NotebookLM Podcast Orchestrator

This is an autonomous orchestrator that automatically plans and generates NotebookLM Audio Overviews (podcasts) using Gemini and the NotebookLM Python SDK.

Features

• Deep Research: Leverages Gemini Pro to automatically generate in-depth research documents based on your topics.
• Automated Audio Generation: Pushes research directly to NotebookLM and starts synthesis using the notebooklm-py SDK.
• Background Tracking: Start a generation, detach your terminal, and resume tracking later.
• Automatic MP3 Downloads: Automatically downloads generated audio directly to podcast_audio/.
• Agent/MCP Ready: Exposes an MCP (Model Context Protocol) server so AI agents can autonomously generate podcasts.

Prerequisites

1. Google Chrome: You must have Google Chrome installed and be logged into notebooklm.google.com. The bot uses your local Chrome cookies to authenticate.
2. Gemini API Key: You need a Gemini API Key to run the research phases.

Setup

# Create a virtual environment (optional but recommended)
python3 -m venv venv
source venv/bin/activate

# Install dependencies
pip install -r requirements.txt

Interactive CLI Usage

Run the interactive orchestrator:

python podcast_cli.py

From the interactive menu, you can generate a new Podcast Plan from a topic or execute an existing JSON plan.

If you start a generation and want to close the terminal, you can! The generation happens in the cloud. You can reconnect and check on the status later by running:

python podcast_cli.py -c

Autonomous Agent (MCP) Usage

You can run the MCP server to expose the podcast generation capabilities to an AI agent:

python mcp_server.py

Agents will be able to use the generate_podcast_plan and execute_podcast_plan tools to run the pipeline autonomously.