assistant

Assistant

A showcase of all Project Think features, built with @cloudflare/think.

What this demonstrates

• Think base class — getModel(), configureSession(), getTools(), maxSteps for a batteries-included agent
• Built-in workspace — file tools (read, write, edit, find, grep, delete) auto-wired on every turn
• Sandboxed code execution — createExecuteTool lets the LLM write and run JavaScript in a Dynamic Worker via @cloudflare/codemode
• Self-authored extensions — extensionLoader + createExtensionTools let the agent create new tools at runtime
• Persistent memory — context blocks (soul, memory) the model can read and write across sessions
• Non-destructive compaction — older messages summarized when context overflows, originals preserved
• Searchable knowledge base — FTS5-backed AgentSearchProvider with search_context and set_context tools
• Dynamic configuration — typed AgentConfig with model tier and persona, persisted in SQLite
• Server-side tools — getWeather, calculate execute on the server
• Client-side tools — getUserTimezone runs in the browser via onToolCall
• Tool approval — calculate requires user approval for large numbers
• MCP integration — connect external tool servers, tools appear in the chat
• Lifecycle hooks — beforeTurn, beforeToolCall, afterToolCall, onStepFinish, onChatResponse
• Durable chat recovery — chatRecovery wraps turns in fibers for crash recovery
• Scheduled proactive turns — daily summary via saveMessages from a cron schedule
• Regeneration with branch navigation — v1/v2/v3 response versions via getBranches
• Stream resumption — page refresh replays the active stream (built into Think)
• useAgentChat — Think speaks the same CF_AGENT protocol as AIChatAgent

How to run

npm install
npm start

Key code

Server (src/server.ts):

export class MyAssistant extends Think<Env> {
  chatRecovery = true;
  extensionLoader = this.env.LOADER;

  getModel() { /* model tier from config */ }
  configureSession(session) {
    return session
      .withContext("memory", { ... })
      .onCompaction(createCompactFunction({ ... }))
      .compactAfter(50000)
      .withContext("knowledge", { provider: new AgentSearchProvider(this) })
      .withCachedPrompt();
  }
  getTools() {
    return {
      execute: createExecuteTool({ ... }),
      ...createExtensionTools({ ... }),
      getWeather: tool({ ... }),
      calculate: tool({ needsApproval: ..., ... })
    };
  }
}

Client (src/client.tsx) — uses useAgentChat from @cloudflare/ai-chat/react, with panels for workspace browsing, extension management, and dynamic configuration.

Related

• Think docs
• Think tools
• Think lifecycle hooks