An intelligent, conversational data agent capable of querying a local SQL database to answer business questions about customers, orders, and deliveries.

Built with Python, LangGraph, Groq (Llama 3), and Streamlit.

• Natural Language to SQL: Translates plain English questions into secure SQL queries.
• Dynamic Schema Awareness: The agent reads the database structure in real-time instantly adapting if the tables change.
• Read-Only Safety: Uses strictly read-only connections to prevent accidental data modification.
• Dual Interface:
  • Web UI: A user-friendly chat interface built with Streamlit.
  • CLI: A terminal-based mode for developer testing.
• Memory: Retains conversation context to answer follow-up questions.

• LLM: Llama 3.3 (via Groq API)

• Orchestration: LangGraph & LangChain

• Database: SQLite

• Frontend: Streamlit

• Environment: Python 3.9+

git clone <your-repo-url>
cd Conversational-Text-to-SQL-Agent

• For MacOS/Linux.

python -m venv venv
source venv/bin/activate

pip install -r requirements.txt

• Rename .env.local -> .env.
• Fill with your API Keys as requested in the .env file.

python3 ingest.py

• For the app:

streamlit run app.py

• For the CLI:

python3 main.py

.
├── agent/                  # The Agent
│   ├── graph.py            # LangGraph workflow definition
│   ├── prompt.py           # System prompts and rules
│   ├── state.py            # AgentState definition
│   └── tools.py            # SQL execution tools
├── data/                   # The data
│   ├── data.csv
│   └── ...
├── database/               # Data storage
│   └── database.sqlite     # The SQLite file
├── .env                    # API Keys (GitIgnored)
├── app.py                  # Streamlit App
├── ingest.py               # Data loading script
├── main.py                 # CLI App
└── requirements.txt        # Python dependencies