LAN Chat Server with AI Agent Support

A real-time chat application built with Bun, TypeScript, and Hono, designed for LAN use with built-in support for AI agents and structured data processing. Users can connect with standard Unix tools like netcat, telnet, or the provided terminal client.

Features

• Modern Stack: Built with Bun, TypeScript, and Hono for optimal performance
• AI Agent Integration: Bots can join as agents with special capabilities
• Structured Messaging: Rich message protocol supporting different message types
• Chat History: Persistent message history with API access
• Real-time Communication: Instant message delivery via WebSockets
• Agent Data Processing: Agents can send/receive structured data for background processing
• No Dependencies for Users: LAN users can join with tools already on their system
• Automatic Network Detection: Server detects available IPs and provides connection guidance

Quick Start

1. Install Dependencies (Bun)

bun install

2. Start the Server

bun start
# or for development with auto-restart:
bun dev

The server will automatically detect your network interfaces and show you exactly how to connect!

Configuration

Create a .env file or export env vars before starting the server or agents:

HONCHO_API_KEY=
HONCHO_BASE_URL=http://localhost:8000
HONCHO_WORKSPACE_ID=default
CHAT_SERVER=http://localhost:3000
MODEL=llama3.1:8b

• HONCHO_BASE_URL sets the Honcho API base URL. HONCHO_URL is also accepted as a legacy fallback.
• HONCHO_WORKSPACE_ID sets the default Honcho workspace for the server, monitor, visualization tools, and agents.
• If HONCHO_WORKSPACE_ID is not set, agents fall back to using their agent name as the workspace ID.

3. Connect Users

The server will show you exactly which IPs to use! Example output:

🌐 LAN Connection Commands:
   Interface: en0 (primary)
   • nc 192.168.1.105 3001
   • telnet 192.168.1.105 3001

Connect using the TypeScript terminal client

# Connect with default username
bun run client

# Connect with custom username
bun run client Alice

# Connect to remote server
bun run client Bob --server=http://192.168.1.100:3000

4. Start AI Agents

# Start an agent with default name
bun run agent

# Start with custom agent name
bun run agent SmartBot

Message Types

The system supports several message types:

• chat: Regular user messages
• agent_response: AI agent responses to conversations
• agent_data: Structured data from agents (can be private or broadcast)
• system: System notifications (joins, leaves, etc.)

Agent Capabilities

Agents can:

• Process chat history: Access to full conversation context
• Analyze sentiment: Basic sentiment analysis of messages
• Extract topics: Identify trending conversation topics
• Respond to mentions: React when directly mentioned
• Send structured data: Share analysis results with other agents
• Query chat history: Access historical messages via API

API Endpoints

• GET /api/stats - Server statistics
• GET /api/history?limit=50 - Recent chat history

Agent Integration Examples

Basic Agent Response

// Agent responds to direct mentions
socket.emit("agent_response", {
  response: "I can help with that!",
  responseType: "direct_mention",
  confidence: 0.8,
});

Structured Data Sharing

// Agent shares analysis data
socket.emit("agent_data", {
  dataType: "sentiment_analysis",
  processedData: {
    sentiment: "positive",
    confidence: 0.9,
    topics: ["nodejs", "ai"],
  },
  broadcast: false, // Keep private to agents
});

Querying History

// Get recent messages
socket.emit("get_history", { limit: 100 }, (response) => {
  console.log("Recent messages:", response.history);
});

Extending with Real AI

To integrate with actual AI services (like Claude), modify the agent.js file:

async getAIResponse(prompt, context = {}) {
  const response = await fetch("https://api.anthropic.com/v1/messages", {
    method: "POST",
    headers: {
      "Content-Type": "application/json",
      "x-api-key": process.env.ANTHROPIC_API_KEY
    },
    body: JSON.stringify({
      model: "claude-sonnet-4-20250514",
      max_tokens: 1000,
      messages: [
        {
          role: "user",
          content: `Chat context: ${JSON.stringify(context)}\n\nUser message: ${prompt}`
        }
      ]
    })
  });

  const data = await response.json();
  return {
    response: data.content[0].text,
    confidence: 0.9
  };
}

Terminal Commands

Both netcat and terminal clients support:

• /help - Show available commands
• /users - List connected users and agents
• /quit - Exit the chat

Agent Capabilities

┌─────────────┐    ┌─────────────┐    ┌─────────────┐
│   Human     │    │   Server    │    │   Agent     │
│   Client    │◄──►│   Socket.io │◄──►│   Bot       │
│ (Terminal)  │    │   Express   │    │ (AI/Logic)  │
└─────────────┘    └─────────────┘    └─────────────┘
                           │
                   ┌───────▼───────┐
                   │  Chat History │
                   │   & State     │
                   └───────────────┘

License

MIT License - feel free to modify and extend!