Use `@huggingface/inference` inside `@huggingface/widgets`

[coyotte508]

Can be done by adding "@huggingface/inference": "workspace:^" in the deps and running pnpm install

And then using @huggingface/inference everywhere :)

[kohsheen1234]

Hi @coyotte508, can I work on this

[coyotte508]

Hi @kohsheen1234 there are some validation hurdles (cc @mishig25) before this can be done

We first need a global validation utility (like #385) and you can follow the discussion there, we need to add a source of truth somewhere for schemas.

[coyotte508]

Actually instead of calling the task-specific endpints, maybe we can just call request and streamingRequest from @huggingface/inference (instead of the existing callInferenceApi)

That would already be an improvement, and we can add validation as a second step. If you're interested @kohsheen1234 and feel you can handle the adjusments needed :)

cc @mishig25 @SBrandeis for visibility

Note: possible conflicts with #384

[kohsheen1234]

Thank you @coyotte508 , I'll start working on this then :)

[julien-c]

the widgets are an internal codebase now