Configuration

Client options

const jev = new JevClient({
  apiKey: process.env.TYPESAFE_API_KEY,
  baseUrl: 'https://your-typesafe-compatible-endpoint.example',
  headers: { 'X-Application': 'catalog' },
})

Option	Purpose
apiKey	Authenticates the managed SDK client. Falls back to TYPESAFE_API_KEY; a blank key is rejected.
baseUrl	Overrides the SDK base URL.
headers	Default request headers.
client	Reuses an initialized TypeSafeClient.

Managed options and client injection are mutually exclusive, and combining them throws a TypeError.

Inject the official client

import { TypeSafeClient } from '@typesafe-ai/sdk'

const sdk = new TypeSafeClient({ apiKey: process.env.TYPESAFE_API_KEY! })
const jev = new JevClient({ client: sdk })

Use injection when native TypeSafe APIs and Anvia decisions should share credentials and transport settings. Managed clients disable SDK logging; injected clients retain their logging configuration.

Retries

The adapter disables SDK retries on every call, for managed and injected clients. Use decide({ retries }) or decideBatch({ retries }) so Core's retry policy is the only layer.

Model selection

const model = jev.decisionModel({ modelId: JEV_LATEST })

modelId is required and must be a non-empty string. JEV_LATEST is 'jev-latest'; other IDs remain valid.

Provider options

providerOptions on a decision request forwards extra JSON body fields. The adapter preserves model, state, and questions, so they cannot be replaced.

Runtime

The package is ESM, ships declarations, and depends on @typesafe-ai/sdk, which requires Node.js 20 or newer. Keep it server-side. Install a version of @anvia/core allowed by its peer range.