NotebookLM Integration

MOSAIC can create and populate Google NotebookLM notebooks directly from search results. After finding and downloading open-access PDFs, MOSAIC imports them into a new notebook — ready for AI-powered Q&A, summaries, quizzes, and audio overviews.

Unofficial integration

This feature uses notebooklm-py, an unofficial Python client for NotebookLM. Google may change internal APIs without notice. Use for personal projects and research.

Setup

1. Inject notebooklm-py into MOSAIC

--include-apps exposes the notebooklm CLI to your PATH. [browser] pulls in Playwright, which is needed for the one-time Google sign-in.

# pipx
pipx inject --include-apps mosaic-search "notebooklm-py[browser]"

# uv
uv tool inject --include-apps mosaic-search "notebooklm-py[browser]"

# pip / venv (activate your venv first)
pip install "notebooklm-py[browser]"

2. Install the Chromium browser (one-time)

Playwright is installed inside the MOSAIC tool venv but its own CLI is not exposed to PATH by pipx/uv. Download the Chromium binary by calling it directly from the venv:

# pipx
~/.local/share/pipx/venvs/mosaic-search/bin/playwright install chromium

# uv
~/.local/share/uv/tools/mosaic-search/bin/playwright install chromium

# pip / venv (playwright is on PATH inside the activated venv)
playwright install chromium

This is a ~150 MB one-time download to ~/.cache/ms-playwright/.

3. Authenticate (one-time)

notebooklm login

A Chromium window opens for Google sign-in. After logging in, press Enter in the terminal. Credentials are saved to ~/.notebooklm/ and reused automatically by every future mosaic notebook call.

TIP

After the first sign-in, Playwright and Chromium are never invoked again. You can also copy ~/.notebooklm/ to another machine to skip re-authentication.

3. Authenticate (one-time)

notebooklm login

This opens a Chromium browser window for Google sign-in. Credentials are stored in ~/.config/notebooklm/ and reused automatically on every subsequent mosaic notebook call.

If you are ever signed out, re-run notebooklm login.

Usage

From a search query

Search, download, and import in one command:

# Search all sources, download OA PDFs, create notebook
mosaic notebook create "Attention Mechanisms" \
    --query "attention is all you need" \
    --oa-only --max 10

# Queue an Audio Overview (podcast) after import
mosaic notebook create "Transformers Survey" \
    --query "transformer architecture survey 2023" \
    --oa-only --podcast

# Queue multiple artifacts at once
mosaic notebook create "CRISPR 2024" \
    --query "CRISPR gene editing 2024" \
    --oa-only --briefing --quiz --mind-map

From a local directory

If you already have PDFs downloaded, import them directly:

mosaic notebook create "My Reading List" --from-dir ~/mosaic-papers/

# Import and queue a slide deck
mosaic notebook create "My Reading List" --from-dir ~/mosaic-papers/ --slide-deck

Artifact generation flags

After the notebook is populated, MOSAIC can queue one or more AI-generated artifacts. All flags are optional and can be combined freely. Artifacts are queued immediately after the notebook is created and generated asynchronously by NotebookLM — check the notebook in a few minutes.

Flag	Artifact type
--podcast	Audio Overview (podcast)
--video	Video Overview
--briefing	Briefing Doc
--study-guide	Study Guide
--quiz	Quiz
--flashcards	Flashcards
--infographic	Infographic
--slide-deck	Slide Deck
--data-table	Data Table
--mind-map	Mind Map

Audio Overview (--podcast)

A conversational AI-generated podcast — two synthetic hosts discuss the key findings, debates, and open questions across all sources in the notebook. Typical length is 10–20 minutes. Ideal for absorbing a body of literature while commuting or away from a screen. The audio file can be downloaded from the NotebookLM Studio panel.

Best for: broad literature surveys, getting a high-level feel for a new topic.

mosaic notebook create "Diffusion Models" \
    --query "diffusion probabilistic models image generation" \
    --oa-only --podcast

Video Overview (--video)

A short video presentation — an AI avatar narrates a visual summary of the notebook sources, similar to the Audio Overview but with synchronized slides or imagery. Rendered as an MP4 that can be downloaded or shared.

Best for: visual learners; presentations where a recorded explainer is more engaging than a written summary.

mosaic notebook create "AlphaFold Review" \
    --query "alphafold protein structure 2023" \
    --oa-only --video

Briefing Doc (--briefing)

A structured prose document — typically 2–5 pages — that synthesises the most important themes, claims, and supporting evidence found across the sources. Sections cover background, key findings, points of consensus, open questions, and a short conclusion. Exported as formatted text you can copy into a report or share directly.

Best for: quick executive summaries; writing literature-review introductions; getting up to speed before a meeting.

mosaic notebook create "CRISPR Therapeutics" \
    --query "CRISPR in vivo gene therapy clinical trial" \
    --oa-only -y 2022-2025 --briefing

Study Guide (--study-guide)

A learning-oriented document organised around key concepts, definitions, and review questions. NotebookLM structures the material as an outline with subsections per concept, each accompanied by a plain-language explanation grounded in the sources. Suitable for exam preparation or onboarding new team members.

Best for: teaching a topic, onboarding, systematic self-study.

mosaic notebook create "Reinforcement Learning Basics" \
    --query "reinforcement learning policy gradient survey" \
    --oa-only --study-guide

Quiz (--quiz)

A set of multiple-choice and short-answer questions drawn directly from the notebook sources, complete with correct answers and source citations. Difficulty and quantity are chosen automatically by NotebookLM based on the content. Questions can be used as-is for self-assessment or adapted into a formal assessment.

Best for: self-testing comprehension; generating questions for a journal club or course assignment.

mosaic notebook create "Attention Mechanisms" \
    --query "self-attention transformer architecture" \
    --oa-only --quiz

Flashcards (--flashcards)

A deck of term/definition or question/answer cards extracted from the sources. Each card is concise — front side poses a concept or question, back side gives the answer with a brief explanation. The deck can be reviewed inside NotebookLM or exported for import into Anki or similar spaced-repetition tools.

Best for: memorising terminology, methods, or key facts from a dense literature.

mosaic notebook create "Genomics Terminology" \
    --query "single cell RNA sequencing methods review" \
    --oa-only --flashcards

Infographic (--infographic)

A single-page visual summary — key statistics, findings, and relationships rendered as a designed graphic with icons, callouts, and short text blocks. Produced as an image file (PNG or PDF). Layout and detail level are chosen automatically.

Best for: posters, social media summaries, slide backgrounds, or any context where a dense text document is not appropriate.

mosaic notebook create "Climate Models 2024" \
    --query "climate model uncertainty projection 2024" \
    --oa-only --infographic

Slide Deck (--slide-deck)

A presentation exported as a set of slides (Google Slides format or equivalent), covering the topic from introduction through methodology, key results, and conclusions. Each slide has a title, bullet points, and — where appropriate — placeholder notes for a speaker. Ready to open, customise, and present.

Best for: conference talks, group seminars, lab meetings, course lectures.

mosaic notebook create "Graph Neural Networks" \
    --query "graph neural network survey applications" \
    --oa-only -y 2021-2025 --slide-deck

Data Table (--data-table)

A structured table that extracts comparable data points — numbers, metrics, experimental conditions, model names, benchmark scores, dates — from the sources and aligns them in rows and columns. Useful for spotting trends, comparing methods, or preparing a results table for a paper.

Best for: systematic reviews, benchmark comparisons, meta-analyses.

mosaic notebook create "LLM Benchmarks" \
    --query "large language model benchmark evaluation 2024" \
    --oa-only --data-table

Mind Map (--mind-map)

An interactive, zoomable graph of concepts, sub-topics, and the relationships between them, extracted from the notebook sources. Nodes represent ideas or entities; edges show how they connect. Viewable and navigable inside NotebookLM Studio.

Best for: exploring a new field, identifying gaps, planning a research agenda, or explaining a complex topic visually.

mosaic notebook create "Metabolomics Overview" \
    --query "metabolomics mass spectrometry data analysis" \
    --oa-only --mind-map

Full option reference

Option	Default	Description
NAME	(required)	Notebook name
--query, -q	—	Search query (alternative to --from-dir)
--from-dir	—	Directory of PDFs to import (alternative to --query)
--max, -n	10	Max results per source when using --query
--oa-only	off	Only include open-access papers
--pdf-only	off	Only include papers with a known PDF URL
--year, -y	—	Year filter (same formats as search)
--author, -a	—	Author filter, repeatable
--journal, -j	—	Journal name substring filter
--field, -f	all	Scope query to title, abstract, or all
--raw-query	—	Raw query sent directly to APIs, bypasses all transforms
--download-dir	config	Override PDF download directory for this run only
--podcast	off	Queue an Audio Overview after import
--video	off	Queue a Video Overview after import
--briefing	off	Queue a Briefing Doc after import
--study-guide	off	Queue a Study Guide after import
--quiz	off	Queue a Quiz after import
--flashcards	off	Queue Flashcards after import
--infographic	off	Queue an Infographic after import
--slide-deck	off	Queue a Slide Deck after import
--data-table	off	Queue a Data Table after import
--mind-map	off	Queue a Mind Map after import

What happens under the hood

1. MOSAIC searches all enabled sources with the given query.
2. For each result, it attempts to download the PDF (with Unpaywall fallback).
3. Papers with a local PDF are uploaded as files; the rest are added by URL.
4. Up to 50 sources are imported (NotebookLM's hard limit).
5. Any artifact flags (--podcast, --briefing, etc.) are queued — they appear in NotebookLM Studio within a few minutes.

Typical workflow

# 1. Find and collect papers on a topic
mosaic notebook create "Protein Folding 2024" \
    --query "protein structure prediction alphafold" \
    --oa-only --max 15 --podcast --briefing

# 2. Open the link printed by MOSAIC in your browser
# 3. Chat with the notebook; review the briefing doc and listen to the podcast

Notes

• Source limit: NotebookLM accepts at most 50 sources per notebook. For larger result sets, only the first 50 are imported.
• Artifact timing: Generation takes a few minutes. The notebook is usable immediately while artifacts render in the background.
• Credentials: notebooklm login stores a session cookie locally. Re-run if you are signed out.