AIStora

AIStora is an end-to-end AI analytics platform powered by a custom in-memory data engine and a natural-language query layer. It enables users to upload CSV files, explore them using SQL-like logic, and generate visual insights simply by chatting with an AI assistant.

Built with Flask, PostgreSQL, Docker, and a lightweight custom engine, AIStora performs fast projections, filters, joins, and aggregations without relying on Pandas. Gemini AI converts natural-language questions into executable data operations.

🎥 Demo

A short walkthrough demonstrating:

CSV upload and logical database creation
Natural-language query execution
AI-generated query plans
Real-time tables and visualizations

▶️ Demo video: <img src="./demo/demo.gif" width="800" height="800"/>

Features

AI-Driven Data Exploration
Ask questions in plain English and let the system translate them into executable data operations.

Custom In-Memory Engine
A lightweight engine optimized for speed, supporting:

Column projection and row filtering
Grouping and aggregation
Inner joins and type inference
Streaming CSV parsing for large datasets

Instant Visualizations
Generate tables, counts, bar charts, and line plots in real time.

Secure Authentication
User registration, login, and session-based access control.

Logical Databases
Isolated workspaces with support for multiple CSVs per database.

Zero Pandas Dependency
All data logic is implemented natively.

Tech Stack

Backend: Flask, SQLAlchemy, Gunicorn
Database: PostgreSQL
Engine: Custom DataFrame implementation with streaming CSV parser
AI / LLM: Google Gemini
Frontend: Vanilla JavaScript, Tailwind CSS, HTML templates
Infrastructure: Docker, Docker Compose
CI/CD: GitHub Actions

Architecture

graph TD
    User[User] -->|Browser / Chat| UI[Frontend UI]
    UI -->|HTTP Requests| API[Flask Backend]
    API -->|Auth| DB[(PostgreSQL)]
    API -->|Natural Language Query| LLM[Gemini AI]
    LLM -->|Query Plan| Engine[Custom Data Engine]
    Engine -->|Read| CSV[CSV Files]
    Engine -->|Results| API

Installation & Setup

Prerequisites

Docker Desktop
Git
Google Gemini API key

1. Clone the Repository

git clone https://github.com/naga251602/AIStora.git
cd AIStora

2. Configure Environment Variables

cp .env.example .env

Add your Gemini API key and other configuration values.

3. Run with Docker

docker-compose up --build

Run in detached mode:

docker-compose up --build -d

4. Access the Application

http://localhost:5001

Stop the application:

docker-compose down

Development Workflow

Directory Structure

/engine         # Custom DataFrame and parser logic
/services       # LLM integration and chart generation
/routes         # Blueprints for auth, chat, and data operations
/templates      # Jinja2 HTML templates
/static         # JavaScript and Tailwind CSS
/tests          # Pytest suite

Running Tests

docker-compose exec web pytest

Why This Project Matters

Custom Compute Engine Demonstrates strong understanding of data structures, parsing, and query execution.

System Design Integrates AI, persistent storage, and in-memory computation in a Dockerized architecture.

Modularity Clear separation between the data engine, web layer, and AI services.

License

MIT License

AiStora

Install / Use

README