VidChat

VidChat is a small Streamlit app for asking questions about a YouTube video. It downloads/transcribes the video audio with OpenAI Whisper, stores transcript chunks in a FAISS vector index, and uses Gemini to answer from the retrieved transcript context.

Features

• Import a YouTube video by URL.
• Transcribe video audio.
• Ask questions about the transcript.
• Show the transcript snippets used as references.

Tech Stack

• Python
• Streamlit
• LangChain
• OpenAI Whisper and embeddings
• Google Gemini
• FAISS

Setup

1. Install dependencies:

pip install -r requirements.txt

2. Create a local .env file from the example:

cp .env.example .env

3. Add your API keys:

GOOGLE_KEY=your_google_api_key
OPENAI_KEY=your_openai_api_key

4. Run the app:

streamlit run app.py

Notes

• Real .env files are ignored by Git.
• Downloaded audio files are saved in audios/ and ignored by Git.
• The search tab is present in the UI, but the current app flow only handles direct YouTube URL import.

Security

Do not commit API keys. This repository previously tracked .env; if real keys were committed, rotate those keys in the OpenAI and Google dashboards.