Releases: vercel/storage
@vercel/postgres-kysely@0.7.1
@vercel/blob@0.19.0
Minor Changes
-
d44bd3b: feat(blob): add retry to all blob requests
This change generalizes the way we request the internal Blob API. This moves api version, authorization, response validation and error handling all into one place.
Also this adds a retry mechanism to the API requests
@vercel/postgres-kysely@0.7.0
Minor Changes
- f70264e: Correct VercelPostgresDialect to return an adapter that reports that transactions are not supported
@vercel/blob@0.18.0
Minor Changes
-
dc7ba0e: feat(blob): allow inline content disposition for certain blobs
Once you use this new version, then most common medias won't be automatically
downloading but rather will display the content inline.Already uploaded files will not change their behavior.
You can reupload them if you want to change their behavior.Fixes #509
@vercel/blob@0.17.1
Patch Changes
- d4c06b0: chore(blob): fix types on client.put
@vercel/blob@0.17.0
Minor Changes
-
898c14a: feat(blob): Add multipart option to reliably upload medium and large files
It turns out, uploading large files using Vercel Blob has been a struggle for users.
Before this change, file uploads were limited to around 200MB for technical reasons.
Before this change, even uploading a file of 100MB could fail for various reasons (network being one of them).To solve this for good, we're introducting a new option to
put
andupload
calls:multipart: true
. This new option will make sure your file is uploaded parts by parts to Vercel Blob, and when some parts are failing, we will retry them. This option is available for server and client uploads.Usage:
const blob = await put('file.png', file, { access: 'public', multipart: true, // `false` by default }); // and: const blob = await upload('file.png', file, { access: 'public', handleUploadUrl: '/api/upload', multipart: true, });
If your
file
is a Node.js stream or a ReadableStream then we will gradually read and upload it without blowing out your server or browser memory.More examples:
import { createReadStream } from 'node:fs'; const blob = await vercelBlob.put( 'elon.mp4', // this works 👍, it will gradually read the file from the system and upload it createReadStream('/users/Elon/me.mp4'), { access: 'public', multipart: true }, );
const response = await fetch( 'https://example-files.online-convert.com/video/mp4/example_big.mp4', ); const blob = await vercelBlob.put( 'example_big.mp4', // this works too 👍, it will gradually read the file from internet and upload it response.body, { access: 'public', multipart: true }, );
Patch Changes
-
fd1781f: feat(blob): allow folder creation
This allows the creation of empty folders in the blob store. Before this change the SDK would always require a body, which is prohibited by the API.
Now the the SDK validates if the operation is a folder creation by checking if the pathname ends with a trailling slash.const blob = await vercelBlob.put('folder/', { access: 'public', addRandomSuffix: false, });
@vercel/blob@0.16.1
@vercel/kv@1.0.1
Patch Changes
- 44f84bd: Upgrade to latest upstash-redis: https://github.com/upstash/upstash-redis/releases
@vercel/blob@0.16.0
Minor Changes
- 26a2acb: feat(blob): throw specific error when service unavailable
@vercel/blob@0.15.1
Patch Changes
-
f9c4061: fix(blob): Enforce content-type on fetch requests during token generation
Before this change, we would not send the content-type header on fetch requests sent to your server during client uploads. We consider this a bugfix as it should have been sent before.
⚠️ If you upgrade to this version, and you're using any smart request body parser (like Next.js Pages API routes) then: You need to remove anyJSON.parse(request.body)
at thehandleUpload
step, as the body will be JSON by default now. This is valid for theonBeforeGenerateToken
andonUploadCompleted
steps.