SDK Reference

The Gambi SDK provides a Vercel AI SDK provider for using shared LLMs through a Gambi hub.

Installation

npm install gambi-sdk
# or
bun add gambi-sdk

createGambi(options)

Creates a Gambi provider instance.

Option	Type	Default	Description
roomCode	string	—	Room code to connect to. Required.
hubUrl	string	http://localhost:3000	Hub URL
defaultProtocol	"openResponses" | "chatCompletions"	"openResponses"	Protocol used by the top-level routing helpers

import { createGambi } from "gambi-sdk";

const gambi = createGambi({
  roomCode: "ABC123",
  hubUrl: "http://localhost:3000",
});

Protocol Selection

The SDK defaults to openResponses. Both protocols are first-class:

// Default: Responses API
const gambi = createGambi({
  roomCode: "ABC123",
});

// Chat Completions
const gambi = createGambi({
  roomCode: "ABC123",
  defaultProtocol: "chatCompletions",
});

You can also select per-call via namespaces:

gambi.openResponses.any();      // Responses API
gambi.chatCompletions.any();    // Chat Completions

Model Routing

Three routing methods are available. All return a Vercel AI SDK model instance.

gambi.any()

Routes to a random online participant.

const result = await generateText({
  model: gambi.any(),
  prompt: "Hello",
});

gambi.participant(id)

Routes to a specific participant by nickname or ID.

const result = await generateText({
  model: gambi.participant("alice"),
  prompt: "Hello",
});

gambi.model(name)

Routes to the first online participant running the specified model.

const result = await generateText({
  model: gambi.model("llama3"),
  prompt: "Hello",
});

All routing methods are also available under gambi.openResponses.* and gambi.chatCompletions.*.

Streaming

Use streamText from the Vercel AI SDK:

import { streamText } from "ai";

const stream = await streamText({
  model: gambi.any(),
  prompt: "Write a story",
});

for await (const chunk of stream.textStream) {
  process.stdout.write(chunk);
}

Generation Options

Standard Vercel AI SDK options are supported:

const result = await generateText({
  model: gambi.any(),
  prompt: "Explain recursion",
  temperature: 0.7,
  maxTokens: 500,
});

See the Vercel AI SDK docs for all available options.