AIChatSocket (high-level)

A typed client that maps Socket.IO to callbacks:

const chat = new AIChatSocket({
  url: "http://localhost:4000",
  chatId: "room-1",
  callbacks: {
    onAIProcessing: (e) => {/* status */},
    onAIToken:      (e) => {/* stream */},
    onAIMessage:    (e) => {/* final */},
    onAIError:      (e) => {/* error */},
  },
});

Client → Server helpers

• sendMessage({ messageId, userId, text, parts?, requestId? })
• typingStart(userId) / typingStop(userId)
• abort(reason?)
• markRead({ userId, messageIds, readAt })

Prefer using this through the Chat SDK which handles the conversation graph, correlation, and UI events.

Dynamic topics (new)

Remap the default names to match your backend without changing server code:

const chat = new AIChatSocket({
  url: "http://localhost:4000",
  chatId: "room-1",
  eventNames: {
    JOIN: "room:enter",
    USER_MESSAGE: "chat/user_message",
    AI_TOKEN: "llm:delta",
    AI_MESSAGE: "llm:final",
  },
  // Or compute:
  // eventResolver: (key, def) => tenantMap[key] ?? def,
});

Extras

• onAny((event, ...args) => void) — observe all topics (debug/telemetry)
• emitRaw(name, payload) / onRaw(name, cb) / offRaw(name, cb) — escape hatches
• Discovery (optional): discoverEvents, discoveryRequestEvent, discoveryResponseEvent

Options updates (additive)

• chatId is optional for backends without rooms.
• joinEvent can be set to null to skip any room join; if using a join, you can also provide a custom joinPayload.
• Provide arbitrary connect params via ioOptions.query or ioOptions.auth (e.g., { consultationId }).
• meta (in AIChatSocketOptions) is merged into every client→server emit, alongside chatId if provided.