feat: Extended NotebookLM capabilities - source management, media generation, intelligent filtering

CHANGELOG.md

@@ -5,6 +5,41 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [1.4.0] - 2026-01-24
+
+### Added
+- **Intelligent Source Filtering** - LLM-based relevance scoring before each query
+  - Uses Gemini or Claude to score each source (1-10) based on semantic relevance
+  - Automatically deselects irrelevant sources in browser for faster responses
+  - Fallback chain: Gemini CLI → Claude CLI → keyword matching
+  - Configurable threshold (`--threshold N`) for filtering sensitivity
+  - `--keyword-filter` flag to use keyword matching instead of LLM
+  - `--no-filter` flag to disable filtering entirely
+
+- **Source Summary Caching** - Automated extraction and caching of source metadata
+  - New `source_extractor.py` - Click-based extraction of Source Guide content
+  - New `source_filter.py` - LLM scoring and browser checkbox automation
+  - Auto-extracts source summaries when adding notebooks
+  - Cached in `data/library-source-summary/{notebook-id}.md`
+  - Incremental updates - only fetches NEW sources not in cache
+
+- **Source Management Commands**
+  - `update-sources --id ID` - Update source summary for a notebook
+  - `update-sources --id ID --force` - Re-extract all sources
+  - `update-sources --all` - Update all notebooks
+  - `show-sources --id ID` - Display cached source summaries
+
+### Changed
+- **Python 3 Enforcement** - All scripts now use `python3` instead of `python`
+- **Browser Switch** - Changed from Chrome to Chromium for better compatibility
+- **ask_question.py** - Now integrates source filtering before each query
+
+### Fixed
+- **Checkbox Automation** - Uses Playwright native clicks for reliable Angular interaction
+  - JavaScript `.click()` didn't trigger Angular change detection
+  - Now uses `checkbox.click(force=True)` for proper event handling
+  - Each deselection is verified after clicking
+
 ## [1.3.0] - 2025-11-21
 
 ### Added

README.md

@@ -161,21 +161,30 @@ This is a **Claude Code Skill** - a local folder containing instructions and scr
 
 ```
 ~/.claude/skills/notebooklm/
-├── SKILL.md              # Instructions for Claude
-├── scripts/              # Python automation scripts
-│   ├── ask_question.py   # Query NotebookLM
-│   ├── notebook_manager.py # Library management
-│   └── auth_manager.py   # Google authentication
-├── .venv/                # Isolated Python environment (auto-created)
-└── data/                 # Local notebook library
+├── SKILL.md                # Instructions for Claude
+├── scripts/                # Python automation scripts
+│   ├── ask_question.py     # Query NotebookLM with source filtering
+│   ├── notebook_manager.py # Library + source summary management
+│   ├── auth_manager.py     # Google authentication
+│   ├── source_filter.py    # LLM-based relevance scoring
+│   ├── source_extractor.py # Extract source summaries from notebooks
+│   ├── add_source.py       # Add sources (files, URLs, text)
+│   ├── list_sources.py     # List sources via UI scraping
+│   ├── remove_source.py    # Remove sources (with safety confirmation)
+│   ├── audio_generator.py  # Generate Audio Overviews
+│   └── video_generator.py  # Generate Video Overviews
+├── .venv/                  # Isolated Python environment (auto-created)
+└── data/                   # Local notebook library + source summaries
 ```
 
 When you mention NotebookLM or send a notebook URL, Claude:
 1. Loads the skill instructions
 2. Runs the appropriate Python script
-3. Opens a browser, asks your question
-4. Returns the answer directly to you
-5. Uses that knowledge to help with your task
+3. **Scores source relevance** using Gemini/Claude LLM
+4. **Deselects irrelevant sources** in the browser
+5. Opens a browser, asks your question
+6. Returns the answer directly to you
+7. Uses that knowledge to help with your task
 
 ---
 
@@ -190,6 +199,39 @@ No copy-paste between browser and editor. Claude asks and receives answers progr
 ### **Smart Library Management**
 Save NotebookLM links with tags and descriptions. Claude auto-selects the right notebook for your task.
 
+### **Intelligent Source Filtering**
+Before each query, the skill analyzes which sources are relevant:
+- **LLM-based scoring**: Uses Gemini or Claude to semantically score each source (1-10)
+- **Automatic deselection**: Unchecks irrelevant sources in the browser
+- **Faster responses**: NotebookLM only searches relevant documents
+- **Fallback support**: Gemini → Claude → keyword matching
+
+### **Source Summary Caching**
+When you add a notebook, the skill automatically:
+- Extracts summaries for each source document
+- Caches them locally for instant relevance scoring
+- Updates incrementally (only fetches new sources)
+
+### **Source Management**
+Add, list, and remove sources directly from Claude Code:
+- **Add sources**: Upload PDFs, paste URLs (websites/YouTube), or add text directly
+- **List sources**: View all sources in a notebook via reliable UI scraping
+- **Remove sources**: Permanently delete sources with safety confirmation
+
+### **Audio Overview Generation**
+Generate podcast-style audio summaries of your notebook content:
+- **Formats**: Deep Dive, Brief, Critique, Debate
+- **Lengths**: Short, Default, Long
+- **Multi-language support**: English, Spanish, Japanese, and more
+- **Custom instructions**: Guide what the AI hosts focus on
+
+### **Video Overview Generation**
+Generate video summaries with customizable visual styles:
+- **Formats**: Explainer, Brief
+- **Visual styles**: Auto, Classic, Whiteboard, Kawaii, Anime
+- **Multi-language support**
+- **Custom instructions**: Target specific use cases or audiences
+
 ### **Automatic Authentication**
 One-time Google login, then authentication persists across sessions.
 
@@ -206,10 +248,18 @@ Uses realistic typing speeds and interaction patterns to avoid detection.
 | What you say | What happens |
 |--------------|--------------|
 | *"Set up NotebookLM authentication"* | Opens Chrome for Google login |
-| *"Add [link] to my NotebookLM library"* | Saves notebook with metadata |
+| *"Add [link] to my NotebookLM library"* | Saves notebook with metadata + extracts source summaries |
 | *"Show my NotebookLM notebooks"* | Lists all saved notebooks |
-| *"Ask my API docs about [topic]"* | Queries the relevant notebook |
+| *"Ask my API docs about [topic]"* | Queries with intelligent source filtering |
 | *"Use the React notebook"* | Sets active notebook |
+| *"Update sources for my notebook"* | Re-fetches source summaries |
+| *"Show sources in my notebook"* | Displays cached source summaries |
+| *"Add this PDF to my notebook"* | Uploads file as a new source |
+| *"Add this URL to my notebook"* | Adds website or YouTube video as source |
+| *"List sources in my notebook"* | Shows all sources via UI scraping |
+| *"Remove [source] from my notebook"* | Permanently deletes a source (with confirmation) |
+| *"Generate an audio overview"* | Creates podcast-style audio summary |
+| *"Generate a video overview"* | Creates video summary with visuals |
 | *"Clear NotebookLM data"* | Fresh start (keeps library) |
 
 ---
@@ -264,11 +314,16 @@ All data is stored locally within the skill directory:
 
 ```
 ~/.claude/skills/notebooklm/data/
-├── library.json       - Your notebook library with metadata
-├── auth_info.json     - Authentication status info
-└── browser_state/     - Browser cookies and session data
+├── library.json              - Your notebook library with metadata
+├── auth_info.json            - Authentication status info
+├── browser_state/            - Browser cookies and session data
+└── library-source-summary/   - Cached source summaries
+    └── {notebook-id}.md      - Source titles and summaries for each notebook
 ```
 
+**Source Summary Files:**
+Each notebook gets a markdown file with summaries of all sources. These are used for intelligent source filtering - the LLM scores each source's relevance to your question before querying.
+
 **Important Security Note:**
 - The `data/` directory contains sensitive authentication data and personal notebooks
 - It's automatically excluded from git via `.gitignore`