Hello everyone, I am integrating the workerAI with Langchain4j to bring visibility to this great ass

Hello everyone, I am integrating the workerAI with Langchain4j to bring visibility to this great asset https://github.com/clun/langchain4j/tree/provider-workerai. Most of the features are their. I have question regarding the

text-to_Image

text-to_Image

text-to_Image

text-to_Image (1) a few times i got the "upstream too busy", (2) is there any plan to add the resolution in the input request (3) what would be great value for num steps... i can try and guess but explanation would help

Pato•11/27/23, 7:12 PM

hey im getting this error with wrangler

✘ [ERROR] Error while creating remote dev session: TypeError: fetch failed
      at fetch
 {
    cause: Error: connect ETIMEDOUT 2606:4700:3031::ac43:cb34:443
        at TCPConnectWrap.afterConnect [as oncomplete] (node:net:1495:16) {
      errno: -60,
      code: 'ETIMEDOUT',
      syscall: 'connect',
      address: '2606:4700:3031::ac43:cb34',
      port: 443
    }
  }

✘ [ERROR] Error while creating remote dev session: TypeError: fetch failed
      at fetch
 {
    cause: Error: connect ETIMEDOUT 2606:4700:3031::ac43:cb34:443
        at TCPConnectWrap.afterConnect [as oncomplete] (node:net:1495:16) {
      errno: -60,
      code: 'ETIMEDOUT',
      syscall: 'connect',
      address: '2606:4700:3031::ac43:cb34',
      port: 443
    }
  }

Pato•11/27/23, 7:12 PM

it worked fine like 5 minutes ago

Pato•11/27/23, 7:26 PM

anyone knows what might be going on? There`s nothing on cloudflares status page

CCedrick Hello everyone, I am integrating the workerAI with Langchain4j to bring visibili...

CedrickOP•11/28/23, 1:15 AM

https://github.com/langchain4j/langchain4j/pull/320

GitHub

Support for WorkerAI Provider by clun · Pull Request #320 · langcha...

Support for Worker AI
CloudFlare opened a Beta of a new service Called WorkerAI
https://developers.cloudflare.com/workers-ai/
They do provide a REST API for different operations.

Text Generation
T...

gonz•11/28/23, 10:53 AM

Hi all. Say I have a zone called foo.bar.comfoo.bar.com and a workers route for *.foo.bar.com/**.foo.bar.com/* that goes to my worker, plus a proxied DNS record for *.foo.bar.com*.foo.bar.com. All that works fine. Then we decide to create a zone for bar.combar.com and migrate off foo.bar.comfoo.bar.com into this zone. Added the same proxied DNS record to this bar.combar.com zone and the same workers route. Issue we're seeing is that the requests seem to go straight to the origin server but they don't seem to be hitting the worker at all.

Any ideas/thoughts?

Correction: it seems to break whenever I remove the workers route from the old foo.bar.comfoo.bar.com zone. I want the new bar.combar.com to process these requests, not the old zone

Fun Miles•11/28/23, 1:44 PM

What is the CPU limit accounting when using a bound worker with the new billing system? Does the worker called with fetch get counted as part of the caller's CPU usage, or does it get its own limit?

Fun Miles•11/28/23, 1:46 PM

If it were called via the network, it would get its own limit and the caller time would pause while awaiting for the result of the fetch, right? Is it similar for a bound worker call?

liam | helius.dev•11/28/23, 8:34 PM

Are there any plans to reduce the cold start latency of workers? Its quite bad right now

Original message was deleted

liam | helius.dev•11/28/23, 8:43 PM

Do you have any idea how much this can help/what target cold start time is? Articles like this even claim there is zero cold start, but in practice I see anywhere between 50-500ms

liam | helius.dev•11/28/23, 8:50 PM

Spot checking response times from different locations, comparing first to subsequent requests. I had a lengthy convo with a support engineer a while ago but he wasn't able to figure out why with worker was impacted in this way, but without wasn't, so I just assume its something weird in the worker runtime

Lliam | helius.dev Do you have any idea how much this can help/what target cold start time is? Arti...

Hello, I’m Allie!•11/28/23, 8:56 PM

Basically, below 1 MiB you use Quicksilver, the DB that supports Zero Cold Starts. Above 1 MiB, you run from Cold Storage, which does not

HHello, I’m Allie!Basically, below 1 MiB you use Quicksilver, the DB that supports Zero Cold Start...

liam | helius.dev•11/28/23, 8:57 PM

Got it, do you have any sense of the latency distribution for running off of quicksilver vs cold storage?

Lliam | helius.dev Got it, do you have any sense of the latency distribution for running off of qui...

Hello, I’m Allie!•11/28/23, 8:58 PM

It should just be an instant drop off, like you are observing

Hello, I’m Allie!•11/28/23, 8:58 PM

If you get it below the limit, then it should just be 0(minus TLS start/other stuff)

HHello, I’m Allie!It should just be an instant drop off, like you are observing

liam | helius.dev•11/28/23, 9:00 PM

When I deploy this is what I see, is the limit based on compressed or uncompressed? Seems its already under but maybe something is still sending it to cold storage since its close?

Total Upload: 957.07 KiB / gzip: 182.93 KiB

Total Upload: 957.07 KiB / gzip: 182.93 KiB

Lliam | helius.dev When I deploy this is what I see, is the limit based on compressed or uncompress...

Hello, I’m Allie!•11/28/23, 9:00 PM

Compressed. Just to be sure, are you paying for Workers atm?

HHello, I’m Allie!Compressed. Just to be sure, are you paying for Workers atm?

liam | helius.dev•11/28/23, 9:01 PM

yeah, enterprise and unbound

Lliam | helius.dev yeah, enterprise and unbound

Hello, I’m Allie!•11/28/23, 9:06 PM

Weird…

Hello, I’m Allie!•11/28/23, 9:06 PM

Yeah, not sure what else could be causing that

Chaika•11/28/23, 9:10 PM

The other thing is about how your worker runs. If you use a route that specifies a specific path the magical cold start stuff can't work

Chaika•11/28/23, 9:10 PM

https://blog.cloudflare.com/eliminating-cold-starts-with-cloudflare-workers/

For now, this is only available for Workers that are deployed to a “root” hostname like “example.com” and not specific paths like “example.com/path/to/something.” We plan to introduce more optimizations in the future that can preload specific paths.

Edit: If it's more then ~50ms then it sounds more like global scope then cold start not working though, tcp connect isn't that long

Constantin•11/29/23, 1:54 AM

hey, i'm creating a minimal email worker that's making a post request to a slack webhook. all good and working. the problem is i'm missing email body. i was looking for it on

message.raw

message.raw

but is not there

CConstantin hey, i'm creating a minimal email worker that's making a post request to a slack...

Unsmart•11/29/23, 1:57 AM

https://github.com/edevil/email_worker_parser/blob/main/src/index.js

GitHub

email_worker_parser/src/index.js at main · edevil/email_worker_parser

An example of a worker that parses an email. Contribute to edevil/email_worker_parser development by creating an account on GitHub.

Constantin•11/29/23, 1:58 AM

aha, i see ty @Unsmart | Tech debt

amorfati•11/29/23, 2:36 AM

every time I deploy my worker the kv bindings is removed from my worker. I tried:

wrangler deploy --keep-vars

wrangler deploy --keep-vars

keep_vars = true
kv_namespaces = [
    { binding = "XXXXX", id = "XXXXXXX" },
]

keep_vars = true
kv_namespaces = [
    { binding = "XXXXX", id = "XXXXXXX" },
]

Constantin•11/29/23, 3:21 AM

does email workers are supported in wrangler?

Ddevnull I have an AI model I'd like to run, along side the llama & stable diffusion mode...

Isaac McFadyen•11/29/23, 4:21 AM

Not sure if it would be a matter of getting credit, but more that you can't run your own models on Workers AI yet.

Original message was deleted

liam | helius.dev•11/29/23, 5:40 AM

I tested with a simple proxy and same issue

CChaika The other thing is about how your worker runs. If you use a route that specifies...

liam | helius.dev•11/29/23, 5:41 AM

Does this apply if I have it at root but add a query string?

Also what do you mean by global scope being an issue?

masd4•11/29/23, 8:06 AM

Can the automatic running of cloudflare worker code while I'm typing be turned off? I can't work like this when I get basic auth popups every split second and I dont see the advantage of my worker constantly running on each keypress when I can use the Send button anytime I need to run

Mmasd4 Can the automatic running of cloudflare worker code while I'm typing be turned o...

Hello, I’m Allie!•11/29/23, 8:07 AM

You may be able to disable it in the settings, but otherwise, you can develop locally and push

masd4•11/29/23, 8:14 AM

I dont see an option for this but I think I may have a workaround.. I have to cause a syntax error and then it doesnt run constantly

masd4•11/29/23, 8:14 AM

nevermind it still runs

banbanboi•11/29/23, 8:16 AM

Hi, I am having problem with r2 audio/video file uploaded via presigned URL (PutObjectCommand). It was being uploaded but when we are going to use it via signedUrl(GetObjectCommand) for transcription, file cannot be read for transcription. The get presigned and signed URL are generated through Workers. I tried to download it and played in media player, and still it was unable to play the file. Did I missed something in upload using presigned url ?

but when I upload the file directly in R2, then transcribed the file via signedUrl, the transcription service that we are using are successfully able to transcribed it

Zetax•11/29/23, 5:39 PM

Hey,

I'm currently trying to minimize latency as much as possible. I am currently using Cloudflare hyperdrive with my postgres db + a cloudflare kv cache for my entries, but I am still getting around 100ms latency. Is there any way to reduce this further? I am just running a simple select query and my return size is about 4KB. I attached an image of what my route looks like, and an image of what the respective latency for each operation is. I also attached an image for my db middleware to allow me to use my db in ever call. I know this is not very effective, but almost all my routes use a DB call anyway. Or is there any way to reuse an existing connection?

Unsmart•11/29/23, 5:47 PM

I would remove the cache and remove the 2nd db connection that doesnt use hyperdrive and it should be faster

Unsmart•11/29/23, 5:50 PM

and if theres some reason you really dont want to use hyperdrive for a specific route I would make it where you return only the hyperdrive connection or only the direct connection instead of opening both

UUnsmart I would remove the cache and remove the 2nd db connection that doesnt use hyperd...

Zetax•11/29/23, 5:52 PM

Why remove the cache? It seems to be much faster. Just trying to understand

Unsmart•11/29/23, 5:52 PM

hyperdrive already adds a cache so that would just double cache now

UUnsmart hyperdrive already adds a cache so that would just double cache now

Zetax•11/29/23, 5:55 PM

True, but for some reason the Hyperdrive cache is way slower. It adds about 80ms of latency

UUnsmart hyperdrive already adds a cache so that would just double cache now

Zetax•11/29/23, 6:09 PM

Would cloudflare D1 be better for my use case? I just want the lowest latency possible

Isaac McFadyen•11/29/23, 6:54 PM

D1 would be lower latency in the location where it's created, but it's also SQLite (not Postgres) so anything Postgres specific won't work with D1.

Deo Kumar•11/29/23, 8:12 PM

I am using photon-rs using WebAssembly, and the worker throws

Error 1102: Workers Resources Limit

Error 1102: Workers Resources Limit

, I think it's due to CPU time limit. How can I check the cpu time taken by a fetch handler?

elegant-idn•11/29/23, 9:51 PM

Hi, everyone. Sorry for bothering you, but I need your help
I hosted a static website using Cloudflare workers to optimize the performance
how can I increase the page speed? what should I do, like configuring or anything else?

johnkush•11/29/23, 10:16 PM

I just switched to paid workers plan. Is there something like a safety limit or something, so I don't end up spending big money by a mistake, since it's paid on additional requests as-you-go?

Eelegant-idn Hi, everyone. Sorry for bothering you, but I need your help I hosted a static we...

Cyb3r-Jak3•11/29/23, 10:17 PM

If your site is purely static then I would recommend pages

Jjohnkush I just switched to paid workers plan. Is there something like a safety limit or ...

Unsmart•11/29/23, 10:26 PM

There's no hard limits you can impose, setting up WAF rules to try and block any bot traffic and rate limiting rules to disallow spam would be things you can do to try and ensure expected usage levels.

Normally you could also setup notifications for when you reach a certain request amount but seems due to the billing changes recently made to workers you cannot do this yet with the updated billing model.

crossbeau•11/29/23, 11:57 PM

NuxtJS SSR worker, my code is fetching my dotnet endpoint in azure using a cloudflare origin certificate. I have strict TLS turned on and my nuxt worker is getting 522 errors on request to dotnet endpoint. Do i need to somehow get worker to do an origin pull? If i turn off strict TLS worker goes back to functional

Hello everyone, I am integrating the workerAI with Langchain4j to bring visibility to this great ass

Similar Threads