Using cache.put returns "Web Cache is not available in this context."

[benan789]

Bug report

• I confirm this is a bug with Supabase, not with my own application.
• I confirm I have searched the Docs, GitHub Discussions, and Discord.

Describe the bug

Using cache.put returns Web Cache is not available in this context. when the cache api example uses it:

edge-runtime/examples/cache-api/index.ts

Line 20 in e126e1f

await cache.put(url, response.clone());

To Reproduce

Follow this example: https://github.com/supabase/edge-runtime/blob/develop/examples/cache-api/index.ts

Expected behavior

Not sure if the example is suppose to work or if cache.put is disabled or if additional steps are needed.

[nyannyacha]

Hello @benan789

CacheStorage APIs are currently limited in use as they are only used for caching AI model data.

edge-runtime/ext/ai/js/onnxruntime/cache_adapter.js

Lines 63 to 78 in e126e1f

	// NOTE(kallebysantos): Full `Web Cache API` is not allowed so we disable all other methods. In a
	// scenario where `Web Cache` is free to use, the `ai: Cache Adapter` should act as request
	// middleware ie. add new behaviour on top of `next()`
	async function notAvailable() {
	// Ignore errors when `debug = true`
	if (internals, internals.bootstrapArgs.opts.debug) {
	return;
	}
	throw new Error("Web Cache is not available in this context");
	}
	CacheStoragePrototype.has = notAvailable;
	CacheStoragePrototype.delete = notAvailable;
	CachePrototype.delete = notAvailable;
	CachePrototype.put = notAvailable;

We expect that in the future there will be discussions about whether to enable these APIs, but at this time, we do not plan to do so.

cc @kallebysantos

[benan789]

I am trying to cache an AI model actually haha. But how does that work without put? Or do I need to enable it somehow?

[kallebysantos]

Hi guys sorry for my delay 😅

@benan789, like Nya said we did allow it for ai models but just under transformers-cache tag:

edge-runtime/ext/ai/js/onnxruntime/cache_adapter.js

Line 5 in e126e1f

const ALLOWED_CACHE_NAMES = ["transformers-cache"];

this cache adapter is part of our native transformers.js integration #446

---

I am trying to cache an AI model actually haha.

If you're working with onnx models you can easily use the rust onnx runtime support by supplying the model url as String buffer, example

By using the native rust ort it will add the caching as well optimise the execution performance

There's already an Inference API pull request that add better onnx support on top of our native APIs - but since isn't merged yet you can simply add it manually by using this JS sample