MODEL REGISTRY

import { ModelRegistry } from '@huggingface/transformers';

const task = 'feature-extraction';
const modelId = 'onnx-community/all-MiniLM-L6-v2-ONNX';
const modelOptions = { dtype: 'fp32' };

const files = await ModelRegistry.get_pipeline_files(task, modelId, modelOptions);
// Example: ['config.json', 'onnx/model.onnx', 'tokenizer.json', ...]

const metadata = await ModelRegistry.get_file_metadata(modelId, 'onnx/model.onnx');
console.log(metadata);

const cached = await ModelRegistry.is_pipeline_cached(task, modelId, modelOptions);
console.log(cached ? 'Ready offline' : 'Needs download');

await ModelRegistry.clear_pipeline_cache(task, modelId, modelOptions);

const dtypes = await ModelRegistry.get_available_dtypes(modelId);
// Example: ['fp32', 'fp16', 'q4', 'q4f16']

import { ModelRegistry, pipeline } from '@huggingface/transformers';

const task = 'feature-extraction';
const modelId = 'onnx-community/all-MiniLM-L6-v2-ONNX';
const modelOptions = { dtype: 'q8' };

const files = await ModelRegistry.get_pipeline_files(task, modelId, modelOptions);

const metadata = await Promise.all(
  files.map((file) => ModelRegistry.get_file_metadata(modelId, file))
);

const totalBytes = metadata.reduce((sum, item) => sum + (item?.size ?? 0), 0);
const totalMB = (totalBytes / 1024 / 1024).toFixed(2);

const cached = await ModelRegistry.is_pipeline_cached(task, modelId, modelOptions);
console.log({ fileCount: files.length, totalMB, cached });

const pipe = await pipeline(task, modelId, {
  ...modelOptions,
  progress_callback: (info) => {
    if (info.status === 'progress_total') {
      console.log(`Loading: ${info.progress.toFixed(1)}%`);
    }
  },
});

await pipe.dispose();

import { ModelRegistry, pipeline } from '@huggingface/transformers';

const task = 'text-generation';
const modelId = 'onnx-community/Qwen2.5-0.5B-Instruct';

const dtypes = await ModelRegistry.get_available_dtypes(modelId);
const preferred = dtypes.includes('q4') ? 'q4' : dtypes[0] ?? 'fp32';

const generator = await pipeline(task, modelId, { dtype: preferred });
// ... inference
await generator.dispose();

import { ModelRegistry } from '@huggingface/transformers';

await ModelRegistry.clear_pipeline_cache(
  'feature-extraction',
  'onnx-community/all-MiniLM-L6-v2-ONNX',
  { dtype: 'fp32' }
);

import { ModelRegistry, env, pipeline } from '@huggingface/transformers';

const task = 'feature-extraction';
const modelId = 'onnx-community/all-MiniLM-L6-v2-ONNX';
const modelOptions = { dtype: 'q8' };

const cached = await ModelRegistry.is_pipeline_cached(task, modelId, modelOptions);

if (!cached) {
  throw new Error('Model not cached yet. Connect once to download assets.');
}

env.allowRemoteModels = false;
const pipe = await pipeline(task, modelId, { ...modelOptions, local_files_only: true });