A personal knowledge base powered by AI with hybrid search capabilities. Built with Cloudflare Workers, AI Search, and the Agents SDK.

• Hybrid Search — Combines vector similarity (semantic) + keyword matching (BM25) with RRF fusion
• Document Ingestion — Add journal entries, articles, notes, goals, and health data
• AI-Powered Chat — Ask questions about your knowledge base with cited sources
• Real-time Sync — WebSocket connection with message persistence
• Scheduled Tasks — Set reminders and recurring tasks
• MCP Integration — Connect external tools via Model Context Protocol

# Install dependencies
npm install

# Run locally (AI Search requires deployment)
npm run dev

Open http://localhost:5173 to see your wiki.

npm run deploy

This creates your AI Search instance and deploys the worker to Cloudflare.

Once deployed, you can:

• Journal entries — "Add a journal entry about my trip to Tokyo"
• Articles — "Save this article about machine learning"
• Notes — "Create a note about project ideas"
• Goals — "Track my goal to run a marathon"
• Health data — "Log my workout for today"

• Ask questions — "What did I learn about neural networks?"
• Find entries — "Show me my journal entries from last month"
• Compare methods — Use /compare <query> to see vector vs keyword vs hybrid results

• /stats — Get wiki statistics
• /lint — Check wiki health
• /compare <query> — Compare search methods side-by-side
• /debug — Show debug info

src/
  server.ts    # Chat agent with wiki tools and AI Search integration
  app.tsx      # Chat UI with wiki-specific branding
  client.tsx   # React entry point
  styles.css   # Tailwind + Kumo styles

┌─────────────┐     WebSocket      ┌─────────────────┐
│   Client    │ ◄────────────────► │  ChatAgent (DO) │
│  (React)    │                    │                 │
└─────────────┘                    │  ┌───────────┐  │
                                   │  │ AI Search │  │
                                   │  │  Instance │  │
                                   │  └───────────┘  │
                                   │        ▲        │
                                   │        │        │
                                   │  ┌─────┴─────┐  │
                                   │  │  Vector   │  │
                                   │  │  Keyword  │  │
                                   │  │   Index   │  │
                                   │  └───────────┘  │
                                   └─────────────────┘

Edit server.ts:

const result = streamText({
  model: workersai("@cf/moonshotai/kimi-k2.5")
  // ...
});

Modify the search options in server.ts:

return instance.search({
  messages: [{ role: "user", content: query }],
  ai_search_options: {
    retrieval: {
      retrieval_type: "hybrid",
      fusion_method: "rrf",
      match_threshold: 0.4, // Adjust relevance threshold
      max_num_results: 10, // Adjust result count
      boost_by: [{ field: "timestamp", direction: "desc" }], // Add relevance boosting

    },
    reranking: {
      enabled: true,
      model: "@cf/baai/bge-reranker-base"
    }
  }
});

In the ingestDocument tool, extend the docType enum:

docType: z.enum([
  "journal",
  "article",
  "note",
  "goal",
  "health",
  "recipe" // Add your custom type
]);

• AI Search Documentation
• Agents SDK
• Workers AI Models

MIT