I do need to read. In each request, I'm essentially checking if the value has reached a certain numb

Qquisi.do I do need to read. In each request, I'm essentially checking if the value has re...

H

So it is an increment, plus a read? What you could do is push to AE, then have a cron trigger that checks whether it went over, and pushes that state to KV

H

Hello, I’m Allie!•3/1/24, 6:34 AM

This has the added benefit of being comparatively scalable

H

Hello, I’m Allie!•3/1/24, 6:34 AM

Since you aren’t hammering the data store on every request

Original message was deleted

Q

quisi.doOP•3/1/24, 6:36 AM

Hmm. I don't see it in the limits docs but I hadn't considered throttle limits.

Qquisi.do Hmm. I don't see it in the limits docs but I hadn't considered throttle limits.

H

Hello, I’m Allie!•3/1/24, 6:37 AM

That one isn’t an imposed limit iirc, but just with how fast D1 can process a query, coupled with the fact that it isn’t multithreaded

HHello, I’m Allie!So it is an increment, plus a read? What you could do is push to AE, then have a...

Q

quisi.doOP•3/1/24, 6:38 AM

I don't believe I currently have access to DO on free tier. I imagine I'll inevitably require paid tier but right now I'm just building PoCs so the gated services kinda make that hard.

H

Hello, I’m Allie!•3/1/24, 6:38 AM

You could also use R2, if that is the issue

Q

quisi.doOP•3/1/24, 6:39 AM

Wrong reply, sorry.

Qquisi.do I don't believe I currently have access to DO on free tier. I imagine I'll inevi...

H

Hello, I’m Allie!•3/1/24, 6:39 AM

My method doesn’t require a DO?

H

Hello, I’m Allie!•3/1/24, 6:39 AM

All good

Q

quisi.doOP•3/1/24, 6:39 AM

I meant the quote above yours.

Q

quisi.doOP•3/1/24, 6:40 AM

I wish the limits and pricing for every service were side by side. Maybe I'll make a table for that just to keep my sanity.

Q

quisi.doOP•3/1/24, 6:45 AM

Trade off is more expensive for less latency and faster consistency, right?

Qquisi.do Trade off is more expensive for less latency and faster consistency, right?

H

Hello, I’m Allie!•3/1/24, 6:45 AM

The big thing is that KV scales basically infinitely

Q

quisi.doOP•3/1/24, 6:45 AM

I wonder if I should be concerned with D1 read latency on every request

Q

quisi.doOP•3/1/24, 6:46 AM

My wallet doesn't lmao

H

Hello, I’m Allie!•3/1/24, 6:46 AM

D1, since it is a single isolate somewhere, can only handle so many requests

Q

quisi.doOP•3/1/24, 6:46 AM

It's in us east IIRC

Qquisi.do It's in us east IIRC

H

Hello, I’m Allie!•3/1/24, 6:47 AM

I meant more, because it is only one isolate, it can at best handle a few thousand rps(and that is without all of the SQLite machinery running every query)

H

Hello, I’m Allie!•3/1/24, 6:48 AM

And the values aren’t cached either, so every request is backhauled to the database. You can remedy that yourself tho

H

Hello, I’m Allie!•3/1/24, 6:49 AM

This would also be fixed by read replicas, but those don’t exist yet

RRasel Rana Rocky `vitest` doesn't seem to work with `@cloudflare/itty-router-openapi`, there's a ...

D

Dragon•3/1/24, 7:09 AM

Hi @Rasel Rana Rocky - Do you have a sample project to get started with a simple server having - Router, D1 connectivity, CORS ?

C

cj•3/1/24, 7:27 AM

Is there a Cloudflare service I can use to render the canvas?

ScreenshotOne seems to be using Cloudflare, but it appears they route the call to their server to make the render.

I see there is a new service - Browser-Rendering. It has a waiting list. If I join today, how long will it take for me to get approved? I would like to launch the app this weekend.

ScreenshotOne · Built with Workers · Cloudflare Workers®

Render screenshots in one simple API call

SS0ns0fAnarchy Hi, I am using the cloudflare stack for an app, I use KV store but since we sent...

H

Hello, I’m Allie!•3/1/24, 10:17 AM

Yes it is, though it isn't guaranteed to be persisted for the TTL you give it

R

Rayane•3/1/24, 11:13 AM

Hello everyone,

Can someone explain me how I can have a median cpu time of 78ms while the limit is 10 ms? (I use Page Functions free)

A

anurag•3/1/24, 11:51 AM

Cloudflare Worker is waiting for the whole response before streaming the output to the client, which defeats the entire point of streaming. I don't know what I am doing wrong.

router.post("/v1/streaming", async (request, { OPENAI_API }) => {
    let { readable, writable } = new TransformStream();
    const client = new ChatOpenAI({
      openAIApiKey: OPENAI_API,
    });

    const stream = await client.stream(
      "Write a 2 sentence poem about a sunset."
    );
    let writer = writable.getWriter();
    const textEncoder = new TextEncoder();

    for await (const part of stream) {
      if (part.content.length) {
        // Assuming part.content is a string containing the text
        const words = part.content.split(" "); // Split the text into words
        for (const word of words) {
          const sseFormattedData = `data: ${word}\n\n`; // Format each word as SSE
          await writer.write(textEncoder.encode(sseFormattedData));
          console.log("🚀 - forawait - sseFormattedData:", sseFormattedData);
          await new Promise((r) => setTimeout(r, 1000)); // Wait for 1 second before sending the next word
        }
      }
    }

    writer.close();

    return new Response(readable, {
      // headers: {
      //   "Content-Type": "text/event-stream",
      //   "Cache-Control": "no-cache",
      //   Connection: "keep-alive",
      // },
    });
  })

router.post("/v1/streaming", async (request, { OPENAI_API }) => {
    let { readable, writable } = new TransformStream();
    const client = new ChatOpenAI({
      openAIApiKey: OPENAI_API,
    });

    const stream = await client.stream(
      "Write a 2 sentence poem about a sunset."
    );
    let writer = writable.getWriter();
    const textEncoder = new TextEncoder();

    for await (const part of stream) {
      if (part.content.length) {
        // Assuming part.content is a string containing the text
        const words = part.content.split(" "); // Split the text into words
        for (const word of words) {
          const sseFormattedData = `data: ${word}\n\n`; // Format each word as SSE
          await writer.write(textEncoder.encode(sseFormattedData));
          console.log("🚀 - forawait - sseFormattedData:", sseFormattedData);
          await new Promise((r) => setTimeout(r, 1000)); // Wait for 1 second before sending the next word
        }
      }
    }

    writer.close();

    return new Response(readable, {
      // headers: {
      //   "Content-Type": "text/event-stream",
      //   "Cache-Control": "no-cache",
      //   Connection: "keep-alive",
      // },
    });
  })

router.post("/v1/streaming", async (request, { OPENAI_API }) => {
    let { readable, writable } = new TransformStream();
    const client = new ChatOpenAI({
      openAIApiKey: OPENAI_API,
    });

    const stream = await client.stream(
      "Write a 2 sentence poem about a sunset."
    );
    let writer = writable.getWriter();
    const textEncoder = new TextEncoder();

    for await (const part of stream) {
      if (part.content.length) {
        // Assuming part.content is a string containing the text
        const words = part.content.split(" "); // Split the text into words
        for (const word of words) {
          const sseFormattedData = `data: ${word}\n\n`; // Format each word as SSE
          await writer.write(textEncoder.encode(sseFormattedData));
          console.log("🚀 - forawait - sseFormattedData:", sseFormattedData);
          await new Promise((r) => setTimeout(r, 1000)); // Wait for 1 second before sending the next word
        }
      }
    }

    writer.close();

    return new Response(readable, {
      // headers: {
      //   "Content-Type": "text/event-stream",
      //   "Cache-Control": "no-cache",
      //   Connection: "keep-alive",
      // },
    });
  })

router.post("/v1/streaming", async (request, { OPENAI_API }) => {
    let { readable, writable } = new TransformStream();
    const client = new ChatOpenAI({
      openAIApiKey: OPENAI_API,
    });

    const stream = await client.stream(
      "Write a 2 sentence poem about a sunset."
    );
    let writer = writable.getWriter();
    const textEncoder = new TextEncoder();

    for await (const part of stream) {
      if (part.content.length) {
        // Assuming part.content is a string containing the text
        const words = part.content.split(" "); // Split the text into words
        for (const word of words) {
          const sseFormattedData = `data: ${word}\n\n`; // Format each word as SSE
          await writer.write(textEncoder.encode(sseFormattedData));
          console.log("🚀 - forawait - sseFormattedData:", sseFormattedData);
          await new Promise((r) => setTimeout(r, 1000)); // Wait for 1 second before sending the next word
        }
      }
    }

    writer.close();

    return new Response(readable, {
      // headers: {
      //   "Content-Type": "text/event-stream",
      //   "Cache-Control": "no-cache",
      //   Connection: "keep-alive",
      // },
    });
  })

K

kian•3/1/24, 2:22 PM

Not currently.

DDragon Hi @Rasel Rana Rocky - Do you have a sample project to get started with a simple...

R

Rasel Rana Rocky•3/1/24, 3:57 PM

Not yet. I'm cooking a terraform solution. Once I'm done, I'm gonna make a boilerplate around the whole thing to kick start projects.

Nnwdles Hi there! I have two questions. 1. Is it possible to create a token with permis...

C

Cyb3r-Jak3•3/1/24, 4:02 PM

1. No, api tokens get all workers. 2. I believe the /aaa/ will only fire on domain.com/aaa

S

Sean Kerr•3/1/24, 4:08 PM

Is it possible to have multiple workers run on the same custom domain, or do I need to have one worker that handles routes for the entire domain? It seems when I try to deploy a second worker, it wants to link the new worker the the already existing custom domain and remove it from the old worker. If this is the case, what is the point of the Workers Routes section under the domain?

Nnwdles thank you, and also, sorry, it's formatting, I asked about `* / bbb/ *` and `* /...

C

Cyb3r-Jak3•3/1/24, 4:09 PM

I know the most specific route wins but not sure how it would be calculated. I can play around with it later today and see

K

kian•3/1/24, 4:13 PM

I don't think you can even have infix wildcards can you?

C

Chaika•3/1/24, 4:14 PM

There's a ton of docs on routes here: https://developers.cloudflare.com/workers/configuration/routing/routes/#validity

C

Chaika•3/1/24, 4:15 PM

Route patterns may not contain infix wildcards or query parameters. For example, neither
example.com/*.jpg
example.com/*.jpg
example.com/*.jpg
example.com/*.jpg nor
example.com/?foo=*
example.com/?foo=*
example.com/?foo=*
example.com/?foo=* are valid route patterns.
When more than one route pattern could match a request URL, the most specific route pattern wins. For example, the pattern
www.example.com/*
www.example.com/*
www.example.com/*
www.example.com/* would take precedence over
*.example.com/*
*.example.com/*
*.example.com/*
*.example.com/* when matching a request for
https://www.example.com/
https://www.example.com/
https://www.example.com/
https://www.example.com/. The pattern
example.com/hello/*
example.com/hello/*
example.com/hello/*
example.com/hello/* would take precedence over
example.com/*
example.com/*
example.com/*
example.com/* when matching a request for example.com/hello/world.

RRayane Hello everyone,👋 Can someone explain me how I can have a median cpu time of 78m...

R

Rayane•3/1/24, 5:08 PM

Up

RRayane Up ⬆️

C

Chaika•3/1/24, 5:14 PM

I would read this: How are CPU limits enforced in the free tier?

C

Chaika•3/1/24, 5:14 PM

tldr being you don't have enough requests to trigger it/the limit is loose and you get an initial quota

C

Chaika•3/1/24, 5:15 PM

if you get enough requests, and those requests all exceed the time, you'd start seeing errors

C

Chaika•3/1/24, 5:20 PM

(isolates in that is referring to how Workers run. When you send a request to a Worker, it hits a CF machine/metal, and creates an Isolate for your Worker if there isn't one already on that machine. You stay connected to the same machine thanks to keepalive, but otherwise routing is random/doesn't prefer metals where you already have isolates afaik, so with little requests over a decent period of time, you could exceed the limit regularly until you get enough requests/popular enough)
Not really abusable because if your service becomes popular/gets a consistent amount of requests you'll burn through the initial quota and have to stay within cpu time

HHello, I’m Allie!30 seconds of CPU time, and theoretically indefinite runtime, though it may be e...

J

Joy•3/1/24, 7:06 PM

Hey @HardlyWorkin'
is it 10ms or 30s of CPU time for free plan. Because in a blog I found it to be 30s for standard plan
Blog Link : https://blog.cloudflare.com/workers-pricing-scale-to-zero/

The Cloudflare Blog

New Workers pricing — never pay to wait on I/O again

Announcing new pricing for Cloudflare Workers, where you are billed based on CPU time, and never for the idle time that your Worker spends waiting on network requests and other I/O.

K

kian•3/1/24, 7:08 PM

10ms

K

kian•3/1/24, 7:08 PM

Standard pricing is part of the Workers Paid plan

I do need to read. In each request, I'm essentially checking if the value has reached a certain numb

Similar Threads

Similar Threads

Similar Threads