I suspect that some of the metadata you insert has string with `U+2028` in it withing a JSON string.

I suspect that some of the metadata you insert has string with

U+2028

U+2028

in it withing a JSON string. We have reports of this issue recently.

yevgenOP•6/9/25, 1:31 PM

I recommend testing vectors insert without metadata attached to confirm the issue.

Yyevgen I suspect that some of the metadata you insert has string with `U+2028` in it wi...

yinxingmaiming6409•6/9/25, 2:57 PM

Error code: 400 - {'result': None, 'success': False, 'errors': [{'code': 1005, 'message': 'vectorize.unknown_content_type'}], 'messages': []}

yinxingmaiming6409•6/9/25, 2:57 PM

He's still like this, it's too difficult.

yinxingmaiming6409•6/9/25, 2:58 PM

I also use the official SDK, and for the simplest data, I don't think there should be any errors unless there is a problem with the official SDK?

Yyevgen please read https://developers.cloudflare.com/vectorize/best-practices/insert-ve...

yinxingmaiming6409•6/9/25, 3:08 PM

How else can I investigate? Can you help me?

yinxingmaiming6409•6/9/25, 3:12 PM

Yyevgen I suspect that some of the metadata you insert has string with `U+2028` in it wi...

yinxingmaiming6409•6/9/25, 3:12 PM

Do you think my data is correct?

Original message was deleted

yinxingmaiming6409•6/9/25, 3:26 PM

I read an article saying that you want this format? Isn't ndjson this format?
I'll find you where the article is.

yinxingmaiming6409•6/9/25, 3:27 PM

yinxingmaiming6409•6/9/25, 3:27 PM

https://developers.cloudflare.com/vectorize/best-practices/insert-vectors/#examples

Cloudflare Docs

Insert vectors

Vectorize indexes allow you to insert vectors at any point: Vectorize will optimize the index behind the scenes to ensure that vector search remains efficient, even as new vectors are added or existing vectors updated.

yinxingmaiming6409•6/9/25, 3:28 PM

The file must be in newline-delimited JSON (NDJSON format): each complete vector must be newline separated, and not within an array or object.

Original message was deleted

yinxingmaiming6409•6/9/25, 3:30 PM

Because I only have one data and there is a line break at the end, right?

yinxingmaiming6409•6/9/25, 3:39 PM

I don't think there is any difference between me and the data format in the example. Can't you tell where the difference lies?

AAlex Graham You don't need to use NDJSON when you are using the SDKs (either in Python or No...

yinxingmaiming6409•6/10/25, 1:03 AM

I think he still needs a string type. How should I transmit it?

yinxingmaiming6409•6/10/25, 1:05 AM

yinxingmaiming6409•6/10/25, 1:05 AM

I feel like I misunderstood his comments. He only accepts str types

yinxingmaiming6409•6/10/25, 1:08 AM

The error has not changed,

yinxingmaiming6409•6/10/25, 1:34 AM

Has anyone really succeeded with SDK?

yinxingmaiming6409•6/10/25, 8:35 AM

Is billing based on the query vector dimension? Or the number of queries? For example, for my 512 dimensional vector, is querying once counted as 512?

Original message was deleted

yinxingmaiming6409•6/10/25, 9:18 AM

Got it. But I still can't upload data. What should I do?

yinxingmaiming6409•6/10/25, 9:19 AM

Please help me.

yinxingmaiming6409•6/10/25, 1:17 PM

Are you guys at work

yinxingmaiming6409•6/10/25, 3:37 PM

A problem that I haven't been able to solve for many days, who can help me? Is this a problem with the SDK or mine?

yinxingmaiming6409•6/10/25, 4:05 PM

https://stackoverflow.com/questions/79274252/vectorize-upstream-error-failed-to-parse-upsert-vectors-request-in-ndjson-form/79282741#79282741

yinxingmaiming6409•6/10/25, 4:05 PM

At present, it seems that there is an error in the SDK encapsulation

yinxingmaiming6409•6/11/25, 1:18 AM

@Leo

yinxingmaiming6409•6/11/25, 2:02 AM

yinxingmaiming6409•6/11/25, 2:05 AM

yinxingmaiming6409•6/11/25, 2:05 AM

yinxingmaiming6409•6/11/25, 2:05 AM

I currently feel that it is an internal error in the SDK. No one can use it.

Original message was deleted

yinxingmaiming6409•6/11/25, 5:27 AM

I am currently 90% certain that it is an issue with the official SDK, but I am not sure if it is due to my inability to use it. Could you please confirm if it is an issue with the SDK? If so, I can provide feedback to the official personnel for repair to avoid any inconvenience caused by future users

Original message was deleted

yinxingmaiming6409•6/11/25, 5:28 AM

I can request it myself without any problem. I made a simple demo with the same data format

yinxingmaiming6409•6/11/25, 9:55 AM

Why do API calls occasionally encounter 410 errors? Request once every second, which is not too fast.

yinxingmaiming6409•6/11/25, 9:55 AM

yinxingmaiming6409•6/11/25, 10:41 AM

Can't filter the score?

yinxingmaiming6409•6/14/25, 12:47 AM

yinxingmaiming6409•6/14/25, 12:48 AM

Is there no other way for me to search for more information? Can the same user only have a maximum of 1200 times in 5 minutes?

yinxingmaiming6409•6/14/25, 12:48 AM

What constitutes the same user?

Yyinxingmaiming6409 Click to see attachment

1984 Ford Laser•6/14/25, 1:12 AM

That's for REST APIs, if you're accessing Vectorize via a Worker the same limits don't apply

11984 Ford Laser That's for REST APIs, if you're accessing Vectorize via a Worker the same limits...

yinxingmaiming6409•6/14/25, 1:14 AM

Are you sure? How are workers restricted? Can you send it to me? I have never encountered any examples of workers. And I see that the indexes name bound to workers is fixed, but I have a dynamic name.

yinxingmaiming6409•6/14/25, 1:15 AM

If I invite one more person to manage, will it be an additional 1200 times?

yinxingmaiming6409•6/14/25, 2:21 AM

Who knows if workers limit the access speed of vectorize?

Yyinxingmaiming6409 Who knows if workers limit the access speed of vectorize?

1984 Ford Laser•6/14/25, 2:40 AM

Have you tried reading this?
https://developers.cloudflare.com/vectorize/platform/limits/
I think main limit you might be concerned about would be the max upsert per batch

Cloudflare Docs

Limits

The following limits apply to accounts, indexes and vectors (as specified):

Yyinxingmaiming6409 Are you sure? How are workers restricted? Can you send it to me? I have never en...

U.Senthil Kumar | Contributor•6/14/25, 6:13 AM

Workers have some restrictions like fixed binding names (no dynamic binding names at runtime). You can’t dynamically create bindings per user, but you can handle routing logic inside the Worker code. Adding a teammate doesn’t multiply cost, pricing is based on usage, not the number of collaborators.

11984 Ford Laser Have you tried reading this? https://developers.cloudflare.com/vectorize/platfor...

U.Senthil Kumar | Contributor•6/14/25, 6:15 AM

Yes, the main limit you'd hit is the max upsert batch size: it's capped at 1,000 vectors per batch when using Workers, and 5,000 via the HTTP API, So if you're sending large batches from a Worker, you'll need to chunk them into batches of ≤ 1,000 vectors.Also note index name limits (≤ 64 bytes), but chunking your upserts is the key concern.

11984 Ford Laser Have you tried reading this? https://developers.cloudflare.com/vectorize/platfor...

yinxingmaiming6409•6/14/25, 7:18 AM

No, I'm not worried about data insertion restrictions, I'm mainly concerned about queries. Because I am working on an on-site service software that takes photos and checks for similar images. So there may be more queries.

yinxingmaiming6409•6/14/25, 11:39 AM

Does anyone know that if vectorize is operated by workers, there is no speed limit?

beingmudit [WePlayDOS]•6/15/25, 9:28 AM

Hey Team, is there a way to delete older vectors in a single go or if I want to clean up my entire index.

Bbeingmudit [WePlayDOS]Hey Team, is there a way to delete older vectors in a single go or if I want to ...

U.Senthil Kumar | Contributor•6/15/25, 1:06 PM

Yes, you can delete older vectors or clean up the entire index using a batch deletedelete operation or by dropping and recreating the index.

I suspect that some of the metadata you insert has string with `U+2028` in it withing a JSON string.

Similar Threads

Similar Threads

Similar Threads