How hemanthvarmakonduru/texttosqlagent works
Project type and purpose
• This is a full-stack interactive data-analysis agent built in Python.
File structure
• Top-level files:
• app.py — Streamlit entry point (frontend)
• database_setup.py — DB schema and seed logic
• requirements.txt / setup.py — dependency definitions
• config/ and src/config/ — environment and application settings
• src/
• core/text_to_sql_agent.py — agent logic (transforms NL → SQL)
• utils/database.py — DB connection & query execution
• utils/visualization.py — charting & plots
• data/ — CSV files used to initialize the database
• docs/ — user and development documentation
• tests/ — unit & integration tests validating components and DB interactions
README for additional insights
• Tech Stack: Python, Streamlit (frontend), OpenAI API (external NL model), PostgreSQL (DB), pandas/visualization libs
• Workflow described: install → set env vars → setup DB → run Streamlit
a. Main components
1. Streamlit UI (app.py)
2. Text-to-SQL Agent (core/text_to_sql_agent.py)
3. Database Layer (PostgreSQL + src/utils/database.py)
4. Visualization Module (src/utils/visualization.py)
5. External OpenAI API
6. Configuration Management (config/settings.py, .env)
b. Relationships & interactions
• User → Streamlit UI: inputs natural-language query
• UI → Agent: sends NL text for conversion
• Agent → OpenAI API: issues API call to produce SQL
• Agent → Database Layer: executes SQL and fetches results
• Results → Visualization Module: generates charts & tables
• UI ← Visualization: displays output
• Config service provides API keys and DB credentials to all modules
c. Architectural patterns & design principles
• Clear separation of concerns:
– Presentation layer (UI),
– Business logic (Agent),
– Data access layer
• Modularity: utils, core, config packages
• Single-responsibility classes/functions in utils and core
• Dependency injection of settings from config modules
d. Key technologies & frameworks
• Python 3.x
• Streamlit for frontend
• OpenAI Python SDK
• psycopg2 or SQLAlchemy (via src/utils/database.py)
• pandas / matplotlib or altair (via visualization.py)
• PostgreSQL
1. Sketch high-level layers: Presentation → Business Logic → Data
2. Populate each layer with the actual components from the file tree
5. Surround environment/config files in a boundary labeled “Configuration”