Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.arkor.ai/llms.txt

Use this file to discover all available pages before exploring further.

​
SDK overview

The arkor package gives you three primitives that cover the typical project: define a trainer, wrap it in a project manifest, and let the CLI / Studio find it. 
// src/arkor/trainer.ts
import { createTrainer } from "arkor";

export const trainer = createTrainer({
  name: "support-bot-v1",
  model: "unsloth/gemma-4-E4B-it",
  dataset: { type: "huggingface", name: "arkorlab/triage-demo" },
  lora: { r: 16, alpha: 16 },
  maxSteps: 100,
});

// src/arkor/index.ts
import { createArkor } from "arkor";
import { trainer } from "./trainer";

export const arkor = createArkor({ trainer });
That is the whole shape arkor dev and arkor start discover.

​
What’s here

TopicWhat you do with it
createTrainerDefine a fine-tuning run: model, dataset, LoRA, hyperparameters.
createArkorWrap a trainer into the manifest the CLI reads.
DatasetSourceTell the trainer where data comes from (HuggingFace name or blob URL).
CallbacksHook into the run: forward metrics, evaluate checkpoints, send notifications.
inferCall the in-flight model from inside onCheckpoint.
Trainer controlstart, wait, cancel, abortSignal. Use these when running outside the CLI.

​
Reference

For full type signatures, all TrainerInput fields, the auxiliary helpers (runTrainer, readCredentials, etc.), and the public-vs-internal export contract, see the SDK Reference.