notebooklm-distiller

# NotebookLM Distiller

Automated knowledge extraction pipeline: search NotebookLM notebooks by keyword → generate deep questions or structured summaries → write linked Obsidian markdown notes.

**Five subcommands:**
- `distill` — extract knowledge from existing notebooks (qa / summary / glossary)
- `quiz` — generate quiz questions as JSON for Discord-based interactive sessions
- `evaluate` — evaluate a user's answer against notebook sources (JSON output)
- `research` — start a web research session inside NotebookLM on any topic
- `persist` — write any markdown content directly into the Obsidian vault

## When to use (trigger phrases)

Trigger `distill` subcommand when:
- User types `/notebooklm-distill` or `/notebooklm-distill-summary`
- User says "蒸馏", "提取知识", "distill notebooks", "extract from notebook"
- User wants NotebookLM content structured into Obsidian notes

Trigger `research` subcommand when:
- User says "研究一下 <topic>", "做网络调研", "research this topic in NotebookLM"
- User wants NotebookLM to gather web sources on a topic without providing URLs

Trigger `quiz` + `evaluate` subcommands when:
- User says "quiz me on X", "考考我", "出题测试我", "测验"
- User wants an interactive Q&A session in Discord on a NotebookLM topic
- **Orchestration flow (Discord)**:
  1. Call `quiz --keywords X` → get JSON with `notebook_id` + `notebook_name` + `questions[]`
  2. **MUST** announce source before Q1: `来，N 道题（来源：{notebook_name} · ID: {notebook_id[:8]}）`
  3. Send Q1 to Discord, wait for user reply
  4. Call `evaluate --notebook-id X --question Q1 --answer <reply>` → get JSON feedback
  5. Post feedback to Discord, proceed to Q2
  6. Repeat until all questions done or user says stop
- **CRITICAL**: Always show notebook source so user can verify questions came from NLM, not agent knowledge

Trigger `persist` subcommand when:
- User says "存到 Obsidian", "把这段内容写入知识库", "persist this to vault"
- User wants to archive discussion output or raw...

# NotebookLM Distiller

An [OpenClaw](https://github.com/openclaw) skill that extracts knowledge from Google NotebookLM notebooks and writes structured Markdown notes to your Obsidian vault.

> **Version 2.0** — Now with three subcommands: `distill`, `research`, and `persist`.

---

## Features

- **`distill`** — Extract knowledge from existing NotebookLM notebooks into Obsidian
  - Three modes: `qa` (15-20 deep Q&A pairs + common misconception per question), `summary` (5-section expert knowledge map), `glossary` (15-30 domain terms with expert vs beginner usage)
  - Keyword-based notebook matching (case-insensitive substring)
  - Auto-generated YAML frontmatter compatible with Obsidian
- **`quiz`** — Generate quiz questions as JSON for agent-orchestrated interactive sessions (e.g. Discord)
- **`evaluate`** — Evaluate a user's answer against notebook sources; returns structured feedback as JSON
- **`research`** — Start a NotebookLM web research session on any topic, wait for completion, output the notebook ID for follow-up distillation
- **`persist`** — Write any Markdown content directly into your Obsidian vault with frontmatter

No web-scraping dependencies required — pairs with [DeepReader](https://github.com/astonysh/OpenClaw-DeepReeder) for full URL-to-Obsidian automation.

---

## Installation

**1. Copy the skill into OpenClaw:**
```bash
cp -r notebooklm-distiller ~/.openclaw/skills/
```

**2. Install the NotebookLM CLI:**
```bash
pip3 install notebooklm-py
```

**3. Authenticate with Google (once only):**
```bash
notebooklm login
# Opens a browser — log in with your Google account linked to NotebookLM
```

**Requirements:** Python 3.10+, no extra pip packages beyond `notebooklm-py`.

---

## Usage

### Subcommand: `distill`

Extract knowledge from one or more notebooks whose titles match your keywords.

```bash
python3 ~/.openclaw/skills/notebooklm-distiller/scripts/distill.py distill \
  --keywords "machine learning" "transformer" \
  --topic "ML Research"...