Cloudflare Developers•2y ago

Try MMR search - I think what you are looking for is diversity in results

Ppkpio Try MMR search - I think what you are looking for is diversity in results

scotto•4/24/24, 11:35 AM

cloudflare vectorize supports inserting multiples vectors to the same id?

Sscotto cloudflare vectorize supports inserting multiples vectors to the same id?

pkpioOP•4/24/24, 1:21 PM

Don't know

Sscotto cloudflare vectorize supports inserting multiples vectors to the same id?

Saqib•4/24/24, 4:11 PM

I think it does not allow.

Wang•4/25/24, 11:45 AM

I suddenly realized that the functionality of my product is similar to AWS Bedrock's knowledge base management , and I don't have confidence in being able to compete with AWS. I want to know if it's possible for me to join Cloudflare and do the same thing?:blobhaj:

Wang•4/25/24, 11:47 AM

I use the technology stack of Cloudflare Pages for full-stack development, Cloudflare D1, Cloudflare AI, Cloudflare R2, as well as Vectorize for knowledge training and chatbot construction.

Wang•4/25/24, 11:47 AM

:dogekek:

scotto•4/25/24, 12:46 PM

i'm getting an empty 200 response from vectorize rest API, index has data

scotto•4/25/24, 12:46 PM

no errors just an empty response

scotto•4/25/24, 1:02 PM

how to open a support ticket here? i can send a video, i can reproduce the problem quite easily

Sscotto how to open a support ticket here? i can send a video, i can reproduce the probl...

Isaac McFadyen•4/25/24, 1:07 PM

?support

Flare•4/25/24, 1:07 PM

To contact Cloudflare Support about an issue, please visit the Support Portal and fill in the form on the portal. After submission, you will receive confirmation over email.

Some issues, such as Account or Billing related issues, cannot be solved by the community.
Any plan level (including Free plans) can open tickets for Account, Billing or Registrar ticket categories. Make sure to select the correct category to ensure it goes to the right place.
For more information on the methods by which you can contact Support for your plan level, see Contacting Cloudflare Support - Cloudflare Docs

Sscotto i'm getting an empty 200 response from vectorize rest API, index has data

Isaac McFadyen•4/25/24, 1:16 PM

Can you provide more info about the response you're getting? If it's a widespread issue we might be able to escalate here.

Isaac McFadyen•4/25/24, 1:16 PM

Are you getting an empty array in the response? An empty 200 body? Something else?

Isaac McFadyen•4/25/24, 1:16 PM

And what is your Vectorize index dimensions and type (i.e. cosine, etc)?

scotto•4/25/24, 7:38 PM

it's possible to see the amount of vectors inside a index?

Sscotto it's possible to see the amount of vectors inside a index?

thipperz•4/25/24, 8:06 PM

such basic stuff right

thipperz•4/25/24, 8:06 PM

i'm hoping to see a lot of things when Vectorize goes GA

Bram Adams•4/29/24, 3:38 PM

would really love any kind of way to introspect anything about the vector db. a ui similar to supabase where i could see the metadata at least would go a really long way!

Sweetbasil•4/29/24, 5:42 PM

Hi Cloudflare team
Is Vectorize database HIPPA compliant?

antonioballs•4/29/24, 11:55 PM

Any update on the unreliability of querying Vectorize via REST API, @cfnathan ? Discussed here: Can you provide more info about the. Can provide more info you need. Cheers.

Discord

Discord - A New Way to Chat with Friends & Communities

Discord is the easiest way to communicate over voice, video, and text. Chat, hang out, and stay close with your friends and communities.

Bonadio•5/1/24, 5:07 AM

Vectorize seems very slow. I have a simple RAG application that uses vectorize with an index of size 1536 with cosine, with a single document of 5 pages , I was timing the code and vectorize takes almost 2 seconds to respond.
Here are my timing
(log) Embedding time 378
(log) Vectorize time 1844
(log) Document time 171
(log) LLM time 1518

the code is this

        const vectorizeStartTime: any = new Date()
        const queryVec = await c.env.VECTORIZE_TESTE_1536.query(embQuery, { 
            topK: 3, 
            returnValues: false, 
            returnMetadata: true, 
            namespace: data.customerId.toString()+"_"+data.repoId.toString()
        });
        const vectorizeEndtime: any= new Date()
        console.log("Vectorize time", vectorizeEndtime - vectorizeStartTime)

        const vectorizeStartTime: any = new Date()
        const queryVec = await c.env.VECTORIZE_TESTE_1536.query(embQuery, { 
            topK: 3, 
            returnValues: false, 
            returnMetadata: true, 
            namespace: data.customerId.toString()+"_"+data.repoId.toString()
        });
        const vectorizeEndtime: any= new Date()
        console.log("Vectorize time", vectorizeEndtime - vectorizeStartTime)

Is this the expected speed?

Bonadio•5/1/24, 5:11 AM

If I run the same code using chromadb in my local machine it only takes 102 milliseconds to respond

nuno•5/2/24, 11:24 AM

Hi guys, is it possible to query a vector and filter by multiple namespaces?

Ccfnathan This is a symptom of how we have the beta deployed. As mentioned in the above th...

Bonadio•5/3/24, 1:23 AM

Hi @cfnathan, thanks for the feedback.

Ccfnathan This is a symptom of how we have the beta deployed. As mentioned in the above th...

scotto•5/3/24, 10:29 PM

any ETA for this new version?

notorious_tom•5/3/24, 11:55 PM

hey @George Portillo just wanted to confirm I'm seeing the same issue with code that was working before :\

Jens•5/6/24, 7:40 PM

The pageContent text is not stored and is only used for embedding to create the vectors, right?

So that I have to save the text separately and cannot simply get it back later from the vector db.

const store = new CloudflareVectorizeStore(embeddings, {
      index: env.VECTORIZE_INDEX,
});
await store.addDocuments(
  [
    {
      pageContent: "hello",
      metadata: {},
    },
    {
      pageContent: "world",
      metadata: {},
    },
    {
      pageContent: "hi",
      metadata: {},
    },
  ],
  { ids: ["id1", "id2", "id3"] }
);

const store = new CloudflareVectorizeStore(embeddings, {
      index: env.VECTORIZE_INDEX,
});
await store.addDocuments(
  [
    {
      pageContent: "hello",
      metadata: {},
    },
    {
      pageContent: "world",
      metadata: {},
    },
    {
      pageContent: "hi",
      metadata: {},
    },
  ],
  { ids: ["id1", "id2", "id3"] }
);

Ccfnathan This is a symptom of how we have the beta deployed. As mentioned in the above th...

Jens•5/6/24, 8:38 PM

Where is the datacenter?

Original message was deleted

Jens•5/7/24, 10:03 AM

Yes thanks, I thought it was probably in the USA. I am Europa and the performance wasn't great. Now I tried something else it was faster.

Aim Salam•5/10/24, 1:47 AM

Quick question:

Is there any chance 'Maximum vectors per index' will be unlimited.

Wang•5/15/24, 2:19 AM

When can the GA version be released? :bibicat:

antonioballs•5/16/24, 10:49 AM

@cfnathan had a chance to try querying vectorize again and it now works about 30% of the time when testing from location outside US. The rest of the time the error is 'Request failed with status code 502'. statusText: 'Bad Gateway'. And some HTML is returned which includes a Cloudflare Ray ID: 884acb211000270d if that helps. Better but still unreliable

zensin•5/22/24, 1:01 PM

needs better embedding models once stable

Aantonioballs @cfnathan had a chance to try querying vectorize again and it now works about 30...

zensin•5/22/24, 1:01 PM

I'm curious if the issue you mentioned is still happening

Zzensin I'm curious if the issue you mentioned is still happening

antonioballs•5/23/24, 10:29 AM

Yes it's unusable at the moment. I've switched to Pinecone

scotto•5/25/24, 7:08 PM

Any news on stabilization from outside US?

imnotgoodatchoosingusernames•5/27/24, 8:06 PM

Hi! Just a quick question, but can we use namespaces with the LangChain Cloudflare Vectorize integration?

Isaac McFadyen•5/27/24, 10:33 PM

(for context: the message was removed)

Sscotto Any news on stabilization from outside US?

yevgen•5/28/24, 10:01 AM

The team is focused to release next version of Vectorize with 2 MM per index support end of June in private beta. In the meantime for folks experiencing issues with REST query API outside of US we recommend using Workers bindings which doesn't need to go through API Gateway service.

antonioballs•5/28/24, 2:35 PM

Hi @yevgen, do you mean service bindings, where we make a HTTP request to one worker which calls another worker that handles the requests to Vectorize?

Because I've tried that approach and still no data is returned from Vectorize. Let me know if I'm missing something.

Aantonioballs Hi @yevgen, do you mean service bindings, where we make a HTTP request to one w...

yevgen•5/28/24, 3:08 PM

yes, this approach removes API Gateway from the call sequence. I don't think there is anything specific with query Vectorize outside of US if done through workers bindings.

matt•5/29/24, 2:46 AM

anyone else getting "VECTOR_QUERY_ERROR (code = 1005): Internal Error: Internal error" ? was working yesterday.

matt•5/29/24, 4:06 AM

working now. must've been related to the D1 issues.

Yyevgen yes, this approach removes API Gateway from the call sequence. I don't think the...

antonioballs•5/29/24, 2:52 PM

Tried this approach and still doesnt work (from outside the US at least). No data is returned from Vectorize.

const vectorizeStartTime: any = new Date() const queryVec = await c.env.VECTORIZE_TESTE_1536.query(embQuery, { topK: 3, returnValues: false, returnMetadata: true, namespace: data.customerId.toString()+"_"+data.repoId.toString() }); const vectorizeEndtime: any= new Date() console.log("Vectorize time", vectorizeEndtime - vectorizeStartTime)

const store = new CloudflareVectorizeStore(embeddings, { index: env.VECTORIZE_INDEX, }); await store.addDocuments( [ { pageContent: "hello", metadata: {}, }, { pageContent: "world", metadata: {}, }, { pageContent: "hi", metadata: {}, }, ], { ids: ["id1", "id2", "id3"] } );

Try MMR search - I think what you are looking for is diversity in results

Similar Threads

Similar Threads

Similar Threads