AI Data Analyst

Upload a CSV or connect a database, ask questions in plain English, and get tables, summaries, or charts back.

Live App | Snapshot | Architecture | Tech Stack | Run locally

AI Data Analyst is a no-code analysis app for CSV files and relational databases. It inspects the provided data source, decides whether the user needs a SQL answer or a visualization, generates the right query or code, executes it, and returns a concise result.

Product Snapshot

The app starts with a simple data-source picker. Users can upload a CSV file or provide a database URL, then move into a chat-style analysis flow.

What It Does

Capability	Description
CSV upload	Uploads CSV files and stores them in Supabase Storage.
Database input	Accepts relational database URLs and reads schema metadata through SQLAlchemy.
Natural-language analysis	Converts plain-English questions into SQL or visualization workflows.
SQL answers	Generates, validates, executes, and summarizes SQL query results.
Visual analysis	Generates Python chart code and runs it inside an E2B sandbox.
Artifact storage	Stores generated chart images in Supabase and returns public URLs.

Why Not Just Use ChatGPT?

AI Data Analyst is built for users who want the convenience of an AI analyst without handing database API keys directly to a general-purpose chat product. Secrets stay in your own deployment environment, and database access flows through this app's backend instead of being pasted into an external assistant.

It is also open source, so the trust model is inspectable: you can read the code, see how uploads, database URLs, query generation, and execution are handled, and change anything that does not match your security posture.

The product UX is purpose-built for data analysis. Instead of a generic chat upload flow, the app separates CSV/database connection, schema extraction, query answering, visualization, and artifact storage. It can also work with large CSV files, including million-row datasets, where general chat tools often hit upload or context limits.

Architecture

Tools And Technologies

Layer	Tools
Frontend	HTML, Tailwind CSS, vanilla JavaScript
API	FastAPI, Uvicorn, Pydantic settings
Agent workflow	LangGraph, OpenAI, LangChain model bindings
Data execution	DuckDB for CSV queries, SQLAlchemy for database URLs, pandas for tabular results
Visualization	E2B Code Interpreter sandbox, Python chart generation
Storage	Supabase Storage for CSV files and generated images
Deployment	Docker Compose, Caddy reverse proxy, Hostinger VPS
CI/CD	GitHub Actions
Documentation	Mermaid CLI for generated architecture diagrams

Analysis Flow

Step	What happens
1	User uploads a CSV or submits a database URL.
2	/upload stores CSV files in Supabase, while database URLs pass directly into the analysis flow.
3	/query invokes the LangGraph analyst agent with the user's question and data source.
4	The agent extracts schema/sample metadata and classifies the request as SQL or visualization.
5	SQL requests run through DuckDB or SQLAlchemy; visualization requests run generated Python code in E2B.
6	The API returns rows, summaries, and chart image URLs.

API Surface

Endpoint	Purpose
GET /health	Health check for the deployed API.
POST /upload	Upload a CSV file or pass through a database URL.
POST /query	Ask a question against a csv_url or db_url.

Example query payload:

{
  "query": "What is the total revenue by region?",
  "csv_url": "https://..."
}

Environment

Variable	Purpose
OPENAI_API_KEY	Model calls for query, code, and answer generation.
E2B_API_KEY	Sandboxed Python execution for visualizations.
SUPABASE_URL	Supabase project URL.
SUPABASE_KEY	Supabase service/API key.
CSV_BUCKET_NAME	Bucket for uploaded CSV files. Defaults to data_csv.
IMAGE_BUCKET_NAME	Bucket for generated chart images. Defaults to data_image.
DATABASE_URL	Optional local/default database URL for development.

Deployment

Production is deployed at:

https://aidataanalyst.pushpitkamboj.com

The Hostinger VPS stack runs the FastAPI app behind Caddy with automatic HTTPS. GitHub Actions are included for CI and Hostinger deployment.

License

MIT

Local Setup

Requires Python 3.11+.

git clone https://github.com/pushpitkamboj/AIDataAnalyst
cd AIDataAnalyst
python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

Create .env:

OPENAI_API_KEY=
E2B_API_KEY=
SUPABASE_URL=
SUPABASE_KEY=
CSV_BUCKET_NAME=data_csv
IMAGE_BUCKET_NAME=data_image

Run the app:

PYTHONPATH=src uvicorn api.main:app --reload --host 127.0.0.1 --port 8000

Open:

http://127.0.0.1:8000

Or run with Docker:

docker compose up --build