Anarix AI Agent Task (FastAPI + Gemini API, PostgreSQL)

This project is the Challenge Provided by the Anarix to create a AI-powered E-Commerce data assistant that:

- Understands the Natural Language Questions
- Converets them to the any Database Query (I have selected the PostgreSQL) using the LLM (gemini-2.5-flash)
- Returns the tabulated response in the terminal using the tabulate library
- Visualizes the results as charts, bars when applicable

---

Project Directory Structure

Anarix_Task/
|__ .env => (Stores API_KEY and DB_URL for secure environment config)
|__ data/
|   |__ Product-Level Ad Sales and Metrics.csv
|   |__ Product-Level Eligibility Table.csv
|   |__ Product-Level Total Sales and Metrics.csv
|
|__ db/
|   |__ connection.py => (Sets up DB connection and extracts table schemas)
|   |__ queries.py => (Executes SQL queries and generates visualizations by matplotlib)
|
|__ llm/
|   |__ agent.py => (Uses Gemini API to convert user Natural Language Question's into SQL queries)
|
|__ init_db.py => (Loads CSV files into the database as tables)
|__ requirements.py => (Script to define dependencies of the Project)
|__ TASK DESCRIPTION_GenAI Intern VIT.pdf 
|__ README.md => (Project documentation)

---

Dependencies

    FastAPI, Uvicorn (Web Server)
    SQLAlchemy (Database Access)
    pandas (Data Handling)
    Matplotlib (library for Chart/Visuvalization)
    tabulate (Terminal Table)
    python-dotenv (Environment variable loader)
    google-generatveai (Gemini SDK)

Set UP Instructions

1. Add the Environment for your Database (postgres) DB_URL and the Gemini_API_KEY

API_KEY = your_gemini_api_key
You can Collect you API_KEY from here: https://aistudio.google.com/apikey
DB_URL = postgresql://<username>:<your-db-password>@localhost:5432/<your_db>

2. Create and Activate Python Virtual Environment, Then Install Dependencies

# Create virtual environment
python -m venv <your-venv-name>

Activate on PowerShell (Windows):

.\<your-venv-name>\Scripts\activate

Activate on Linux / Git Bash:

source <your-venv-name>/Scripts/activate

Install project dependencies:

pip install -r requirements.txt

3. Load the csv files to your PostgreSQL

python init_db.py or python -m init_db

4. Run the AI Agent by Running the main.py

python main.py or python -m main
The output runs in the following server: http://127.0.0.1:8000/docs or http://localhost:8000/docs

---

Screenshots

1. Terminal Interaction

---

2. Swagger UI (/docs)

---

3. Auto-Generated Line Chart

Query Asked:

Result:

---

4. Auto-Generated Bar Chart

Query Asked:

Result: