Apppublic

Nantha997/Zero-ETL

sourceHugging Faceupdated 6mo agoView on Hugging Face

0likes

App README

Zero-ETL · AI Data Transformation Engine

Type a plain-English command → AI generates and executes the data transformation instantly.

🚀 Quick Start

1. Get a Groq API Key

Go to Groq Console and create a free API key to access fast, open-source models like Llama 3.3.

2. Configure `.env`

Open .env and replace the placeholder:

GROQ_API_KEY=your_actual_key_here

3. Run

bat

run.bat

The browser will open automatically. The backend runs at http://localhost:8000.

💬 Example Commands

You can chain commands natively! Once a file is created (like a Parquet file), you can immediately run transformations on it.

Command	What it does
`load sales.csv and convert to parquet`	Reads CSV, writes `sales.parquet`
`show first 20 rows of sales.parquet`	Seamlessly reads and previews the new Parquet file
`join customers.csv with orders.csv on customer_id`	Merges two datasets
`filter sales.parquet where status is Completed`	Applies row filter to a Parquet file
`fill nulls in age column with median`	Imputes missing values
`show summary statistics`	Runs `.describe()` on the dataset
`run SQL: SELECT * FROM data.csv LIMIT 5`	DuckDB SQL query
`calculate total revenue by product_category in sales.csv`	Performs groupby and aggregation
`list all files in workspace`	Shows available files in `backend/data/`

🏗 Architecture

frontend/
├── index.html        ← Dark-mode chat UI
├── style.css         ← Glassmorphism styling
└── app.js            ← API calls, table rendering

backend/
├── main.py           ← FastAPI server (REST API)
├── requirements.txt  ← Python deps
├── data/             ← Your data files (auto-created)
└── engine/
    ├── nl_parser.py  ← Groq LLM → Python code
    ├── executor.py   ← Safe code execution
    └── file_manager.py ← File ops (upload/list/delete)

🔌 API Endpoints

Method	Endpoint	Description
`POST`	`/execute`	Run a NL command
`GET`	`/files`	List workspace files
`POST`	`/upload`	Upload a data file
`GET`	`/preview/{filename}`	Preview first N rows
`GET`	`/download/{filename}`	Download a file
`DELETE`	`/files/{filename}`	Delete a file
`GET`	`/history`	Session command history

📦 Dependencies

Package	Purpose
`fastapi` + `uvicorn`	REST API server
`groq`	Groq LLM inference (Llama-3.3)
`pandas`	DataFrame operations
`pyarrow`	Parquet read/write
`duckdb`	In-process SQL engine
`openpyxl`	Excel support

🔒 Safety Notes

—LLM-generated code runs in a controlled namespace with only safe imports.
—All file I/O is restricted to the /backend/data/ folder.
—No network calls from within generated code.