EchoNotes 📝

An audio recording and transcription tool built with Private AI endpoints.

Features

• 🎤 Recording: Native cross-platform audio recording
• 📝 Transcription: Hosted Whisper API with multi-language support (hosted by Cloudera AI Inference service)
• 💬 Chat with Transcriptions: Ask questions, generate notes, and run configurable Quick Actions (OpenAI-compatible LLM hosted by Cloudera AI Inference service)
• 📁 Project Organization: Per-transcription folders with original + transcript
• ⚡ AMP-ready: One-click import to Cloudera AI

Installation

Running in Cloudera AI

This repo includes an AMP manifest (.project-metadata.yaml). To deploy via Cloudera AI Workbench:

1. In Cloudera AI, create a new AMP project from this repo (URL or uploaded archive)
2. Set these environment variables in the AMP form:
   • WHISPER_BASE_URL (required)
   • LLM_BASE_URL (required)
   • LLM_MODEL_ID (required)
   • API_KEY (optional; if not provided, the app reads /tmp/jwt and uses access_token automatically)
   • MEETINGS_DIR, TRANSCRIPTIONS_DIR (optional overrides)
3. During deployment, Step 1 (Session) runs pip install .; Step 2 (Application) starts Streamlit process
4. Launch the Application; the app will be reachable at the AMP-provided URL

Running Locally

1. Clone the repository:

   git clone <repository-url>
   cd echo_notes

2. Install dependencies:

   uv sync

3. Configure API credentials (REQUIRED):

   # Copy the example configuration file
   cp .env.example .env
   
   # Edit .env with your actual API credentials
   # All required variables must be set - see .env.example for details

   Required environment variables:

   • API_KEY: Your API key for both Whisper and LLM services
   • WHISPER_BASE_URL: Whisper transcription service endpoint
   • LLM_BASE_URL: LLM service endpoint (OpenAI-compatible)
   • LLM_MODEL_ID: LLM model identifier

4. Start Application:

# Start the application
./start_ui.sh

# Or directly with uv
uv run python echo_notes_app.py

Advanced Configuration

Quick Actions (Chat) overrides:

• QUICK_ACTIONS_FILE: Path to JSON file with an array of {label, prompt}
• QUICK_ACTIONS: JSON string for the same structure
• Defaults: Summary, Sentiment, Action Items

Example QUICK_ACTIONS JSON:

[
  {"label": "🧾 Summary", "prompt": "Write a concise summary of the transcription focusing on key points."},
  {"label": "🙂 Sentiment", "prompt": "Analyze the overall sentiment and tone."},
  {"label": "✅ Action Items", "prompt": "Extract action items with owners and suggested due dates."}
]