Your first QVAC app

import {
  loadModel,
  completion,
  unloadModel,
  close,
  LLAMA_3_2_1B_INST_Q4_0,
} from "@qvac/sdk";

let modelId;
try {
  // Downloads the model on the first run, then loads it into memory.
  modelId = await loadModel({
    modelSrc: LLAMA_3_2_1B_INST_Q4_0,
    modelType: "llm",
  });

  const run = completion({
    modelId,
    history: [{ role: "user", content: "In one sentence, why run AI locally?" }],
    stream: true,
  });

  for await (const token of run.tokenStream) process.stdout.write(token);

  const stats = await run.stats;
  console.log(`\n\n${stats?.tokensPerSecond?.toFixed(1)} tokens/s on ${stats?.backendDevice}`);
} finally {
  if (modelId) await unloadModel({ modelId });
  await close();
}

import { loadModel, ocr, unloadModel, close, OCR_LATIN } from "@qvac/sdk";

let modelId;
try {
  modelId = await loadModel({
    modelSrc: OCR_LATIN,
    modelConfig: { langList: ["en"] },
  });

  const run = ocr({ modelId, image: "sample.png", options: { paragraph: false } });
  const blocks = await run.blocks;

  for (const block of blocks) console.log(block);
} finally {
  if (modelId) await unloadModel({ modelId });
  await close();
}