notebooklm

NotebookLM 自动化

适用于 Google NotebookLM（notebooklm-py）的非官方 Python CLI 和 API。提供完整的程序化访问能力，包括 Web UI 未公开的功能。

先决条件

• Python 3.10+
• 可访问 NotebookLM 的 Google 账号
• 通过 Playwright 进行一次性浏览器登录

安装

# Install with browser login support
pip install "notebooklm-py[browser]"
playwright install chromium

# Linux only: also run
playwright install-deps chromium

身份验证

首次设置需要通过浏览器登录：

notebooklm login
# Opens Chromium → sign into Google → press Enter when done
# Session saved to ~/.notebooklm/storage_state.json

检查身份验证状态：notebooklm auth check --test

对于无头/CI 环境，可从本地计算机复制 storage_state.json，或设置 NOTEBOOKLM_AUTH_JSON 环境变量。

环境变量

变量	说明	默认值
NOTEBOOKLM_HOME	配置目录	~/.notebooklm
NOTEBOOKLM_AUTH_JSON	内嵌身份验证 JSON（CI/CD）	—
NOTEBOOKLM_LOG_LEVEL	DEBUG/INFO/WARNING/ERROR	WARNING
NOTEBOOKLM_DEBUG_RPC	启用 RPC 调试（1）	false

核心工作流

典型工作流为：创建笔记本 → 添加来源 → 生成内容 → 下载。

1. 笔记本管理

notebooklm create "My Research"       # Create notebook
notebooklm list                        # List all notebooks
notebooklm use <id>                    # Set active notebook (supports partial ID)
notebooklm summary                     # AI summary of current notebook
notebooklm rename "New Title"          # Rename
notebooklm delete <id>                 # Delete

2. 添加来源

系统会根据类型自动检测来源：

notebooklm source add "https://example.com/article"          # URL
notebooklm source add "https://youtube.com/watch?v=..."       # YouTube
notebooklm source add ./document.pdf                           # File (PDF, MD, DOCX, TXT, audio, video, images)
notebooklm source add-drive <drive-file-id> "Title"           # Google Drive
notebooklm source add-research "climate policy" --mode deep --import-all  # Research agent

其他来源命令：

notebooklm source list                # List sources
notebooklm source fulltext <id>       # Get source full text
notebooklm source guide <id>          # AI-generated source guide
notebooklm source rename <id> "New"   # Rename
notebooklm source refresh <id>        # Re-fetch URL source
notebooklm source delete <id>         # Delete

3. 聊天 / 问答

notebooklm ask "What are the key findings?" -s <source_id>
notebooklm ask "Compare sources" --json --save-as-note --note-title "Comparison"
notebooklm history                     # View chat history
notebooklm history --save              # Save history as note

4. 内容生成

所有生成命令均支持：-s/--source（可重复使用，仅限指定来源）、--json、--language、--retry N。

大多数命令都是异步的——使用 --wait 阻塞等待完成。

音频概览（播客）

notebooklm generate audio "Focus on practical applications" \
  --format deep-dive \    # deep-dive | brief | critique | debate
  --length long \         # short | default | long
  --wait

视频概览

notebooklm generate video "Explain the architecture" \
  --format explainer \    # explainer | brief
  --style whiteboard \    # auto | classic | whiteboard | kawaii | anime | watercolor | retro-print | heritage | paper-craft
  --wait

幻灯片

notebooklm generate slide-deck "Executive summary" \
  --format detailed \     # detailed | presenter
  --length default \      # default | short
  --wait

# Revise a specific slide
notebooklm generate revise-slide "Add more data points" \
  -a <artifact_id> --slide 2 --wait    # slide is zero-based

学习材料

# Quizzes
notebooklm generate quiz --difficulty hard --quantity more --wait

# Flashcards
notebooklm generate flashcards --difficulty medium --wait

视觉内容与数据

# Infographic
notebooklm generate infographic \
  --orientation landscape \   # landscape | portrait | square
  --detail detailed \         # concise | standard | detailed
  --wait

# Mind map (synchronous, no --wait needed)
notebooklm generate mind-map

# Data table
notebooklm generate data-table "Compare metrics across studies" --wait

报告

notebooklm generate report "Security analysis" \
  --format briefing-doc \     # briefing-doc | study-guide | blog-post | custom
  --append "Include threat modeling" \
  --wait

5. 下载内容

所有下载命令均支持：-a/--artifact、--all、--latest、--earliest、--name、--force、--no-clobber、--dry-run、--json。

notebooklm download audio ./podcast.mp3
notebooklm download video ./overview.mp4
notebooklm download slide-deck ./slides.pptx --format pptx   # or pdf (default)
notebooklm download infographic ./info.png
notebooklm download report ./report.md
notebooklm download mind-map ./map.json
notebooklm download data-table ./data.csv
notebooklm download quiz --format json ./quiz.json       # json | markdown | html
notebooklm download flashcards --format markdown ./cards.md

6. 分享

notebooklm share status
notebooklm share public --enable           # Create public link
notebooklm share view-level full           # full | chat
notebooklm share add [email protected] --permission editor -m "Check this out"
notebooklm share remove [email protected]

7. 语言

notebooklm language list                   # 80+ languages
notebooklm language get
notebooklm language set ja                 # Set to Japanese

Python API

用于程序化工作流的完整异步 API：

import asyncio
from notebooklm import NotebookLMClient

async def main():
    async with await NotebookLMClient.from_storage() as client:
        # Create notebook and add sources
        nb = await client.notebooks.create("Research")
        await client.sources.add_url(nb.id, "https://example.com")

        # Generate audio overview
        artifact = await client.artifacts.generate_audio(
            nb.id, description="Deep dive on findings",
            format=AudioFormat.DEEP_DIVE, length=AudioLength.LONG
        )

        # Wait and download
        await client.artifacts.wait(nb.id, artifact.id)
        await client.artifacts.download_audio(nb.id, artifact.id, "output.mp3")

        # Chat with sources
        result = await client.chat.ask(nb.id, "Summarize key points")
        print(result.answer)

asyncio.run(main())

API 模块：client.notebooks、client.sources、client.artifacts、client.chat、client.research、client.notes、client.settings、client.sharing

常见示例

从研究到播客的工作流

notebooklm create "Climate Research"
notebooklm use <id>
notebooklm source add "https://en.wikipedia.org/wiki/Climate_change"
notebooklm source add-research "climate change solutions 2025" --mode deep --import-all
notebooklm generate audio "Focus on actionable solutions" --format debate --length long --wait
notebooklm download audio ./climate-debate.mp3

将文档分析转为学习材料

notebooklm create "Exam Prep"
notebooklm use <id>
notebooklm source add ./textbook.pdf
notebooklm generate quiz --difficulty hard --quantity more --wait
notebooklm generate flashcards --wait
notebooklm download quiz --format markdown ./quiz.md
notebooklm download flashcards --format json ./cards.json

批量导入 + 完整报告

notebooklm create "Literature Review"
notebooklm use <id>
for f in ./papers/*.pdf; do notebooklm source add "$f"; done
notebooklm generate report "Systematic review" --format briefing-doc --wait
notebooklm download report ./review.md

故障排除

问题	解决方法
身份验证已过期	再次运行 notebooklm login
找不到 playwright	运行 pip install "notebooklm-py[browser]"，然后运行 playwright install chromium
生成卡住	对待处理来源使用 notebooklm source wait <id>，并检查 --retry 标记
部分 ID 不匹配	使用更多笔记本 ID 字符
调试 API 调用	设置 NOTEBOOKLM_LOG_LEVEL=DEBUG 或 NOTEBOOKLM_DEBUG_RPC=1

---

注意事项

• 来源上传超过 200MB 时，即使请求返回 200 也会静默失败——下游会一直显示“处理中”。上传前请检查文件大小。
• 每个项目的并发音频/视频生成受速率限制——第二个并发生成任务会失败，并显示笼统的“请稍后重试”提示。请串行执行生成任务。
• gcloud 身份验证与 notebooklm-account 不同——登录了 gcloud 的用户可能无法访问 notebooklm；两者不一致时，请显式传入 --account。
• 上传时的来源顺序会影响生成内容中的引用优先级——为调整顺序而重新上传会改变输出风格。
• Studio 类型的超时时间不同——音频约 5 分钟，视频约 15 分钟，幻灯片约 3 分钟。对于耗时较长的类型，CLI 默认的 60 秒超时会导致任务中途退出。