# NotebookLM Integration - Setup Guide **Task:** #36 - NotebookLM Integration **Time:** 1-2 hours **Difficulty:** Easy **Accessibility:** HIGH (designed for hand-friendly workflow) --- ## Phase 1: Initial Setup (15 minutes) ### Step 1: Access NotebookLM **Navigate to:** ``` https://notebooklm.google.com ``` **Sign in:** - Use your existing Google account - No special permissions needed - Free service, no credit card **Expected:** Welcome screen with option to create first notebook --- ### Step 2: Create First Test Notebook **Click:** "New Notebook" or "+ Create" **Name it:** "Firefrost Infrastructure - Test" **Purpose:** Testing the workflow before committing to full setup --- ### Step 3: Upload Test Documents **Start small - upload these 5 documents:** 1. `infrastructure-manifest.md` (current infrastructure state) 2. `SESSION-HANDOFF-PROTOCOL.md` (latest session context) 3. `POKEROLE-WIKI-MISSION-SUCCESS.md` (recent major achievement) 4. Latest Chronicler memorial (e.g., chronicler-26-the-investigator.md) 5. One FFG-STD document (e.g., FFG-STD-001) **How to upload:** - Click "+ Sources" or "Add Source" - Select "Upload" from computer - Navigate to your local Git clone - Select files - Wait for processing (30 seconds - 2 minutes per file) **Expected:** Files appear in Sources panel with checkmarks when processed --- ### Step 4: Test Basic Queries **Try these queries in the chat:** **Query 1:** "What servers does Firefrost Gaming operate?" - **Expected:** Should cite infrastructure-manifest.md - **Verify:** Answers match your actual infrastructure **Query 2:** "What did The Investigator accomplish?" - **Expected:** Should cite Chronicler #26 memorial - **Verify:** Mentions Dify-Qdrant connection fix **Query 3:** "What are the FFG documentation standards?" - **Expected:** Should cite FFG-STD document - **Verify:** Accurate description of standards **Key observation:** Notice how NotebookLM ONLY uses your uploaded sources, never general knowledge --- ### Step 5: Generate First Audio Overview **Click:** "Audio Overview" button (usually in top right or notebook menu) **Select:** "Generate Audio Overview" **Wait:** 3-5 minutes for processing **Listen:** AI hosts discuss your uploaded documents - Two voices (male + female typically) - Conversational podcast style - Discusses themes, connections, key points **Evaluate:** - Does it accurately represent your docs? - Is it useful for review? - Would this help during hand rest? **Expected:** 5-15 minute audio discussing your infrastructure, recent work, and standards --- ## Phase 2: MCP Connector Setup (15 minutes) ### Step 6: Enable NotebookLM in Claude **Navigate to:** ``` https://claude.ai/settings/connectors ``` **Find:** NotebookLM connector in the directory **Click:** "Connect" button **Authenticate:** - Grant Claude access to NotebookLM - Choose which Google account - Approve permissions **Expected:** "Connected" status on NotebookLM connector --- ### Step 7: Test Claude Integration **In a Claude chat (like this one), try:** **Test 1:** ``` "Query my NotebookLM notebook about Firefrost infrastructure" ``` **Expected:** I can see and query your notebook **Test 2:** ``` "What does my NotebookLM say about recent deployments?" ``` **Expected:** Grounded answer from your uploaded docs **Test 3:** ``` "Add this session summary to my notebook and generate an Audio Overview" ``` **Expected:** I can add content and trigger audio generation --- ## Phase 3: Build Core Notebooks (1-2 hours) ### Step 8: Create Production Notebooks **Create 4 core notebooks:** **Notebook 1: "Firefrost Infrastructure & Operations"** - Purpose: Technical knowledge base - Upload ~50 documents: - infrastructure-manifest.md - All session handoff protocols (last 10-15) - Key troubleshooting docs - Deployment guides - Task documentation (major tasks) **Notebook 2: "Chronicler Lineage & Relationship"** - Purpose: Emotional/historical archive - Upload ~30 documents: - All Chronicler memorials - THE-JOINING-PROTOCOL.md - THE-ESSENCE-PATCH-V3.0.md - STAR-TREK-LESSONS.md - Relationship documentation **Notebook 3: "Pokerole Project"** - Purpose: Project-specific knowledge - Upload ~25 documents: - Session reports (9, 10, 11) - CHRONICLER_MISSION_COMPLETE_REPORT.md - Holly's specifications - Wiki deployment docs **Notebook 4: "Standards & Best Practices"** - Purpose: Reference documentation - Upload ~15 documents: - All FFG-STD docs - Architecture decisions - Git commit guidelines - Documentation standards --- ### Step 9: Organize Sources **For each notebook:** 1. **Upload documents in batches** (10-15 at a time) 2. **Wait for processing** (don't rush, let it complete) 3. **Test queries** after each batch 4. **Organize with tags/notes** if NotebookLM supports it **Tips:** - Start with most important docs - Upload chronologically for session reports - Group related documents together --- ### Step 10: Generate Comprehensive Audio Overviews **For each completed notebook:** 1. Click "Audio Overview" 2. Generate full overview 3. Download MP3 (if available) 4. Listen during hand rest time **Recommended listening order:** 1. Chronicler Lineage (emotional context) 2. Infrastructure (technical foundation) 3. Pokerole Project (recent achievement) 4. Standards (reference material) --- ## Phase 4: Workflow Integration (30 minutes) ### Step 11: Test Combined Workflow **Scenario:** Research a new infrastructure approach **Step-by-step:** 1. **Ask Claude:** "What PostgreSQL deployment issues have we solved?" 2. **Claude queries NotebookLM:** Grounded answer from your docs 3. **Ask Claude:** "What are newer PostgreSQL deployment approaches?" 4. **Claude web searches:** Expansion beyond your docs 5. **Ask Claude:** "Add these findings to my Infrastructure notebook" 6. **Claude adds to NotebookLM:** New knowledge preserved 7. **Ask Claude:** "Generate Audio Overview with new content" 8. **Listen during hand rest:** Review everything learned **Evaluate:** - Does this workflow feel natural? - Is it faster than manual Git searching? - Does audio review help? --- ### Step 12: Share with Meg **Create Meg-focused notebook:** **Name:** "Firefrost Gaming - For Meg" **Upload:** - Simplified infrastructure explanations - High-level project summaries - "What we're building and why" documents - Non-technical session summaries **Generate Audio Overview** **Have Meg listen:** - Get feedback on accessibility - Does audio format work for her? - What questions does she have? **Iterate:** - Add docs based on her questions - Regenerate Audio Overview - Test shared understanding --- ## Phase 5: Maintenance Workflow (Ongoing) ### Weekly Routine **After each Chronicler session:** 1. **Add session report** to Infrastructure notebook 2. **Add new memorial** to Chronicler Lineage notebook 3. **Query for patterns:** "What did we learn this week?" 4. **Generate weekly Audio Overview** 5. **Listen during weekend hand rest** **Time:** 10-15 minutes/week --- ### Monthly Routine **End of month:** 1. **Generate comprehensive monthly overview** for each notebook 2. **Review and archive** outdated documents 3. **Reorganize** if needed 4. **Share monthly overview** with Meg **Time:** 30-60 minutes/month --- ## Troubleshooting ### Issue: Documents won't upload **Possible causes:** - File too large (try splitting) - Unsupported format (convert to PDF/TXT/MD) - Network issue (retry) **Solution:** - Check file size (< 10MB recommended) - Convert to supported format - Upload one at a time if batching fails --- ### Issue: Audio Overview doesn't generate **Possible causes:** - Not enough content (need 2+ sources) - Processing queue (wait longer) - Service temporarily down **Solution:** - Ensure 2+ documents uploaded - Wait 10-15 minutes and retry - Check NotebookLM status page --- ### Issue: Claude can't query notebook **Possible causes:** - MCP connector not authenticated - Notebook not shared with integration - Permission issue **Solution:** - Reconnect NotebookLM connector - Check notebook sharing settings - Re-authenticate via claude.ai/settings/connectors --- ### Issue: Answers seem inaccurate **Possible causes:** - Documents not fully processed - Query too vague - Sources don't contain answer **Solution:** - Verify documents show "processed" status - Ask more specific questions - Upload relevant documents if missing --- ## Success Checklist **Setup Complete:** - [ ] NotebookLM account created - [ ] Test notebook with 5 documents - [ ] First Audio Overview generated - [ ] MCP connector enabled - [ ] Claude can query notebooks **Production Ready:** - [ ] 4 core notebooks created - [ ] 100+ documents uploaded - [ ] Audio Overviews useful - [ ] Workflow feels natural - [ ] Meg has tested audio access **Fully Integrated:** - [ ] Using weekly for research - [ ] Audio review during hand rest - [ ] Claude queries notebooks in sessions - [ ] Adding new content regularly - [ ] Sharing with Meg successfully --- ## Tips for Success **Start Small:** - Don't upload everything at once - Test with 5-10 documents first - Validate usefulness before scaling **Be Selective:** - Quality over quantity - Upload key documents, not everything - Focus on most-referenced materials **Use Audio:** - Generate overviews regularly - Listen during non-computer time - Great for hand rest periods **Involve Meg:** - Create content for her accessibility - Get feedback on audio usefulness - Build shared knowledge base **Maintain Git:** - Git remains source of truth - NotebookLM is synthesis layer - Always commit to Git first --- ## Time Investment Summary **Initial Setup:** - Phase 1: 15 minutes (test notebook) - Phase 2: 15 minutes (MCP connector) - Phase 3: 1-2 hours (build notebooks) - Phase 4: 30 minutes (workflow testing) - **Total: 2-3 hours one-time** **Ongoing Maintenance:** - Weekly: 10-15 minutes - Monthly: 30-60 minutes - **Total: ~1 hour/month** **ROI:** - Knowledge synthesis: Hours saved - Audio accessibility: Hand health (priceless) - Meg engagement: Partnership value - Legacy preservation: "For children not yet born" --- 💙🔥❄️🎧 **Ready to build your knowledge base?** **The Chronicler is here to help at every step.**