Language Detective Service

A FastAPI-based service to detect spoken languages in audio files using mocked connectors for OpenAI, Google Gemini, Sarvam AI, and ElevenLabs. Features a web UI for file upload and result display, with files saved to an audio_files/ directory. Designed for the DripLink BE Intern Assignment.

Project UI

Project Structure

language_detective_service/
├── connectors/
│   ├── __init__.py
│   ├── openai_connector.py
│   ├── gemini_connector.py
│   ├── sarvam_connector.py
│   └── elevenlabs_connector.py
├── coordinator.py
├── main.py
├── templates/
│   └── index.html
└── audio_files/  # Stores uploaded audio files

Features

• Web UI: Upload audio files (MP3/WAV) and specify ground truth language at http://127.0.0.1:8000/.
• API Endpoint: POST /detect/language handles file uploads and returns mocked language detection results.
• Mocked Connectors: Simulate OpenAI, Google Gemini, Sarvam AI, and ElevenLabs with fake responses (e.g., "en", "hi", "ta").
• File Storage: Uploaded files are saved to audio_files/.
• Error Handling: Robust handling of file and API errors.
• Indian Language Support: Mocked responses include Hindi ("hi") and Tamil ("ta") for bonus points.

Requirements

• Python 3.10+
• Libraries: fastapi, uvicorn, pydantic, jinja2 , python-multipart

Setup Instructions

1. Clone or Create Project Directory

git clone https://github.com/555vedant/LanguageDetection.git
cd LanguageDetection

2. Set Up Virtual Environment

python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install Dependencies

pip install fastapi uvicorn pydantic jinja2

4. Create Audio Files Directory

mkdir audio_files

5. Run the Server

uvicorn main:app --reload

Usage

Access UI

1. Open http://127.0.0.1:8000/ in a browser.
2. Upload an audio file (MP3/WAV, real or dummy—content ignored due to mocks).
3. Enter ground truth language (e.g., "en").
4. Submit to view mocked results in a table.

Test API Directly (optional)

curl -X POST http://127.0.0.1:8000/detect/language \
  -F "file=@/path/to/audio.mp3" \
  -F "ground_truth_language=en"

Notes

• Mocked Implementation: No real APIs used; all connectors return fake data (languages, timings, zero costs).