Developers · SDKs

Typed clients. Zero boilerplate.

Each SDK ships with built-in retries, exponential backoff, request signing, and streaming helpers — so you focus on the model, not the plumbing.

Python

pip install kepler-qmax

from kepler import Client

q = Client(api_key="sk_live_...")
result = q.predict(
    model="kepler-q-max",
    task="fraud_detection",
    features=[0.12, 0.84, 1.7],
)
print(result.label, result.confidence)

TypeScript

npm install @kepler/qmax

import { Kepler } from "@kepler/qmax";

const q = new Kepler({ apiKey: process.env.KEPLER_KEY! });

const result = await q.predict({
  model: "kepler-q-max",
  task: "fraud_detection",
  features: [0.12, 0.84, 1.7],
});
console.log(result.label, result.confidence);

go get github.com/sonicium/kepler-go

client := kepler.New("sk_live_...")
res, err := client.Predict(ctx, &kepler.PredictRequest{
    Model:    "kepler-q-max",
    Task:     "fraud_detection",
    Features: []float64{0.12, 0.84, 1.7},
})
if err != nil { log.Fatal(err) }
fmt.Println(res.Label, res.Confidence)

Versioning & support

SDKs follow semver. Major versions are supported for 24 months after release. Security patches backported to all supported majors.

Source & issues

All clients are open-source on GitHub under the sonicium organization. PRs welcome — please run make lint test before submitting.