@alfe.ai/chat

Runtime-agnostic WebSocket client for the Alfe chat service

Part of Alfe — the operating system for AI agents: build, deploy, and run agents with persistent memory, identity, integrations, and channels. See the documentation to get started.

Install

npm install @alfe.ai/chat

Usage

import { ChatServiceClient, resolveAlfeChat } from '@alfe.ai/chat'

const { apiKey, chatWsUrl } = resolveAlfeChat()
if (!apiKey || !chatWsUrl) throw new Error('Alfe chat is not configured')

const client = new ChatServiceClient({
  apiKey,
  wsUrl: chatWsUrl,
  onRequest: async (request) => {
    // Dispatch the authenticated request to the agent runtime, then reply with
    // client.sendResponse(request.id, true, result).
  },
  onConnectionChange: (connected) => {
    console.log(`Chat relay ${connected ? 'connected' : 'disconnected'}`)
  },
})

client.start()

start() is idempotent and the client reconnects with bounded exponential backoff and jitter. stop() permanently fences callbacks from the old socket until start() is called again.

Ordinary responses/events sent while disconnected use a bounded in-memory queue. sendRequest() is deliberately different: it requires a live socket and is never moved into the offline queue, so a timed-out RPC cannot execute later after reconnect.

resolveAlfeChat() normalizes explicit configuration, then reads ~/.alfe/config.toml, then falls back to ALFE_API_KEY and ALFE_CHAT_WS_URL.

Links

• 🌐 Website: https://alfe.ai
• 📚 Docs: https://docs.alfe.ai