milliseconds.ai Start building

Developer tools

Ship the idea. Skip the ceremony.

TypeScript, Python, or your terminal. A few lines to get a decision. The rest is up to your imagination.

Get a free test keyRead the quickstart
Before your first call

Create an API key in the console, then set it in your server or shell environment.

Environment
export MS_API_KEY="your-api-key"

TypeScript

Your labels.
Your types.

Result types follow the labels you supply. Inspect the winning label, the full score distribution, and confidence without unpacking an HTTP response.

Install
npm install @cloudraker/milliseconds

TypeScript reference
TypeScript
import { DecisionMachine } from '@cloudraker/milliseconds';

const dm = new DecisionMachine(); // reads MS_API_KEY
const result = await dm.classify('I was charged twice.', {
  billing: 'payments, invoices, charges and refunds',
  shipping: 'delivery, tracking and packages',
  account: 'login, passwords and profile settings',
});

console.log(result.label);
console.log(result.probability, result.confidence);

Python

From a notebook
to your pipeline.

Use sync or async clients for classification, extraction, and source checks. Keep your labels, schemas, and thresholds in code.

Install
pip install cloudraker-milliseconds

Python reference
Python
from milliseconds import DecisionMachine

with DecisionMachine() as dm:  # reads MS_API_KEY
    result = dm.classify('I was charged twice.', {
        'billing': 'payments, invoices, charges and refunds',
        'shipping': 'delivery, tracking and packages',
        'account': 'login, passwords and profile settings',
    })
    print(result.label)
    print(result.probability, result.confidence)

dm1 CLI

Your terminal
can make the call.

Read text from a file or a pipe. Write JSON for another program. Use exit codes to gate a script and inspect token usage on stderr.

Install
npm install -g @cloudraker/milliseconds

dm1 reference
dm1
dm1 classify "I was charged twice." \
  billing="payments, invoices, charges and refunds" \
  shipping="delivery, tracking and packages" \
  account="login, passwords and profile settings" \
  --json --usage

Keep building

The next useful link.

Take a working example into your project, then evaluate it on your own inputs.

26 working demosEach demo includes SDK and CLI examples.Agent skillsTeach your coding agent the API and evaluation patterns.DocumentationCapability guides, recipes, batching, and API reference.SDK sourceInspect the TypeScript, Python, and CLI implementations.

Using an OpenAI client? The API supports JSON-schema extraction and tool-call requests through its compatibility endpoint. Read the supported request shapes.