Hello. I am building a worker application which can create vectorizedb, insert/query the vectors, de

Hello. I am building a worker application which can create vectorizedb, insert/query the vectors, delete vectorizedb. Currently, we need to bind the vectorize db to the worker manually. How can I automate it so that new dbs can be created and vectors can be inserted without manual binding in wrangler.json?

BBob Hello. I am building a worker application which can create vectorizedb, insert/q...

beingmudit [WePlayDOS]•5/9/25, 6:05 AM

Maybe you can us restful APIs if binding is an issue and i don't think dynamic bindings are supported

BobOP•5/9/25, 6:08 AM

But there is a usage limit on http apis. My insert/query calls might be too many to handle in http endpoint. The only other option I see that I have to create dbs beforehand and bind them

Bbeingmudit [WePlayDOS]Maybe you can us restful APIs if binding is an issue and i don't think dynamic b...

BobOP•5/9/25, 6:11 AM

thanks for the clarification

beingmudit [WePlayDOS]•5/9/25, 6:12 AM

I see

beingmudit [WePlayDOS]•5/9/25, 6:13 AM

I was not aware that bindings and rest apis have different limits

BBob But there is a usage limit on http apis. My insert/query calls might be too many...

yevgen•5/9/25, 10:51 AM

We can increase the rate limit to 9999 calls per 5 mins on your account. Please clarify what usage limits you are concerned about.

Alex Izydorczyk•5/12/25, 5:41 AM

Is the instance (index?) of vectorize creating by AutoRag feature just an ordinary instance that can be directly edited / queried

Alex Izydorczyk•5/12/25, 5:42 AM

AutoRAG today doesn’t support metadata filtering — but can I just append metadata anyway and manually query the database with these outside of the AutoRAG endpoint?

Yyevgen We can increase the rate limit to 9999 calls per 5 mins on your account. Please ...

BobOP•5/14/25, 3:19 AM

insert part I can optimize but querying pattern can be unpredictable for my use case. I want to design the app with max possible throughput achievable. I am assuming that query via workers have much more throughput than http endpoints. So I will pre-create the indexes and assign them to my pipelines when needed so that I can bind the indexes in the worker before hand.

I hope there is no limit on number of vectorize bindings in a single worker unless I am reaching max worker size. I don't see any doc for that.

BBob insert part I can optimize but querying pattern can be unpredictable for my use ...

yevgen•5/14/25, 10:42 AM

you should be fine with binding many vectorize indexes in a single worker.

yevgen•5/14/25, 10:43 AM

Turned out we can only increase API rate limits for Enterprise customers, for self-serve customers we can't

ac•5/14/25, 2:31 PM

Is it still not possible to use vectorize bindings locally? I'm dumping envenv and it's completely missing, and running into errors with it. Even when I use --experimental-vectorize-bind-to-prod--experimental-vectorize-bind-to-prod it's failing (definitely have the right binding name etc)

ac•5/14/25, 2:33 PM

Weirdly, when I start up the dev server the Vectorize index shows up under [worker name] has access to the following bindings:[worker name] has access to the following bindings:. But when I try to use it in the code and when I dump env, it's not there

ac•5/14/25, 3:02 PM

Also, it works fine when running vite devvite dev but not when running

wrangler dev

wrangler dev

(I need wrangler dev because I am running two workers at once)

ac•5/14/25, 3:27 PM

Was able to get it working with the auxiliaryWorkersauxiliaryWorkers config through vite instead. Pretty broken with wrangler dev still but at least I'm unblocked. Thanks for coming to my TED talk everyone

felix_m•5/15/25, 1:05 PM

Quick question: For Vectorize billing calculation, is the total count of vectors is used? Or does namespace and metadata filtering reduced the number of billable Queried Vector DimensionsQueried Vector Dimensions for that request? I have multiple tenants with different sets of vector, so looking for an efficient way to manage these. Creating one index per tenant would work, but it seems to complicate the bindings a bit.
https://developers.cloudflare.com/vectorize/platform/pricing/

yinxingmaiming6409•5/16/25, 3:38 AM

Can't the maximum length of Vectorize be 2048?

Yyinxingmaiming6409 Can't the maximum length of Vectorize be 2048?

yevgen•5/19/25, 1:24 PM

Vectorize supports up to 1578 vector dimension

Yyevgen Vectorize supports up to 1578 vector dimension

yinxingmaiming6409•5/20/25, 3:13 AM

Can I use Vectorize using only the HTTP protocol? What is the speed limit for this access.

Yyinxingmaiming6409 Can I use Vectorize using only the HTTP protocol? What is the speed limit for th...

Viktor•5/20/25, 2:21 PM

Vectorize has a REST api -> https://developers.cloudflare.com/api/resources/vectorize/

Cloudflare API | Vectorize

Interact with Cloudflare's products and services via the Cloudflare API

Deda•5/20/25, 8:54 PM

I have a question, is it possible to make an API that takes 2 vectors from the database, and it returns a position of vector B in a table of similarity of vector A against other vectors in the database?

Deda•5/20/25, 8:54 PM

I hope the question is not confusing -.-

Deda•5/20/25, 8:56 PM

example response: Vector B is in position of 52 (meaning 52 vectors are more similar than vector B to the vector A)

Tony•5/21/25, 12:43 AM

Hi all! Quick question just about to evaluate Vectorize. Does Vectorise need a CF Worker to work? Or can I use it to connect from other clients? Thanks!

DDeda I have a question, is it possible to make an API that takes 2 vectors from the d...

Viktor•5/21/25, 9:02 AM

Not really, at least not in a clean way. You could query vector a with topK = 100 (maximum possible when not using metadata) and loop it until you find the id of vector b in there. Index 0 will be Vector A (as it's 100% equal to the one you queried), and index 99 will be the one with the least similarity. But, as said, that's not clean and only works if there are max. 98 more similar vectors then vector B

TTony Hi all! Quick question just about to evaluate Vectorize. Does Vectorise need a C...

Viktor•5/21/25, 9:02 AM

Vectorize has a REST api that does work outside workers -> https://developers.cloudflare.com/api/resources/vectorize/

Cloudflare API | Vectorize

Interact with Cloudflare's products and services via the Cloudflare API

nico escudero•5/21/25, 8:20 PM

Anyone have any repos/guides/videos of RAG pipelines being built fully on Clouldflare stack? Looking for some inspo

Dev•5/24/25, 2:46 AM

Hi, why can't we use BGE-M3 embedding model (@cf/baai/bge-m3) to create vector index?

Only these model are available to choose from: [ "@cf/baai/bge-small-en-v1.5", "@cf/baai/bge-base-en-v1.5", "@cf/baai/bge-large-en-v1.5", "openai/text-embedding-ada-002", "cohere/embed-multilingual-v2.0" ]

yinxingmaiming6409•5/24/25, 5:44 AM

When should we consider encapsulating the official SDK of C #?

DDev Hi, why can't we use BGE-M3 embedding model (@cf/baai/bge-m3) to create vector i...

yevgen•5/25/25, 11:09 AM

please ask in #workers-ai

Steak•5/25/25, 9:13 PM

Has anyone seen this Vectorize error?

VECTOR_QUERY_ERROR (code = 40006): invalid query vector, expected 1024 dimensions, and got 0 dimensions

My index is definitely set to 1024 dimensions, and I’m querying with a plain JS array of length 1024. Even a hardcoded array fails. Storage is under 600 vectors. Is this a known issue or am I missing something obvious?

Steak•5/25/25, 10:30 PM

I adjusted the query method call, and it worked.

7777777•5/29/25, 2:02 AM

I'm getting the following error locally

✘ [ERROR] Uncaught (in promise) TypeError: Cannot read properties of undefined (reading 'insert')

env.VECTORIZE.insert(vectors)

env.VECTORIZE.insert(vectors)

I'm running npm run dev --experimental-vectorize-bind-to-prodnpm run dev --experimental-vectorize-bind-to-prod and wrangler is showing

- Vectorize Indexes:
  - VECTORIZE: <MY_INDEX> [connected to remote resource]

- Vectorize Indexes:
  - VECTORIZE: <MY_INDEX> [connected to remote resource]

on start up?

zensin•6/1/25, 1:37 PM

vectorize.index.invalid_config - Dimensions must
  be in range: [32, 1536] [code: 3003]

vectorize.index.invalid_config - Dimensions must
  be in range: [32, 1536] [code: 3003]

no 30723072

yinxingmaiming6409•6/2/25, 3:19 AM

Is there a way to use CF's AI to extract image embeddings?

beingmudit [WePlayDOS]•6/3/25, 5:49 PM

Hey CF Team, is something off with vectorize indexing latencies? I am continously getting vectors not found via getByIds. Can anyone update the status?

beingmudit [WePlayDOS]•6/3/25, 5:50 PM

@yevgen @garvitg

yinxingmaiming6409•6/6/25, 10:01 AM

Can each vector database have a maximum of 5 million pieces of data?

yinxingmaiming6409•6/6/25, 10:27 AM

What type of data is ndjson insertion?

yinxingmaiming6409•6/6/25, 10:40 AM

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

yinxingmaiming6409•6/6/25, 10:41 AM

yinxingmaiming6409•6/6/25, 10:41 AM

Why can't I upload it?

Yyinxingmaiming6409 {'result': None, 'success': False, 'errors': [{'code': 1005, 'message': 'vectori...

Isaac McFadyen•6/6/25, 1:18 PM

Not sure if this is your specific issue, but the ID must be a string not a number: https://developers.cloudflare.com/vectorize/best-practices/insert-vectors/

IIsaac McFadyen Not sure if this is your specific issue, but the ID must be a string not a numbe...

yinxingmaiming6409•6/6/25, 2:40 PM

I want to know a few questions, can an index only store a maximum of 5 million data?

Yyinxingmaiming6409 I want to know a few questions, can an index only store a maximum of 5 million d...

Isaac McFadyen•6/6/25, 2:45 PM

See https://developers.cloudflare.com/vectorize/platform/limits/

IIsaac McFadyen See https://developers.cloudflare.com/vectorize/platform/limits/

yinxingmaiming6409•6/6/25, 2:52 PM

My translation is not quite accurate. It seems to mean that there is no limit on the frequency of HTTP requests?

IIsaac McFadyen Not sure if this is your specific issue, but the ID must be a string not a numbe...

yinxingmaiming6409•6/6/25, 4:10 PM

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

yinxingmaiming6409•6/6/25, 4:10 PM

yinxingmaiming6409•6/6/25, 4:10 PM

Where exactly is the data format wrong?

yinxingmaiming6409•6/6/25, 4:40 PM

Is there no example explaining how ndJson data format works? I really can't upload it

Hello. I am building a worker application which can create vectorizedb, insert/query the vectors, de

Similar Threads

Similar Threads

Similar Threads