Apppublic

vikesh99/ai-file-converter

sourceHugging Faceupdated 4mo agoView on Hugging Face

0likes

App README

🤖 AI File Converter

A ChatGPT-like file conversion web app that runs 100% locally. Upload a file, describe what you want in plain English, and download the converted result.

Privacy-first: No files are ever sent to any external API or cloud service.

✨ Features

—🧠 AI intent detection — TinyLlama via Ollama understands natural language commands
—📄 Multi-format support — PDF, DOCX, PPTX, XLSX, TXT, CSV, ODT, RTF
—🔒 100% local — runs entirely on your machine
—💬 Chat UI — ChatGPT-like interface with drag & drop upload
—🧹 Auto cleanup — uploaded and converted files deleted after download

📦 Supported Conversions

Input	Outputs
PDF	DOCX, TXT, PPTX
DOCX / DOC	PDF, TXT, ODT
PPTX / PPT	PDF
XLSX / XLS	PDF, CSV
TXT	PDF, DOCX
RTF	PDF
ODT	PDF

⚙️ System Requirements

Tool	Version	Status
Python	3.10+	Required
Ollama	0.23+	✅ Installed
TinyLlama	latest	✅ Pulled
LibreOffice	7.x+	✅ Installed at `C:\Program Files\LibreOffice`
Pandoc	3.x+	✅ Installed

🚀 Quick Start

1. Create & activate virtual environment

powershell

cd ai-file-converter
python -m venv venv
venv\Scripts\activate

2. Install Python dependencies

powershell

pip install -r backend/requirements.txt

3. Start Ollama (if not already running)

powershell

ollama serve

4. Start the FastAPI backend

powershell

cd backend
uvicorn main:app --reload --host 0.0.0.0 --port 8000

5. Open the frontend

Open frontend/index.html in your browser, or serve it:

powershell

cd frontend
python -m http.server 3000
# Then open http://localhost:3000

🗂️ Project Structure

ai-file-converter/
├── backend/
│   ├── main.py              # FastAPI app (endpoints)
│   ├── intent_detector.py   # TinyLlama + rule-based intent parser
│   ├── converter.py         # Conversion routing engine
│   ├── utils.py             # File helpers
│   └── requirements.txt
├── frontend/
│   ├── index.html           # Chat UI
│   ├── app.js               # Frontend logic
│   └── style.css            # Premium dark styling
├── uploads/                 # Temp (auto-cleaned)
├── outputs/                 # Temp (auto-cleaned)
├── .env                     # Config (paths, limits)
└── .gitignore

💬 Example Commands

You type	What happens
`convert this to pdf`	→ PDF output
`i need a docx version`	→ DOCX output
`give me a text file`	→ TXT output
`turn it into a spreadsheet`	→ XLSX output
`make it a presentation`	→ PPTX output

⚠️ Known Limitations

—PDF → PPTX may lose complex formatting
—Scanned PDFs (image-only) cannot be converted to TXT (no OCR)
—LibreOffice conversions may take 2–5 seconds

🔐 Privacy

—No files sent to external APIs (unless configured to use cloud LLMs)
—All processing is local
—Files auto-deleted after each conversion
—No file contents are logged

☁️ Production Hosting (Under 1 GB)

To host this application publicly on cloud platforms (e.g., Render, Railway, DigitalOcean, AWS) with a minimal storage footprint:

1.Enable Cloud LLM API (Reduces size by 600MB+ & RAM by 1GB+) Add your API key to the environment variables (.env file or cloud dashboard):
2.GEMINI_API_KEY: Get a free key from Google AI Studio.
3.GROQ_API_KEY: Get a free key from Groq Console.

If either key is set, the backend will bypass the local Ollama process and use the cloud API for extremely fast (0.1s) and free intent detection.

1.Docker Deployment Use the included Dockerfile to build and deploy the container. It is optimized to install only headless, GUI-less packages for LibreOffice and Pandoc, keeping the image size around ~500 MB total.

To run locally or build using Docker:

bash

   docker build -t ai-file-converter .
   docker run -p 8000:8000 --env GEMINI_API_KEY="your_api_key" ai-file-converter

Built with FastAPI · TinyLlama/Gemini · LibreOffice · Pandoc · pdf2docx