Search Engine Bot (Hugging Face demo)

This is a small FastAPI demo that accepts an audio upload and uses the Hugging Face Inference API to transcribe audio and translate it to English. Results are stored in database.json.

Quick start

1. Install dependencies (prefer a virtualenv):

python -m pip install -r requirements.txt

2. Set environment variables (example):

$Env:HF_API_TOKEN = "<your-hf-token>"
# optional: choose another model
$Env:HF_MODEL = "openai/whisper-small"

3. Run the server:

python -m uvicorn main:app --reload --port 8000

4. Send a file (example with curl):

curl -X POST "http://127.0.0.1:8000/talk" -F "file=@path\to\audio.wav"

Notes

• This example uses the Hugging Face Inference API (huggingface-hub). You need an API token (free-tier available on HF).
• The code attempts to store both the original transcript and an English translation in database.json.
• For a quick demo, use openai/whisper-small as the default model. You can switch models by setting HF_MODEL.