Introduction

Modaic is production infrastructure for running and aligning System One models. We provide a sane way to manage and adapt these models to your domain without constant upkeep.

Define the shape of the decision and run alignment to improve it. Modaic systematically trains and recalibrates the model using your feedback, helping you adapt it to your domain with less labeling effort.

Go from calibrated in general → calibrated for your use case

Quickstart

Create a custom model, then run it. Replace acme with your username or organization slug.

import { Modaic } from "modaic";

const modaic = new Modaic({
  apiKey: process.env.MODAIC_API_KEY!,
});

const model = await modaic.models.create({
  workspace: "acme",
  slug: "support-priority",
  model: "mo-fast",
  questions: {
    priority: {
      type: "choice",
      instructions: "Choose the support priority.",
      criteria: {
        low: "No financial or time-sensitive impact.",
        normal: "Routine customer issue.",
        high: "Financial impact or an urgent blocker.",
      },
    },
  },
});

const decision = await model.decisions.create({
  state: {
    ticket: "I was charged twice. Please refund one charge.",
  },
});

console.log(decision.answers.priority);
// { type: "choice", choice: "high", probabilities: { ... }, confidence: 0.94 }

Features

• Make decisions — Ask typed questions and get structured answers, probabilities, and confidence.
• Create models — Save questions and configuration in a versioned model repository.
• Collect feedback — Add examples and corrections from domain experts.
• Align models — Adapt a model to your criteria and data distribution.
• Review in the platform — Inspect decisions, annotate examples, and manage alignment runs.

Get started

1. Quickstart — Make your first decision.
2. Using models — Create and call a versioned model.
3. Aligning your model — Improve it with annotated examples.