AutoData — your local AI data analyst. Upload a CSV or Excel file and get instant data profiling, quality checks, guided cleaning, interactive visualizations, natural-language Q&A, AI insights, export, and a downloadable PDF report — all on your own machine.

---

📑 Table of Contents

• Screenshots
• Why AutoData
• Features
• Quick Start
• Architecture
• Deploy to Render
• Tests
• Contributing
• Privacy Note

---

📸 Screenshots

Landing page with live dashboard preview

Feature highlights and workflow

Dataset library with search, filters, and sample data

🚀 Why AutoData

Most data-analysis tools are either black-box SaaS (your data leaves your machine) or require you to string together a dozen notebooks and scripts. AutoData sits in the middle: a self-contained, run-anywhere pipeline that takes a raw file and walks it through a complete analysis workflow — automatically, with no code, and with every number grounded in your actual data.

✨ Features

• 📥 Upload anything — CSV / TSV / Excel (.xlsx & .xls), up to 50 MB, with automatic encoding & delimiter sniffing and async background jobs for large files
• 🔍 Auto-profiling — column type inference, distributions, summary statistics, semantic hints, and PII / sensitive-column detection
• 🩺 Data quality — missing values, duplicates, outliers, type anomalies, constant/empty columns, and skew, rolled into a 0–100 quality score
• 🧹 Cleaning studio — guided column operations (fill missing, convert numeric, trim, lowercase, parse dates, rename, drop) plus one-click quick fixes; every step is tracked and fully undoable
• 📊 Visualizations — auto-generated histograms, time series, bar/pie breakdowns, scatter plots, correlation heatmaps, plus an advanced library: box plots, violin plots, Q-Q plots, distributions, parallel coordinates, seasonal decomposition, treemaps, sunbursts, radar, bubble and pair plots
• 🗂️ Dataset library — every upload is saved locally with search, sort, favorites, file-type filters, rename, duplicate, delete, and an AI-generated one-line summary; three curated sample datasets for instant exploration
• 💬 AI Analyst chat — ask questions in plain English; answers are grounded in the actual dataset (never invented)
• 🧠 AI insights — deterministic pattern detection (correlations, trends, top performers, outliers), each linked to its chart evidence
• 📤 Export & report — download the cleaned dataset as CSV or XLSX, or generate a shareable report in Markdown, HTML, or PDF

🛠️ Quick start

Requirements: Python 3.11+, Node 18+, or Docker.

Using Docker Compose (Recommended)

docker-compose up --build

Open http://localhost:3000 in your browser.

Using Make (Local Development)

make setup
make dev

Open http://localhost:3000 in your browser.

Environment Variables

Copy backend/.env.example to backend/.env and configure your credentials if you want to unlock AI capabilities.

Variable	Description	Default / Example
USER_LLM_API_KEY	Your API key for the LLM provider.	sk-...
USER_LLM_BASE_URL	The base URL for the OpenAI-compatible API.	https://api.deepseek.com/v1
USER_LLM_MODEL	The specific model to use for AI Q&A.	deepseek-chat
AUTODATA_DATA_DIR	(Optional) Path for durable storage.	/opt/data (on Render)

Note

No API Key? No problem. AutoData will gracefully fall back to local rule-based mode. The AI Analyst will still answer questions and generate insights completely locally using deterministic statistical rules.

🏗️ Architecture

graph TD
    %% Define Styles
    classDef frontend fill:#000000,stroke:#fff,stroke-width:2px,color:#fff;
    classDef backend fill:#009688,stroke:#fff,stroke-width:2px,color:#fff;
    classDef storage fill:#003B57,stroke:#fff,stroke-width:2px,color:#fff;
    classDef ai fill:#8b5cf6,stroke:#fff,stroke-width:2px,color:#fff;

    %% Nodes
    User(("👤 User"))
    NextJS["⚛️ Next.js Frontend (React)"]:::frontend
    FastAPI["⚡ FastAPI Backend (Python)"]:::backend
    SQLite[("🗄️ SQLite Database")]:::storage
    DataEngine["⚙️ Data Engine (pandas/numpy)"]:::backend
    AIEngine["🤖 AI Analyst Engine"]:::ai
    LLM["☁️ LLM Provider (Optional)"]:::ai

    %% Connections
    User -- "Upload CSV/Excel\nInteract with UI" --> NextJS
    NextJS -- "REST API Proxy (/api/*)" --> FastAPI
    
    FastAPI -- "Manage Sessions\nStore History" --> SQLite
    FastAPI -- "Process Data" --> DataEngine
    FastAPI -- "Natural Language Q&A" --> AIEngine
    
    DataEngine -- "Load, Profile, Clean" --> DataEngine
    
    AIEngine -. "API Call" .-> LLM
    AIEngine -- "Fallback" --> DataEngine

    %% Subgraphs for organization
    subgraph "Local Environment"
        NextJS
        FastAPI
        SQLite
        DataEngine
        AIEngine
    end

Loading

☁️ Deploy to Render (free)

Push this repo to GitHub, then on Render: New → Blueprint, pick the repo. The render.yaml at the repo root defines two free web services — no credit card required:

Service	Runtime	What it runs
autodata-backend	Python 3.11	FastAPI + pandas on $PORT, health check /api/health
autodata-frontend	Node 20	Next.js next start, proxies every /api/* request to the backend via BACKEND_URL

Open the frontend's *.onrender.com URL — the browser only ever talks to the frontend, which reverse-proxies /api to the backend, so no CORS setup is needed.

🧪 Tests

make test

🤝 Contributing and Vibe Coding ✨

We love contributions! Whether you're a seasoned developer, a data scientist, or someone who loves "vibe coding" with AI tools like GitHub Copilot or Cursor, you are incredibly welcome here.

Please see our Contributing Guide for details on how to get started, and please adhere to our Code of Conduct in all interactions.

How you can contribute:

• Vibe Coding 🤖: Drop this repository into Cursor, Claude, or your favorite AI IDE, and start chatting to build features! We encourage AI-assisted contributions.
• Code 💻: Found a bug? Have a feature idea? Open a PR! The Makefile and docker-compose.yml make it super easy to spin up the dev environment.
• Ideas & Feedback 💡: Open an Issue or start a Discussion. We want to hear how you use AutoData.
• Spread the word 🌟: If you like what we're building, give us a Star! It helps the project grow.

Don't worry if your code isn't perfect. We are happy to help you get your PR across the finish line!

🔒 Privacy note

Datasets are stored locally in a SQLite database under backend/app/data/ and never leave your machine. Nothing is uploaded to a cloud. When no LLM key is configured, all analysis is computed locally with deterministic rules.

---

Built with ❤️ using FastAPI & Next.js by Jayan Gupta.