Quickstart

This walkthrough mirrors the flow from the Mozaik README: build a context from a developer (system) message and a user message, call the model through OpenAIResponses, merge the returned context items back into the context, and save / reload via a ContextRepository.

Example

import {
  Context,
  DeveloperMessage,
  UserMessage,
  InMemoryContextRepository,
  OpenAIResponses,
  Gpt54,
} from '@mozaik-ai/core';

const message = UserMessage.create('Tell me a joke about birds');
const developerMessage = DeveloperMessage.create(
  'You are a joke teller. You will be given a topic and you will need to tell a joke to the user.',
);

const projectId = `pr-${crypto.randomUUID()}`;
const contextRepository = new InMemoryContextRepository();
const context = Context.create(projectId)
  .addItem(developerMessage)
  .addItem(message);

await contextRepository.save(context);

const openresponses = new OpenAIResponses();
const newContextItems = await openresponses.infer(Gpt54, context);
context.applyModelOutput(newContextItems);

await contextRepository.save(context);
const restoredContexts = await contextRepository.getByProjectId(projectId);
console.log(restoredContexts);

What this is doing

1. Context.create(projectId) — Allocates an empty context keyed by your own id (e.g. conversation or project).
2. addItem(...) — Appends OpenResponses-shaped input items (DeveloperMessage, UserMessage, etc.).
3. OpenAIResponses.infer(model, context) — Runs the context runtime: structured context in, new output items returned (assistant message, tool calls, reasoning segments, depending on the model and settings).
4. context.applyModelOutput(newContextItems) — Appends those outputs to the same context for the next turn or for persistence.
5. InMemoryContextRepository — Example storage; swap for your database or cache in real apps.

Model identifiers such as Gpt54, Gpt54Mini, and Gpt54Nano are exported from the package for use with the OpenAI integration—check the version you installed for the exact set of named models.

Related