Are there any suggestions to go around this? I want to create an index with 1M vectors which will h

Are there any suggestions to go around this?

I want to create an index with 1M vectors which will have around 10K namespaces and each query only needs one subset. If entire index's dimensions are used, we would end up with a huge bill.

Llavisht22 Are there any suggestions to go around this? I want to create an index with 1M ...

Matt Silverlock•3/31/25, 12:05 PM

If you store 1M vectors of 1536 dims and query it 1M times a month, it’s about ~$47 - https://developers.cloudflare.com/vectorize/platform/pricing/#usage-examples

Cloudflare Docs

Pricing · Cloudflare Vectorize docs

Vectorize bills are based on:

MMatt Silverlock If you store 1M vectors of 1536 dims and query it 1M times a month, it’s about ~...

lavisht22OP•3/31/25, 12:35 PM

I think I did my math wrong, let me check this again. Thanks!

lavisht22OP•4/2/25, 10:41 AM

I have an index with almost 1M vectors. I have also created a metadata index on one of the properties. When i query this index with a metadata filter, I am getting back results in 2-3 seconds, sometimes more. Is this kind of latency normal with Vectorize or am I doing something wrong?

lavisht22OP•4/2/25, 10:42 AM

I have the same vectors in postgres with pgvector and I was getting similar performance, wanted to try out vectorize to check if there are any performance gains.

Marius B.•4/2/25, 11:19 AM

Is cloudflare currently the most affordable provider for vector databases?

Llavisht22 I have an index with almost 1M vectors. I have also created a metadata index on ...

yevgen•4/4/25, 8:38 PM

this is for cold starts, it's get much better once you start querying data more actively (have more load, more customers). We do in-memory and CDN caching of vectors data for some period of time.

JJonathan Harel Hey team! Is there a way to delete vectors by namespace + metadata and not by ID...

yevgen•4/4/25, 8:38 PM

not at the moment. It's on the roadmap.

MMarius B.Is cloudflare currently the most affordable provider for vector databases?

yevgen•4/4/25, 8:38 PM

We hope so.

jedi•4/4/25, 9:29 PM

hi guys i'm curious about the pricing models. so the pricing model is based on stored and queried vector dimensions. so what if i use metadata filtering so i'm not querying against the whole vdb but rather a smaller set of data? what would the cost structure look like? is the queried vector dimensions is adjusted based on the queried space instead of the total space?

Neckville•4/7/25, 3:47 PM

is there the concept of semantic caching? What about caching for searches based on similarity

S.•4/9/25, 12:00 PM

Hey, is there any doc / link about using Cloudflare Vectorize with llamaindex ?

S.•4/9/25, 1:37 PM

AI AutoRag funnel led to error page

beingmudit [WePlayDOS]•4/9/25, 10:08 PM

Hi CF team, Has there been an issue with Vectorize lately? The indexing is taking more than 5 minutes to be able to query. I know it is an async process but this is a huge time. Can any one provide more details on this?

Bbeingmudit [WePlayDOS]Hi CF team, Has there been an issue with Vectorize lately? The indexing is takin...

yevgen•4/10/25, 9:59 AM

Hi @beingmudit we are seeing increased Vectorize usage since AutoRAG launch, we are scaling our infra for WAL processing to get the metrics back to normal.

yevgen•4/10/25, 10:26 AM

@beingmudit done, it should be under 30 seconds now. Thanks for reporting. Because it didn't cause any errors on our side, we haven't caught the delay in our monitoring.

beingmudit [WePlayDOS]•4/10/25, 11:12 AM

Thanks @yevgen for the resolving this. Yeah I didnt find any issues in the status page as well.

Harshilm•4/12/25, 9:27 PM

@yevgen Is there any way to resync error files in vector db?

Stephen•4/13/25, 7:28 PM

Hi, is there a way of accessing each vecor to add metadata tags to each vector? Used standard R2 ingest but did not add any meta-data and cannot seem to recover a vector based upon id. ID appears to be random value (hash?) I see that each chunk returns the same basic format:
Chunk 4
Cosine Sim.
0.6107
Relevancy
0.0017
Metadata
450db445-638b-4bb9-a0e8-16af21562e23/291385e0-f9c0-4126-96b4-6b1e0c1dfab1/1997-011.pdf
Chunk 1
Cosine Sim.
0.6169
Relevancy
0.0006
Metadata
450db445-638b-4bb9-a0e8-16af21562e23/291385e0-f9c0-4126-96b4-6b1e0c1dfab1/1997-008.pdf
Chunk
3
Cosine Sim.
0.6140
Relevancy
0.0019
Metadata
450db445-638b-4bb9-a0e8-16af21562e23/291385e0-f9c0-4126-96b4-6b1e0c1dfab1/1997-012.pdf

So the metadata appears to be the same apart from the file name and a chunk number. Is there a way of determining these values? I assume the first one is "processeduptoMutation ID" but not sure what the second value might be or perhaps the other way around. Need to understand how to recover this to be able to recover a specific set of vector values for a chunk.

Stephen•4/13/25, 7:30 PM

Or, a way of exporting the vectors from vector database without knowing the IDs ?

Nick•4/13/25, 8:03 PM

Hey all, I’m experiencing unexpectedly slow insert performance with Cloudflare Vectorize during a large-scale vector insertion. Over 12 hours, I successfully inserted about 2.5 million documents individually or in very small groups (1-2 vectors at a time). However, after about 36 hours, my process is still at around 1.9 million vectors total. It appears that Vectorize is batching inserts at about 1,000 vectors each, rather than the advertised batches of up to 200,000 vectors for improved throughput.

My understanding was that Vectorize would automatically batch inserts at these larger sizes to optimize performance, but this doesn’t seem to be happening. Do I need to explicitly batch my inserts (e.g., in groups of 5,000 vectors) to achieve better efficiency, or is there something else going on here?

Could anyone from Cloudflare clarify how batching works internally with Vectorize and suggest the best practices or architecture adjustments for optimizing large-scale vector insert operations?

Thanks so much

Ashish_S•4/14/25, 5:58 AM

delete vectors from AutoRAG vector store when file is deleted

We have a new AutoRAG projects and we regularly add new files and delete old files from the data store.
it was observed that when we delete some file from data rource i.e. R2 bucket. the vectors fro these files is still available in the vector store and search results return those vectors.

ideally we should have some way to delete or it should get deleted automatically.

Nick•4/15/25, 4:13 PM

Hi Vectorize team,

I’m really enjoying Vectorize—great work on building it! I’m running into issues though with metadata filtering on a number field.

Issue: I have two indexes with ~3M vectors each, with each vector representing a document. I have a metadata index on the field "authored" which contains a UNIX timestamp representing the date of authorship. Doing a query on date ranges like 1970-1979 or 1939-1955 consistently causes a 504 error (code 7009: upstream unavailable) after a long wait. Without the filter, queries work fine. Ranges of 5 years or less usually work, but not always.

What’s happening: It seems like filtering on number fields ($gt and $lt) with large result sets triggers a timeout.

Request: Could you investigate this? It’d be awesome to know any limits on metadata filtering and if this can be fixed in the future.

Thanks for your help and for making a solid product!

-Nick

AndrewA•4/15/25, 11:47 PM

Heya all my friend is looking to use vectorize to have an AI that can read from his bookstack instance. He mentioned vectorize to create the index. Does this make sense

AAndrewA Heya all my friend is looking to use vectorize to have an AI that can read from ...

yevgen•4/16/25, 12:54 PM

I recommend to start with AutoRAG https://blog.cloudflare.com/introducing-autorag-on-cloudflare/

The Cloudflare Blog

Introducing AutoRAG: fully managed Retrieval-Augmented Generation o...

AutoRAG is here: fully managed Retrieval-Augmented Generation (RAG) pipelines powered by Cloudflare's global network and powerful developer ecosystem. Simplify how you build and scale RAG pipelines to power your context-aware AI chatbots and search applications.

yevgen•4/16/25, 12:55 PM

much easier to get it up and running.

Mmatt Any roadmap for when max dimensions will be increased? I'd love to use text-embe...

Sora Morimoto•4/17/25, 9:17 AM

I would like to remind this again, but is there any update?

SSora Morimoto I would like to remind this again, but is there any update?

yevgen•4/17/25, 9:27 AM

Do you expect to have indexes under 1mln vectors with 3072 dimensions? Heavier vectors will be slower to insert, slower to train, and queries will be slower. There a lot of tradeoffs once we allow vectors with 3072 dimensions. It's not on the plans to support in Q2.

Ashley•4/18/25, 11:25 AM

Separately, you get diminishing returns with vector dimensions. It varies per model of course, but with AWS’ Titan Embeddings V2 for example found this:

We measured the accuracy of the vectors generated by Amazon Titan Text Embeddings V2 and we observed that vectors with 512 dimensions keep approximately 99 percent of the accuracy provided by vectors with 1024 dimensions. Vectors with 256 dimensions keep 97 percent of the accuracy. This means that you can save 75 percent in vector storage (from 1024 down to 256 dimensions) and keep approximately 97 percent of the accuracy provided by larger vectors.

I’d wager it’s a similar story for most models, but maybe your use case is different, but 3k dimensions is wildly expensive vs say 1k or 512. Maybe you know that already, and you need the 3k, but just a thought

Isaac McFadyen•4/18/25, 3:51 PM

^ A lot of new models support Matryoshka Learning which means even if they output a vector of 3096 or 2048 you can take the first 512 and the quality won't be degraded too much (often only a few percent like that article says)

Isaac McFadyen•4/18/25, 3:52 PM

Be careful though because some providers (e.g. OpenAI) requires you to specifically specify that you want to enable that setting (by requesting a certain vector length at embedding time)

Puliczek•4/18/25, 7:03 PM

Can I have some questions? 1) What's the best way to get all metadata from vectorize? For example i have 1000 topK i want to return 1000 metadata texts, right now limit is just 20 2) Is it good idea to have namespace per user, or i should go with separate vectorize per user? Thanks for help

Alexander•4/20/25, 4:03 AM

Hey, is it possible to trigger AutoRAG to re-index from the API? Couldn't find any way to do this?

AAlexander Hey, is it possible to trigger AutoRAG to re-index from the API? Couldn't find a...

Isaac McFadyen•4/20/25, 8:34 PM

?crossposting #workers-ai

SuperHelpflare•4/20/25, 8:34 PM

Please do not post your question in multiple channels/post it multiple times per the rules at #

welcome-and-rules. It creates confusion for people trying to help you and doesn't get your issue or question solved any faster.

Isaac McFadyen•4/20/25, 8:34 PM

For future use there's a dedicated #autorag channel

Shadow•4/21/25, 11:51 AM

Hey,
I’m encountering a persistent issue when querying a Cloudflare vector index using env.VECTORS.query() in a worker. Despite passing a valid 768-dimensional vector, I consistently receive the error.
VECTOR_QUERY_ERROR (code = 40006): invalid query vector, expected 768 dimensions, and got 0 dimensionsVECTOR_QUERY_ERROR (code = 40006): invalid query vector, expected 768 dimensions, and got 0 dimensions
I have validated the entries with:
console.log(“Vector length:”, queryVectorRaw.length); // 768console.log(“Vector length:”, queryVectorRaw.length); // 768
console.log(“typeof vector[0]:”, typeof queryVectorRaw[0]); // “number”console.log(“typeof vector[0]:”, typeof queryVectorRaw[0]); // “number”
console.log(“Array.isArray:”, Array.isArray(queryVectorRaw)); // trueconsole.log(“Array.isArray:”, Array.isArray(queryVectorRaw)); // true

Has anyone experienced this before?

CodingCoop | Co-Founder Nullshot•4/22/25, 12:15 PM

Any plans to get vectorize running locally?

CCodingCoop | Co-Founder Nullshot Any plans to get vectorize running locally?

Puliczek•4/22/25, 12:48 PM

you can debug prod vectorize with "wrangler dev --experimental-vectorize-bind-to-prod"

Viktor•4/22/25, 6:59 PM

Hey everyone, is there any chance to get the amount of queries via API (see screenshot)? Already looked in the rest api, but I might be blind or there is none

PPuliczek you can debug prod vectorize with "wrangler dev --experimental-vectorize-bind-to...

CodingCoop | Co-Founder Nullshot•4/23/25, 6:57 AM

Yeah i'm familiar @Puliczek - thanks for the share - but i was on an airplane coding the other day and realized vectorize for local dev ain't a thang.

IIsaac McFadyen ?crossposting #workers-ai

Alexander•4/24/25, 9:20 AM

My bad didn't know where to post, got'ya!

#autorag found it later

nukely•4/26/25, 6:37 PM

Getting "message": "vectorize.upstream_error - failed to parse insert vectors request in ndjson format: line Some(0) was not expected format"

ac•4/28/25, 10:59 PM

Is it possible to rename a Vectorize index?

Yyevgen @beingmudit done, it should be under 30 seconds now. Thanks for reporting. Becau...

beingmudit [WePlayDOS]•4/29/25, 4:26 AM

Hey CF team, is something again wrong with Vectorize indexing? Seems like nothing gets returned even after a min. Can anyone confirm the status?

beingmudit [WePlayDOS]•4/29/25, 4:37 AM

I found an issue in my similarity pipeline that started like a day ago. The issue is now, the similarity scores have been downleveved significantly - for eg, earlier I was getting the score of 0.7 for a few content pieces and now the processing of the same content again are sending the score of ~ 0.4 hence my check is not returning results. Has this behaviour been changed recently?

nukely•5/1/25, 3:08 PM

which vector search algo is used in vectorize

Legit•5/6/25, 5:15 PM

hi does vectorize behave in a similar way to r2? in the sense that we can use preview_index_name in the wrangler.toml? just like preview_bucket_name?

thipperz•5/7/25, 4:37 PM

First time reading about precise scores when querying. How much would be the latency increase?

Are there any suggestions to go around this? I want to create an index with 1M vectors which will h

Similar Threads