Cloudflare file hosting

Uses Workers KV to enable Cloudflare as a file host. MIT Licensed.

Pricing notice

You should know that Cloudflare Workers always runs in front the CF Cache. This means none of the files will be cached, and every request made to these files will count against your Workers quota and pricing ($0.50/million after 10 million).

If you use files >10mb then this script will split the file into multiple 10mb parts. Due to this, the script will incur multiple KV.get calls (one for file split detection, others per each 10mb part), meaning you will go through your included KV read quota quicker.

The $5/mo workers charge gets you 10 million free KV reads. You will exhaust your quota depending on the size of files you have. See this spreadsheet for the pricing quota usage.

Usage

Standalone

See the wiki.

Existing worker

the "get file response" method is provided as a library in order to make integrating this into existing workers projects simple.

The API is similar to that of the caches.default API where you call the function, then check if the return value is null, in which case you continue with other code, or a Response, in which case you would immediately return that response.

Use it:

npm install cloudflare-file-hosting

// ES2015 Modules / typescript
import CF_FILES from "cloudflare-file-hosting";

// require / javascript
let CF_FILES = require("cloudflare-file-hosting");

async function handleRequest(request) {
  // FILES_KV is a KV namespace dedicated to the static files
  let url = new URL(request.url)
  let _filesResponse = await CF_FILES.getFile(url.pathname, FILES_KV);
  if (_filesResponse) {
    return _filesResponse;
  }
  /* other code here for when the file wasn't found */
}

The standalone worker is a direct example of using this API, see workerCode.ts.

Uploading files

To upload files, the following environment variables must be set:

• CLOUDFLARE_API_TOKEN - the API token to use for uploading. Require permission Workers KV Storage:edit on the account that owns the namespace.
• CLOUDFLARE_ACCOUNT_ID
• CLOUDFLARE_ZONE_ID
• CLOUDFLARE_KV_NAMESPACE_ID - the KV namespace ID for where to upload files

With these set, run cfupload --path path/to/root. You might need to use node_modules/.bin/cfupload --path path/to/root instead depending on your PATH setup.

Limitations

Download speeds

Not all Cloudflare regions are optimal for large file downloads. A download from the ATL datacenter with a fiber connection within the same state (Georgia) obtained ~6 MB/s download speeds.

Argo smart routing likely won't have any effect due to the nature of Workers (Argo only routes CF <--> origin better).

Upload size and frequency limits

Your uploading may get rate limited by Cloudflare if you have a lot of data to upload and/or are frequently uploading files.

The Cloudflare API request limit is 5000/hour. Each file upload is 1 request and big files that are being split will incur a request per every ~10mb, so you can upload at most 12.5gb before exhausting the limit.

We cannot use bulk upload since it requires uploaded values be UTF-8 encoded strings, which isn't possible for binary data (base64 is not used due to it taking up memory in the worker, where there is a 128mb limit).

Cache

At the moment, responses under 10mb in size will be pulled from the Cloudflare cache if possible, but this will still incur a workers request charge (but not a KV get charge). The cache is on a per-datacenter basis. As far as I know, due to the fact that this uses streamed responses for >10mb files, we can't use the Cache API to prevent incurring charges for KV reads. This is a tradeoff to support files larger than 128mb.