kivv

# kivv - Quick Answers ## 🔑 Where did the API key come from? **Answer:** The API key `c3c74bbeba60635cf12a6b27e766c8b953fcd70ac4e4347f05d8bc68902d2f1d` was hardcoded in the initial database seed data. **Location:** `shared/schema.sql` line 127: ```sql INSERT OR IGNORE INTO users (username, email, api_key, display_name, is_active) VALUES ('jeff', 'jeffbarron@protonmail.com', 'c3c74bbeba60635cf12a6b27e766c8b953fcd70ac4e4347f05d8bc68902d2f1d', 'Jeff', 1), ('wife', 'wife@example.com', 'e98699bedad9746e231843b96150c0638b7cceb717c44d5f9010a272a5b8de5b', 'Wife', 1); ``` **⚠️ CRITICAL:** This key was exposed in git history (CHUNK4-VERIFICATION.md) and MUST be rotated immediately. --- ## 🔄 How do I rotate the API key? **Quick method:** ```bash cd /home/gat0r/kivv # Generate new random key NEW_KEY=$(openssl rand -hex 32) echo "New key: $NEW_KEY" # Update database npx wrangler d1 execute kivv-db --remote \ --command "UPDATE users SET api_key = '$NEW_KEY' WHERE username = 'jeff'" # Save the new key for Claude Desktop config echo "$NEW_KEY" > ~/kivv-api-key.txt ``` **Automated method:** ```bash cd /home/gat0r/kivv ./GET-STARTED.sh ``` This script will: 1. Generate a new random API key 2. Update the database 3. Test authentication 4. Generate Claude Desktop config for you --- ## 🧪 How do I test MCP integration? ### Test 1: Health Check (No Auth) ```bash curl https://kivv-mcp.jeffbarron.workers.dev/health ``` Expected: `{"status":"healthy",...}` ### Test 2: Authentication ```bash curl -H "x-api-key: YOUR_NEW_KEY" \ https://kivv-mcp.jeffbarron.workers.dev/mcp/status ``` Expected: `{"authenticated":true,"user":{"username":"jeff",...}}` ### Test 3: List Papers ```bash curl -X POST https://kivv-mcp.jeffbarron.workers.dev/mcp/list-library \ -H "x-api-key: YOUR_NEW_KEY" \ -H "Content-Type: application/json" \ -d '{"limit": 10}' ``` ### Test 4: Search Papers ```bash curl -X POST https://kivv-mcp.jeffbarron.workers.dev/mcp/search-papers \ -H "x-api-key: YOUR_NEW_KEY" \ -H "Content-Type: application/json" \ -d '{"query": "transformer", "limit": 5}' ``` ### Test 5: Claude Desktop Integration 1. **Edit config:** `~/.config/claude/claude_desktop_config.json` ```json { "mcpServers": { "kivv": { "url": "https://kivv-mcp.jeffbarron.workers.dev/mcp", "headers": { "x-api-key": "YOUR_NEW_KEY" } } } } ``` 2. **Restart Claude Desktop** (quit completely from menu) 3. **Test in Claude:** - "List my research papers" - "Search for papers about attention mechanisms" - "Mark paper 1 as explored" --- ## 🚀 How do I use kivv? ### Via Claude Desktop (Recommended) Once configured (see above), you can use natural language: **List papers:** ``` List my research papers Show me unexplored papers Show me bookmarked papers ``` **Search papers:** ``` Search for papers about transformers Find papers on AI safety Search for LLM interpretability papers ``` **Mark papers:** ``` Mark paper 1 as explored Bookmark paper 5 Add note to paper 3: "Interesting approach to RLHF" ``` ### Via MCP API Directly **List library:** ```bash curl -X POST https://kivv-mcp.jeffbarron.workers.dev/mcp/list-library \ -H "x-api-key: YOUR_KEY" \ -H "Content-Type: application/json" \ -d '{ "limit": 20, "offset": 0, "explored": false }' ``` **Search papers:** ```bash curl -X POST https://kivv-mcp.jeffbarron.workers.dev/mcp/search-papers \ -H "x-api-key: YOUR_KEY" \ -H "Content-Type: application/json" \ -d '{ "query": "attention mechanism", "limit": 10 }' ``` **Mark as explored:** ```bash curl -X POST https://kivv-mcp.jeffbarron.workers.dev/mcp/mark-explored \ -H "x-api-key: YOUR_KEY" \ -H "Content-Type: application/json" \ -d '{ "paper_id": 1, "explored": true, "bookmarked": true, "notes": "Very interesting paper on transformers" }' ``` ### Via RSS Feed Add to your RSS reader: - **RSS 2.0:** https://kivv-mcp.jeffbarron.workers.dev/feeds/jeff/rss.xml - **Atom 1.0:** https://kivv-mcp.jeffbarron.workers.dev/feeds/jeff/atom.xml Works with: Feedly, Inoreader, NetNewsWire, Reeder, etc. --- ## 📰 Where do I get the RSS feed? **Your RSS feeds are already live:** - **RSS 2.0:** https://kivv-mcp.jeffbarron.workers.dev/feeds/jeff/rss.xml - **Atom 1.0:** https://kivv-mcp.jeffbarron.workers.dev/feeds/jeff/atom.xml **Test in browser:** ```bash curl https://kivv-mcp.jeffbarron.workers.dev/feeds/jeff/rss.xml ``` **Add to RSS reader:** 1. Open your RSS reader (Feedly, Inoreader, etc.) 2. Click "Add feed" or "Subscribe" 3. Paste URL: `https://kivv-mcp.jeffbarron.workers.dev/feeds/jeff/rss.xml` 4. Papers will appear as new feed items **RSS feed features:** - ✅ XML escaping (XSS prevention) - ✅ Includes title, abstract, authors, arXiv link - ✅ Published date from arXiv - ✅ Unique GUID (arxiv_id) - ✅ Auto-updates when new papers collected --- ## 📚 How do I check if there are summarized papers? ### Method 1: Query Database Directly ```bash cd /home/gat0r/kivv # Count total papers npx wrangler d1 execute kivv-db --remote \ --command "SELECT COUNT(*) as total FROM papers" # Count papers with summaries npx wrangler d1 execute kivv-db --remote \ --command "SELECT COUNT(*) as with_summary FROM papers WHERE summary IS NOT NULL" # See recent papers npx wrangler d1 execute kivv-db --remote \ --command "SELECT arxiv_id, title, relevance_score, summary FROM papers ORDER BY created_at DESC LIMIT 5" ``` ### Method 2: Use MCP API ```bash curl -X POST https://kivv-mcp.jeffbarron.workers.dev/mcp/list-library \ -H "x-api-key: YOUR_KEY" \ -H "Content-Type: application/json" \ -d '{"limit": 10}' | jq '.papers[] | {arxiv_id, title, summary}' ``` ### Method 3: Claude Desktop ``` List my research papers ``` Claude will show you all papers in your library with summaries. ### Method 4: RSS Feed ```bash curl https://kivv-mcp.jeffbarron.workers.dev/feeds/jeff/rss.xml ``` Papers with summaries will have `<description>` tags containing the AI-generated summary. --- ## ⏰ When will automation run? **Automatic schedule:** Daily at 6:00 AM UTC **Check schedule:** 1. Go to https://dash.cloudflare.com 2. Workers & Pages → **kivv-automation** 3. **Triggers** tab 4. Should show: `0 6 * * *` **Manual trigger:** ```bash # Get CRON_SECRET from Cloudflare Dashboard # Workers → kivv-automation → Settings → Variables & Secrets curl -X POST https://kivv-automation.jeffbarron.workers.dev/run \ -H "Authorization: Bearer YOUR_CRON_SECRET" ``` **What automation does:** 1. Fetches your enabled topics from database (currently 3 topics for jeff) 2. Searches arXiv for new papers (last 24 hours) 3. Uses Claude Haiku to score relevance (0.0-1.0) 4. Papers scoring ≥0.7 get Claude Sonnet summaries 5. Stores papers and summaries in database 6. Creates RSS feed entries **Expected results:** - **Papers found:** 10-50 per day (depends on topics) - **Papers summarized:** 2-10 per day (only relevant ones) - **Cost:** ~$0.10-0.30 per day (budget cap: $1.00/day) --- ## 🏥 Health Checks ```bash # MCP server status curl https://kivv-mcp.jeffbarron.workers.dev/health # Automation worker status curl https://kivv-automation.jeffbarron.workers.dev/health ``` Both should return: ```json { "status": "healthy", "timestamp": "2025-12-01T00:00:00.000Z", "services": { "database": "connected", "cache": "connected", "storage": "connected" }, "version": "1.0.0" } ``` --- ## 🐛 Troubleshooting **No papers in library:** - Automation hasn't run yet (runs daily 6 AM UTC) - Trigger manually (see "Manual trigger" above) - Check topics are enabled: `SELECT * FROM topics WHERE enabled = 1` **Authentication fails:** - Verify API key is correct (no extra spaces) - Ensure you rotated the exposed key - Test with curl first (see "Test 2: Authentication" above) **MCP not working in Claude Desktop:** - Restart Claude Desktop COMPLETELY (quit from menu) - Check config syntax (valid JSON) - Verify URL includes `/mcp` path For more detailed troubleshooting, see [TROUBLESHOOTING.md](TROUBLESHOOTING.md) --- ## 📝 Your Current Configuration **Deployed URLs:** - MCP Server: https://kivv-mcp.jeffbarron.workers.dev - Automation: https://kivv-automation.jeffbarron.workers.dev **Database:** - Database ID: 1e80f2bf-462d-4d51-8002-a4cf26013933 - KV Namespace ID: 7f6b7437931c4c268c27d01a4169101b - R2 Bucket: kivv-papers **Your Account:** - Username: jeff - Email: jeffbarron@protonmail.com - ⚠️ API Key: **MUST BE ROTATED** (exposed in git history) **Your Topics:** 1. Machine Learning Theory 2. AI Safety & Alignment 3. Large Language Models --- ## 🚀 Quick Start Commands ```bash # 1. Run complete setup walkthrough cd /home/gat0r/kivv ./GET-STARTED.sh # 2. Or do it manually: # Generate new API key NEW_KEY=$(openssl rand -hex 32) # Update database npx wrangler d1 execute kivv-db --remote \ --command "UPDATE users SET api_key = '$NEW_KEY' WHERE username = 'jeff'" # Test authentication curl -H "x-api-key: $NEW_KEY" \ https://kivv-mcp.jeffbarron.workers.dev/mcp/status # Configure Claude Desktop nano ~/.config/claude/claude_desktop_config.json # Test RSS feed curl https://kivv-mcp.jeffbarron.workers.dev/feeds/jeff/rss.xml # Check papers npx wrangler d1 execute kivv-db --remote \ --command "SELECT COUNT(*) FROM papers" ``` --- **Need more help?** See: - [SETUP.md](SETUP.md) - Full setup guide - [TROUBLESHOOTING.md](TROUBLESHOOTING.md) - Common issues - [DEPLOYMENT.md](DEPLOYMENT.md) - Deployment details