Skip to content

Latest commit

 

History

128 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Qumly

Qumly logo

An AI-first SQL assistant that turns natural language and voice into safe SQL and explains the results in plain English.

Release GitHub Stars GitHub Issues Last Commit License

Self-hosted Version : click here

Installation Guide : click here

Working with databases often requires knowing SQL, even for simple questions. Qumly provides a natural-language and voice interface for querying databases without requiring users to write SQL manually.

Qumly uses the connected database schema to generate SQL, validates the query before execution, and keeps the generated SQL visible so users can understand and verify what is being executed.

For example, instead of writing a SQL query manually, you can simply ask:

What were the top 5 best-selling products last month?

Qumly generates and executes the query, then presents the results in a clear, human-readable format.

Qumly

Core Capabilities

  • Natural Language to SQL
  • Voice-Based Queries
  • MySQL & PostgreSQL Support
  • Database Schema Inspection
  • Human-Readable AI Answers
  • Dynamic Result Visualizations

Why It's Different

  • Schema-Aware Query Generation
  • SQL Validation & Automatic Correction
  • Safe Read-Only Query Execution
  • Clarification Handling
  • SQL Preview & Explanation

Impact & Outcomes

  • No Manual SQL Writing
  • Faster Data Exploration
  • Reduced Querying Complexity
  • Improved Query Safety
  • Better Understanding of SQL Results
  • Accessible Database Interaction for Non-SQL Users

Architecture

Qumly Architecture

How It Works

  1. User asks a question in natural language or by voice
  2. Frontend sends the request to FastAPI
  3. Backend retrieves the database schema and conversation context
  4. LLM generates a schema-aware SQL query
  5. SQL is validated and corrected if necessary
  6. Database manager executes the validated query
  7. Results and explanations are returned to the frontend

Evaluation

Qumly was evaluated on 49 test cases using an e-commerce database schema. The evaluation covers SQL generation, database results, filtering, aggregation, joins, complex queries, ambiguity handling, and SQL safety.

Qumly evaluation results

Interaction Efficiency

Qumly achieved correct database results across all 49 test cases. One clear query required an unnecessary clarification before generating the correct SQL.

Interaction Efficiency: 48/49 (97.96%)

Tech Stack

  • React + Vite
  • FastAPI
  • Python
  • Groq
  • MySQL
  • PostgreSQL
  • Git & GitHub

Project Status

Qumly is currently in active development and is not production-ready.

Current development focuses on improving SQL generation, query validation, result explanations, database support, visualizations, and the overall user experience.

Contributing

Contributions, suggestions, and feedback are welcome.

If you find a bug or have an idea for improving Qumly, feel free to open an issue or submit a pull request.

License

Qumly is licensed under the MIT License.

See the LICENSE file for more information.

Author

Muhammad Ali Siddiqui

About

An AI-first SQL assistant that turns natural language and voice into safe SQL and explains the results in plain English.

Topics

Resources

Stars

5 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages