JavaScript SDKs

Packages

@inferada-ai/sdknpm

Typed client for every service endpoint: chat, responses, embeddings, rerank, decisions, models, PII, language, audio, OCR and images. No runtime dependencies.

@inferada-ai/ai-sdk-providernpm

Inferada provider for the Vercel AI SDK 7, including Decisions and typed processor options.

@inferada-ai/agent-kitnpm

Framework for Inferada agents. It includes the SDK.

All three are ESM-only and need Node.js 22 or later (agent-kit: 22.19 or later). Until 1.0, a minor release may contain breaking changes; each package's changelog lists them.

Call the API from TypeScript

npm install @inferada-ai/sdk

import { InferadaClient } from '@inferada-ai/sdk'

const inferada = new InferadaClient({
  baseUrl: process.env.INFERADA_API_URL ?? 'https://api.inferada.com',
  token: process.env.INFERADA_API_TOKEN!
})

// A model ID your organisation has configured (see inferada.models.list()).
const reply = await inferada.chat.completions.create({
  model: process.env.INFERADA_MODEL!,
  messages: [{ role: 'user', content: 'Hello!' }],
  inferada: { processors: ['anonymise'], language: 'en' }
})

Request and response bodies use the same snake_case fields as this API reference, so everything documented here applies unchanged. Options that only exist in the client, such as serviceId and signal, are camelCase. Failed requests throw an InferadaError with the status, the parsed body and the request ID.

Use the Vercel AI SDK

npm install @inferada-ai/ai-sdk-provider ai

import { createInferada } from '@inferada-ai/ai-sdk-provider'
import { generateText } from 'ai'

const inferada = createInferada({ apiKey: process.env.INFERADA_API_TOKEN })

const { text } = await generateText({
  model: inferada(process.env.INFERADA_MODEL!),
  prompt: 'Summarise this ticket in one sentence.',
  providerOptions: { inferada: { processors: ['anonymise'], language: 'en' } }
})

Inferada's request extension (processors, haystack, prompt shield, reasoning, service_id) goes in providerOptions.inferada. Response metadata, including usage.cost_eur, is in providerMetadata.inferada.

Build agents

@inferada-ai/agent-kit runs TypeScript on a schedule or from a webhook. See Agents for a complete example.

Coding agents

Every package ships an AGENTS.md for coding assistants. It explains which capability fits which problem, where each workload should run (local or external models), and which privacy processors to use.