transcriber-rag

A simple voice transcriber with RAG application using Deepgram and LangChain.

Features

• Real-time voice transcriber directly from a microphone or speaker using Deepgram.
• Transcription data is used by RAG (retrieval augmented generation) to answer user queries.
• The RAG implementation is using LangChain with Milvus, Cohere, and Google AI.
• The application user interface is created with Streamlit.

Tutorial

1. Install the dependencies.

uv pip install -e . -U

2. Run the app.

streamlit run app.py

3. Sidebar configurations.

• Input API credentials.
• Input settings.
• Select one of the feature pages:
  • Transcribe: Real-time voice transcription.
  • Question and Answer: Answering queries based on transcription.
• Optionally, click "Reset" button to clear the data.

5. Transcribe.

• Click on the "Start" button to start transcribing.
• Click on the "Stop" button to stop transcribing.

6. Question and Answer.

• Input your query in the text box.
• Click on the "Submit" button.
• Click citation button to view citation details.

Demonstration