers (OpenAI-compatible), auth, and APIs. Node.js/npm required. Demo data included. Production needs SMTP/reverse proxy config. Links to Quick Start, Docs, Data Sources, Roadmap provided.

Highlighted Details

• Supports 28 datasource types.
• Features enterprise semantics for consistent interpretation.
• "Safe by default": read-only queries, isolation, masking, timeouts.
• Full auditability: persisted SQL, tool calls, event streams for replayable histories.
• v0.2.0 adds branchable analysis, evidence-first follow-ups, semantic traces, reusable assets.

Maintenance & Community

Actively developed, preferring focused PRs. Community via Slack, QQ, GitHub issues/discussions. Future roadmap: unified semantic layer, autonomous analyst loops.

Licensing & Compatibility

Apache License 2.0: permits commercial use, closed-source integration. Current code, docs, smoke checks are definitive.

Limitations & Caveats

Active development means significant roadmap items (e.g., unified semantic layer, autonomous analyst loops, enterprise control plane) are not yet implemented. Production features like advanced auth/monitoring require custom design. Local dev hot-reloading is "appendix-only."