// src/arkor/trainer.ts (scaffolded)
import { createTrainer } from "arkor";

export const trainer = createTrainer({
  name: "triage-run",
  model: "unsloth/gemma-4-E4B-it",
  dataset: { type: "huggingface", name: "arkorlab/triage-demo" },
  datasetFormat: { type: "chatml" },
  maxSteps: 100,
  evalSteps: 25,
  lora: { r: 16, alpha: 16, loadIn4bit: false },
  callbacks: {
    onLog: ({ step, loss, evalLoss }) => {
      const lossPart =
        typeof loss === "number" && Number.isFinite(loss)
          ? ` loss=${loss.toFixed(4)}`
          : "";
      const evalPart =
        typeof evalLoss === "number" && Number.isFinite(evalLoss)
          ? ` evalLoss=${evalLoss.toFixed(4)}`
          : "";
      console.log(`step=${step}${lossPart}${evalPart}`);
    },
  },
});

dataset: {
  type: "huggingface",
  name: "your-org/your-private-dataset",
  split: "train",        // 任意。データセットのデフォルト split
  subset: "v3",          // 任意。複数 subset を持つデータセット用
},

dataset: {
  type: "blob",
  url: "https://internal.example.com/data/2026-04.jsonl",
  token: process.env.DATASET_TOKEN,    // バックエンドの blob 取得時に転送（ワイヤーフォーマットはバックエンド側で定義）
},

createTrainer({
  // ...
  lora: {
    r: 32,                 // ランクが高いほどよく学べるがコストも上がる
    alpha: 64,             // しばしば 2 × r
    maxLength: 2048,       // 長いサンプルをこのトークン数で切り詰める
    loadIn4bit: true,      // QLoRA、約 4 倍のメモリ節約、わずかな品質コスト
  },
  maxSteps: 500,
  learningRate: 2e-4,
  batchSize: 8,
  weightDecay: 0.01,
  lrSchedulerType: "cosine",
});

createTrainer({
  // ...
  dryRun: true,
});

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,
  callbacks: {
    onLog: ({ step, loss, evalLoss }) => {
      const lossPart =
        typeof loss === "number" && Number.isFinite(loss)
          ? ` loss=${loss.toFixed(4)}`
          : "";
      const evalPart =
        typeof evalLoss === "number" && Number.isFinite(evalLoss)
          ? ` evalLoss=${evalLoss.toFixed(4)}`
          : "";
      console.log(`step=${step}${lossPart}${evalPart}`);
    },
    onCheckpoint: async ({ step, infer }) => {
      // /ja/cookbook/mid-run-eval を参照
      // 型付きフィールドの抽出は /ja/cookbook/structured-outputs を参照
    },
    onCompleted: async ({ job, artifacts }) => {
      // /ja/cookbook/notifications を参照
    },
    onFailed: async ({ job, error }) => {
      // /ja/cookbook/notifications を参照
    },
  },
});

model: "unsloth/gemma-4-E4B-it",   // 現在サポートされている値

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

const SLACK_WEBHOOK = process.env.SLACK_WEBHOOK_URL;

async function postSlack(text: string): Promise<void> {
  if (!SLACK_WEBHOOK) return;
  try {
    await fetch(SLACK_WEBHOOK, {
      method: "POST",
      headers: { "content-type": "application/json" },
      body: JSON.stringify({ text }),
    });
  } catch (err) {
    console.warn("slack post failed:", err);
  }
}

export const trainer = createTrainer({
  name: "support-bot-v2",
  model: "unsloth/gemma-4-E4B-it",
  dataset: { type: "huggingface", name: "arkorlab/triage-demo" },
  lora: { r: 32, alpha: 64, maxLength: 2048, loadIn4bit: true },
  maxSteps: 500,
  learningRate: 2e-4,
  callbacks: {
    onLog: ({ step, loss, evalLoss }) => {
      const lossPart =
        typeof loss === "number" && Number.isFinite(loss)
          ? ` loss=${loss.toFixed(4)}`
          : "";
      const evalPart =
        typeof evalLoss === "number" && Number.isFinite(evalLoss)
          ? ` evalLoss=${evalLoss.toFixed(4)}`
          : "";
      console.log(`step=${step}${lossPart}${evalPart}`);
    },
    onCompleted: async ({ job, artifacts }) => {
      await postSlack(`✓ ${job.name} done (${artifacts.length} artifact(s))`);
    },
    onFailed: async ({ job, error }) => {
      await postSlack(`✗ ${job.name} failed: ${error}`);
    },
  },
});