Downloads & Retrieval

getUrl(path)

Returns a static public URL built from the driver's baseUrl (or an auto-derived default) plus the encoded path.

Pre-signed URLs

Both S3Driver and CloudinaryDriver can sign time-limited download URLs; only LocalDriver cannot (it has no access control to gate).

S3 — getPresignedUrl(path, { expiresIn }) returns a SigV4-signed GET URL — anyone with the link can download the object without credentials until it expires. Hand these out for private buckets instead of getUrl() results, which assume a publicly readable bucket.

const url = await client.getPresignedUrl("uploads/report.pdf", { expiresIn: 900 });
// https://bucket.s3.us-east-1.amazonaws.com/uploads/report.pdf?X-Amz-Algorithm=...

• expiresIn defaults to 3600 seconds (1 hour) and must be an integer in 1..604800 (AWS's 7-day cap); invalid values throw a ValidationError.
• Signing is done entirely with Web Crypto + your configured credentials — no SDK, no network call. Only the host header is signed with an UNSIGNED-PAYLOAD hash, so the URL works in any browser or client.
• Works with endpoint/forcePathStyle too (MinIO, R2, GCS-S3), as long as the endpoint validates SigV4 presigning (MinIO does).

Cloudinary — getPresignedUrl(publicId, { expiresIn, deliveryType }) returns a signed delivery URL (/s--SIG--/ component + expires_at) using the same resource-type/version resolution chain as getUrl. The signature (first 8 chars of URL-safe base64 SHA-1 of the resource path + API secret) only restricts access for assets with authenticated or private delivery — pass deliveryType: "authenticated" for those. CloudinaryDriver.upload creates public upload assets, so the default signed URL resolves but doesn't restrict anything.

• expiresIn defaults to 3600 seconds and must be a positive integer; no documented maximum.
• Drivers that cannot presign (LocalDriver) omit the method — FilewayClient.getPresignedUrl throws a clear error in that case.

Download / read-back (get(path))

Every driver streams a stored file back as a WHATWG ReadableStream<Uint8Array> — the same type it accepts on upload:

const stream = await client.get("uploads/report.pdf"); // ReadableStream<Uint8Array>
const bytes = new Uint8Array(await new Response(stream).arrayBuffer());

The stream is pull-based with backpressure, so files are never buffered whole on the way down. It can be piped straight to a destination (stream.pipeTo(writable)); for binary files use .arrayBuffer()/.blob() rather than .text(), which would decode binary data as UTF-8 and corrupt it.