AI Kit – Core

👉 Read the full docs: ai.aidalinfo.fr

@ai_kit/core bundles the foundational building blocks of AI Kit: typed workflows (createStep, createMapStep, createWorkflow), agents (Agent), and telemetry utilities. Everything else (server kit, templates, CLI) builds on this package.

Installation

pnpm add @ai_kit/core zod
# or
npm install @ai_kit/core zod

zod is optional but enables automatic type inference via inputSchema / outputSchema.

Quick start

import { createStep, createWorkflow } from "@ai_kit/core";
import { z } from "zod";

const fetchWeather = createStep({
  id: "fetch-weather",
  inputSchema: z.object({ city: z.string().min(1) }),
  outputSchema: z.object({ forecast: z.string() }),
  handler: async ({ input }) => {
    // Call your API here
    return { forecast: `Sunny in ${input.city}` };
  },
});

export const weatherWorkflow = createWorkflow({
  id: "weather-line",
  description: "Minimal weather pipeline",
})
  .then(fetchWeather)
  .commit();

const run = await weatherWorkflow.run({ inputData: { city: "Paris" } });
console.log(run.status, run.result);

Agents & telemetry

@ai_kit/core also ships with:

• Agent – orchestrates model calls (OpenAI, Scaleway, …) and exposes custom tools.
• Telemetry helpers to wire Langfuse / OpenTelemetry (workflow.withTelemetry, run.watch(), run.stream()).

Check the documentation for advanced agent samples, Langfuse integration, and human-in-the-loop steps.

Transcription

@ai_kit/core includes model-agnostic audio transcription support, compatible with any OpenAI-compatible endpoint (Scaleway Whisper large v3, OpenAI whisper-1, etc.).

Three public primitives

| Export | Role | |---|---| | createTranscriptionModel(config) | Creates a TranscriptionModelV3 provider | | transcribe(options) | Standalone function: loads audio (path / URL / buffer), calls the model, returns the transcript | | createTranscriptionTool(model, options?) | Returns an AI SDK tool() to attach directly to an Agent |

createTranscriptionModel

import { createTranscriptionModel } from "@ai_kit/core";

const whisperModel = createTranscriptionModel({
  modelId: "whisper-large-v3",
  apiKey: process.env.SCALEWAY_API_KEY!,
  baseURL: "https://api.scaleway.ai/v1",
  providerName: "scaleway", // optional, used in logs
});

Supports any OpenAI-compatible /audio/transcriptions endpoint (response_format=verbose_json).

transcribe

import { transcribe } from "@ai_kit/core";

// From a file path
const result = await transcribe({
  model: whisperModel,
  audio: "/path/to/audio.wav",
  inputType: "path",         // "path" | "url" | "buffer" — auto-detected if omitted
  language: "fr",            // optional ISO-639-1 code
});

console.log(result.text);
// result.segments → [{ text, startSecond, endSecond }]
// result.language, result.durationInSeconds

audio accepts a file path, an http(s) URL, or a Buffer / Uint8Array. The inputType is auto-detected when omitted.

createTranscriptionTool — attach to an Agent

import { createTranscriptionModel, createTranscriptionTool } from "@ai_kit/core";
import { Agent } from "@ai_kit/core";
import { scaleway } from "@ai_kit/core";

const whisperModel = createTranscriptionModel({
  modelId: "whisper-large-v3",
  apiKey: process.env.SCALEWAY_API_KEY!,
  baseURL: "https://api.scaleway.ai/v1",
});

const agent = new Agent({
  name: "medical-assistant",
  model: scaleway("gpt-oss-120b"),
  tools: {
    transcribeAudio: createTranscriptionTool(whisperModel, {
      description: "Transcrit un enregistrement audio médical en texte",
    }),
  },
});

const result = await agent.generate({
  prompt: "Transcris ce fichier : /recordings/consultation.mp3",
});

The tool schema exposed to the LLM: audio (path / URL / base64), inputType, language.

Supported audio formats

flac, mp3, mp4, mpeg, mpga, m4a, ogg, wav, webm (identical to OpenAI Whisper).

Where does @ai_kit/server fit?

@ai_kit/server complements the core by adding:

• A ready-to-use HTTP server (Express/Fastify) to expose your workflows / agents.
• Streaming endpoints (/runs/:id/stream), human-step resume handlers, supervisory APIs.
• Production guardrails (auth hooks, rate limiting, metrics).

Use them together:

• Define workflows, steps, and agents with @ai_kit/core.
• Install @ai_kit/server when you need HTTP/WebSocket exposure, centralized Langfuse telemetry, or a multi-workflow orchestrator.

Useful links

• Docs: https://ai.aidalinfo.fr
• Full examples (workflows + agents): packages/create-ai-kit/templates/*
• Questions/issues: open a ticket on the main repository.