feat(@huggingface/hub): adding scanCacheDir

[axel7083]

Description

Took https://github.com/huggingface/huggingface_hub/blob/main/src/huggingface_hub/utils/_cache_manager.py as heavy inspiration for the implementation.

This is a very basic implementation, which do not cover the deletion.

Related issue

Fixes #905

Usage

import { scanCacheDir } from "@huggingface/hub";

const result = await scanCacheDir();
console.log(result);

[coyotte508]

Thanks for the PR!

Just to be clear, is this PR intended to interact with the cache folder created by the python lib? (cc @Wauplin)

Because the JS lib does not (yet?) cache files locally

[axel7083]

Just to be clear, is this PR intended to interact with the cache folder created by the python lib?

Yes, and hopefully the js library will allow to download model in the cache later

[coyotte508]

Thanks for the PR!

Lots of comments but it's mostly on the exposed API, technically the PR is good :)

Also, a small section in packages/hub/README.md would be great, and a doc comment on the main functions

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[coyotte508]

Thanks for the feature!

cc @julien-c @Wauplin if you have any input before merging

[Wauplin]

Thanks @axel7083 for the work! The implementation looks good at first glance except a question related to paths. I would also add a warning in the docs to mention that the JS client doesn't use the cache system itself (should avoid some misunderstanding).

[Wauplin]

Looks good!

[coyotte508]

Thanks!