I Turned an Archived 23K-Star Text-to-SQL Project Into a Self-Hosted Tool That Actually Works Out…

2026-04-24 · Source: Towards AI - Medium · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Data Science & Analytics · Depth: Intermediate, short

Summary

DataChat is a new self-hosted text-to-SQL chat interface, forked from the archived Vanna.ai project, designed to provide an out-of-the-box solution for generating SQL queries from natural language. The original Vanna.ai, which garnered over 23,000 GitHub stars, was archived in March 2026, leaving a gap for self-hosting enthusiasts. DataChat addresses several critical issues found in the original codebase, including the lack of a schema explorer, manual schema refresh requirements, hardcoded database switching, complex frontend build processes, and serialization crashes with complex data types. It integrates a full schema sidebar, automatic schema refreshing, command-line database switching, and an automated frontend build. DataChat supports both cloud LLMs like Gemini (gemini-2.5-flash) and local LLMs via Ollama (mistral-small3.1:latest), allowing users to keep data entirely local. It is MIT licensed and currently supports PostgreSQL and BigQuery.

Key takeaway

For AI Engineers or MLOps teams seeking a self-hosted text-to-SQL solution, DataChat offers a significantly improved experience over the archived Vanna.ai. You should consider deploying DataChat to streamline data access for non-SQL users, benefiting from its automated setup, integrated schema explorer, and support for local LLMs, which enhances data privacy. This tool eliminates common friction points, making it easier to integrate natural language querying into your data workflows.

Key insights

DataChat transforms an archived text-to-SQL project into a functional, self-hosted tool by resolving key usability and technical issues.

Principles

• Automate setup complexities.
• Provide clear database context.
• Ensure data type compatibility.

Method

The author forked the Vanna.ai 2.0.2 codebase, implemented a schema sidebar, automated schema refreshes and frontend builds, and patched serialization issues to create DataChat.

In practice

• Use `python start.py --database analytics` for database switching.
• Set `LLM_PROVIDER=ollama` for local LLM inference.
• Delete `dist/` to force a frontend rebuild.

Topics

• Text-to-SQL
• DataChat
• Vanna.ai
• Self-Hosted Tools
• Large Language Models

Code references

• drharunyuksel/datachat

Best for: AI Engineer, MLOps Engineer, Software Engineer

Related on AIssential

• Text-to-SQL solution powered by Amazon Bedrock — Text-to-SQL solutions using Amazon Bedrock democratize data access by translating natural language questions into actionable insights.
• Swiggy Rolls Out Hermes V3: From Text-to-SQL to Conversational AI — Hermes V3 enhances text-to-SQL with vector retrieval, conversational memory, and agentic orchestration for improved accuracy and user trust.
• HuggingChat | Chat with Open Models — HuggingChat unifies access to multiple open-source LLMs through a single, free chat interface.
• Cost-efficient custom text-to-SQL using Amazon Nova Micro and Amazon Bedrock on-demand inference — Fine-tuning Amazon Nova Micro with LoRA on Amazon Bedrock provides cost-efficient, production-ready text-to-SQL for custom dialects.
• The Evolution of Chat-Based AI in the Era of Data and Intelligence - Analytics Insight — Chat-based AI democratizes advanced analytics through natural language interaction and modular, integrated platforms.
• Web Search Tool with Streaming in gpt-oss-chat — LLMs can autonomously decide to use web search and generate queries, enhancing conversational AI.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Towards AI - Medium.