🐝 D-Bee: AI MySQL Co-pilot

D-Bee is an AI co-pilot that transforms plain English into complex executable SQL queries in seconds.

Overview

D-Bee leverages local models installed in machine using Ollama to process and execute queries efficiently. The application is built with a focus on extensibility and ease of use.

How to run the application

1. Ollama is must have to run this project: pip install ollama
2. Install the Llama 3.1 model: python -m ollama.download --model llama3.1
3. Install the required packages with pip: pip install -r requirements.txt
4. Run the application using: uvicorn server.route:app --host 0.0.0.0 --port 5656
5. App shall run in http://localhost:5656/

Features

• Natural language query processing
• Direct query execution and result display
• Context-aware schema understanding
• Modular plug-and-play architecture

Flow Diagram

            ┌──────────────────┐
            │ intent_classifier │
            └─────────┬────────┘
                      │
     ┌────────────────┴────────────────┐
     │                │                │

intent="sql_query" intent="execute" intent="chitchat" │ │ │ └────────────────┴────────────────┘ │ ┌─────────▼─────────┐ │ get_chat_history │ └─────────┬─────────┘ │ ┌────────────────┴────────────────┐ │ │ intent="sql_query"/"execute" intent="chitchat" │ │ ┌────────▼─────────┐ ┌─────▼─────┐ │ get_schema_ctx │ │ chitchat │ └────────┬─────────┘ └─────┬─────┘ │ │ ┌───────▼─────────┐ │ │ generate_sql │ │ └───────┬─────────┘ │ │ │ ┌───────▼─────────┐ │ │ validate │ │ └───────┬─────────┘ │ │ │ ┌───────▼─────────┐ │ │ execute │ │ └───────┬─────────┘ │ │ │ ▼ ▼ END END

Roadmap

Phase 1 - Core Implementation

• Basic application setup with Llama 3.1 integration
• Word-to-query execution system
• Direct result display implementation

Phase 2 - Enhanced Features

• Testing and benchmarking on different models like sqlcoder [wip]
• Improved schema context understanding
• Cache db tables and schema
• Understanding user intent to execution (sql_query" | "chitchat" | "execute" | "explain_results)
• Short term memory (remembering chat context)
• Long term memory
• Planning Mode
• Choose between installed models from ui
• user can modify the query from ui
• update/delete query gaurdrailing
• Remembering chat context
• Advanced result visualization
• Human intervention if required.
• Gaurdrail for Non sql questions shall not be executed.
• Stream response instead of waiting for full response.

Phase 3 - UI

• A user can describe the schema of the database from ui.
• Migrate to Next.js
• New Chat UI
• Store old chat history to prevent re-execution of queries.

Phase 4 - Architecture

• Plug-and-play component system
• Modular extension support
• API integration capabilities

Getting Started

[Installation and setup instructions will evolve as the project develops]

Contributing

Contributions are welcome! Please feel free to submit pull requests.

License

[License information to be added]