Wouldn't it be more cost effective to just keep a DO with a session alive perpetually? It costs just

A

Advany•9/28/23, 6:37 PM

I also dont get the pricing. Would be great to get insights

Hhannes Same question about 10cent session price.. is this per spawned session? How does...

U

Unsmart•9/28/23, 6:56 PM

The per session would be every time you call

puppeteer.launch()

puppeteer.launch()

puppeteer.launch()

puppeteer.launch(). You can reuse the session as explained with DOs in the blogpost which would make it where you can pay per minute with only the initial request warming up the DO costing or anytime the DO gets reset (should mostly code updates / runtime updates)

A

Advany•9/28/23, 7:07 PM

So then keeping do alive would be cheaper right?

U

Unsmart•9/28/23, 7:11 PM

depends on your usage levels

D

DanTheGoodman•9/29/23, 2:13 PM

Does puppeteer stealth work? (ping plz)

S

Sam•9/29/23, 3:43 PM

Hello. I'd like to get access to the browser rendering api. Account: 1448756027c02d2cb72a056a23538f40

S

Sam•9/29/23, 3:43 PM

Thank you

Hhannes It's still weird.. does that mean I only pay 10 cents + DO cost for a day (ca. 4...

M

MackenlyOP•9/29/23, 5:40 PM

From my understanding it would just be $0.10 every time the DO and browser session is instantiated as long as it's kept alive, so not even every day, as long as the browser instance gets a command at least every 60 seconds. If this is as it seems the pricing really doesn't make sense. Either there's an undocumented time limit on session time or the session pricing should be much lower to incentivize users to create new sessions rather than keep one open. Paying $0.10 once feels like a loophole.

If the best practice is to leave a session open, that should definitely be communicated in the pricing/docs since without considering using a DO, the pricing seems super high (who could justify $0.10 to take a screenshot when you can do the same thing with puppeteer on GCP functions or AWS for much much less).

U

Unsmart•9/29/23, 8:21 PM

requests != sessions

U

Unsmart•9/29/23, 8:23 PM

but also 50 sessions would only be $5

C

captain•9/30/23, 5:09 PM

hey guy's made the request to get into this a while back...i even run a server using pupperteer as part of a workflow process...would love to start playing with this...been developing in workers for 2 years...any updates now that its in beta in getting access to this?

D

DanTheGoodman•10/6/23, 4:27 PM

How does the concurrent session pricing work? Do I pay $1 to have the ability to run 2 sessions concurrently per month? (se then I can pay $5 for a concurrency of 6) Or is it $1 instead of $0.1 if there is already a session working? The blog post wasn't clear on that.

A

Aaron Peters•10/9/23, 11:57 AM

The pricing is confusing to me too.
What do I pay if I have one browser instance running all the time and it's constantly doing work (e.g. take screenshot of a page, one after the other .. all the time)
As I read it, that is one browser instance and one (concurrent) session, so the costs are 0 as the Workers paid subscription includes 500 Sessions and 10 Concurrent Sessions. Is this a correct understanding?

P

phillip•10/14/23, 4:56 PM

Hello, looking for access to the Browser Rendering API. Thanks

Pphillip Hello, looking for access to the Browser Rendering API. Thanks 🙂

Z

zegevlier•10/14/23, 5:41 PM

Post your account ID in here browser-rendering-apiBrowser Rendering API Beta Access

V

vassbence•10/15/23, 10:10 AM

Hi! Would love start using this API!

4dc10b1f5e6ded482cc85051fbdf49c9

V

vassbence•10/15/23, 10:11 AM

And so the recommended way to use this service is to keep browsers alive as much as possible with Durable Objects?

Vvassbence Hi! Would love start using this API! 4dc10b1f5e6ded482cc85051fbdf49c9

Z

zegevlier•10/15/23, 10:11 AM

Look one message above yours

V

vassbence•10/15/23, 10:12 AM

My bad! Posted in the correct thread. My question still stands

V

vassbence•10/15/23, 10:13 AM

Also, would love to know how one should handle large amounts of traffic, what would be the easiest way to build a queue

Vvassbence And so the recommended way to use this service is to keep browsers alive as much...

Z

zegevlier•10/15/23, 10:14 AM

AFAIK yes, how the pricing works exactly is still a bit unclear, but it seems like there is little harm in keeping sessions open as long as possible to re-use them.

Vvassbence Also, would love to know how one should handle large amounts of traffic, what wo...

Z

zegevlier•10/15/23, 10:14 AM

Probably #queues-beta

V

vassbence•10/15/23, 10:14 AM

Let's say I'm generating PDF of tickets, and an event goes live, 10k people purchase a ticket. I put each ticketId in the queue which gets consumed by a worker which is calling this API via a DO

V

vassbence•10/15/23, 10:14 AM

but that means there is literally 1 consumer for the queue running 1 request / sec (or worse if it takes longer)

V

vassbence•10/15/23, 10:15 AM

Which would be really bad

V

vassbence•10/15/23, 10:15 AM

Or I can actually fan out the work to multiple browsers, it's jut that it's going to cost a lot more?

Vvassbence Or I can actually fan out the work to multiple browsers, it's jut that it's goin...

Z

zegevlier•10/15/23, 10:18 AM

You can absolutely do that: Keep in mind that the current limits for the beta only allow 2 concurrent browsers (https://developers.cloudflare.com/browser-rendering/platform/limits/), and the current pricing (from https://blog.cloudflare.com/running-serverless-puppeteer-workers-durable-objects/) means that fanning out will increase the cost by quite a lot.

V

vassbence•10/15/23, 10:19 AM

Yeah... so to keep using Cloudflare workers i'd need to write logic to fan out to a minimal amount of browsers but keep them 100% working all the time

V

vassbence•10/15/23, 10:20 AM

AWS Lambda is just a better solution for this currently, at least for my use case, as each lambda can start their own puppeteer, i can easily call 10k lambdas and the pricing is a lot better.

V

vassbence•10/16/23, 5:04 PM

Is the API supposed to work in localhost?

V

vassbence•10/16/23, 5:05 PM

C

Chao•10/16/23, 6:51 PM

how much durable object storage does a browser object take up? Say for a browser without any pages opened?

T

Tom•10/19/23, 1:16 PM

Hello all - how can I get access to the beta? I'm looking to replace prerender.io in our stack, and would love to use Browser Rendering API. We already love using pages, workers and the new Workers AI too

TTom 👋 Hello all - how can I get access to the beta? I'm looking to replace prerende...

Z

zegevlier•10/19/23, 1:16 PM

Put your account ID here browser-rendering-apiBrowser Rendering API Beta Access

Zzegevlier Put your account ID here https://discord.com/channels/595317990191398933/1112675...

T

Tom•10/19/23, 1:20 PM

Will do, thanks!

T

Tom•10/19/23, 2:00 PM

fyi - The docs here (https://developers.cloudflare.com/browser-rendering/platform/wrangler/#bindings) mention

node_compat

node_compat

but in the getting started example (https://developers.cloudflare.com/browser-rendering/get-started/screenshots/#4-configure-wranglertoml) its

compatibility_flags = [ "nodejs_compat" ]

compatibility_flags = [ "nodejs_compat" ]

A

Ashley•10/20/23, 11:48 AM

how long does it take to spin up a new browser instance? I am running this code:

    const browser = await puppeteer.launch(env.BROWSER);
    const page = await browser.newPage();

    const [res] = await Promise.all([
        page.waitForResponse(res => res.url().includes('something') && res.request().method().toUpperCase() !== "OPTIONS", {timeout: 10 * 1000}),
        page.goto('https://someurl.com', {waitUntil: "domcontentloaded"}),
    ]);

    const browser = await puppeteer.launch(env.BROWSER);
    const page = await browser.newPage();

    const [res] = await Promise.all([
        page.waitForResponse(res => res.url().includes('something') && res.request().method().toUpperCase() !== "OPTIONS", {timeout: 10 * 1000}),
        page.goto('https://someurl.com', {waitUntil: "domcontentloaded"}),
    ]);

    const browser = await puppeteer.launch(env.BROWSER);
    const page = await browser.newPage();

    const [res] = await Promise.all([
        page.waitForResponse(res => res.url().includes('something') && res.request().method().toUpperCase() !== "OPTIONS", {timeout: 10 * 1000}),
        page.goto('https://someurl.com', {waitUntil: "domcontentloaded"}),
    ]);

    const browser = await puppeteer.launch(env.BROWSER);
    const page = await browser.newPage();

    const [res] = await Promise.all([
        page.waitForResponse(res => res.url().includes('something') && res.request().method().toUpperCase() !== "OPTIONS", {timeout: 10 * 1000}),
        page.goto('https://someurl.com', {waitUntil: "domcontentloaded"}),
    ]);

And locally, on my Mac, it takes 2-3 seconds in total. In Cloudflare however, it seems to be taking a lot longer perhaps? I don't really know, but it's getting exceeded resources and the log tail says CPU exceeded. It says for CPU, but again, I can't see how one request like this would use so much CPU... it's listed as using 75ms of CPU time, so I'm just really confused!

A

Ashley•10/20/23, 12:16 PM

This makes it even more confusing

Varying CPU times, some exceeded, some not

K

Kuririn•10/20/23, 4:21 PM

How can I check if I have access to the beta?

K

Kuririn•10/20/23, 4:22 PM

I filled out the form a few weeks ago.

K

Kuririn•10/20/23, 4:27 PM

I put my account id in the #Browser Rendering API Beta Access

KKuririn How can I check if I have access to the beta?

C

Chaika•10/20/23, 4:58 PM

I think the only way would just be trying to deploy a worker using the browser rendering binding: https://developers.cloudflare.com/browser-rendering/get-started/screenshots/
If it tells you that you don't have access.. then you don't

KKuririn I filled out the form a few weeks ago.

C

Chaika•10/20/23, 4:58 PM

If you never got anything back about filling in the form, I would assume you're not in it yet. Vero replies to everyone who puts their ID in that thread when they get enabled -- when that happens, you'd know for sure

Original message was deleted

M

MackenlyOP•10/21/23, 5:14 PM

It often seems a lot of forms get posted in blog posts and what not, maybe looked at a few times, and not used any further. Would be great if there were some sort of beta/early access program/request tool within the account dashboard. Really thankful for the team working on this discord.

Z

zegevlier•10/21/23, 6:04 PM

This is a bit off-topic for this channel, but I know that #magic-network-monitoring prioritized people in the discord because they usually give more feedback.

FFirefox When running the following code I keep getting a type error of `readable is not ...

I

ItsWendell•10/23/23, 6:53 PM

Any solutions found? It seems like what is returned in the PDF is not an Buffer, debugging this now more. Edit: Okay so

.pdf()

.pdf()

.pdf()

.pdf() doesn't work but this does:

        const qoutePdf = await page.createPDFStream({
            printBackground: true,
            format: 'a4',
        });

        // Create web compatible readable stream
        const { readable, writable } = new TransformStream();

        const writer = writable.getWriter();

        qoutePdf.on('data', (chunk) => {
            writer.write(chunk);
        });

        qoutePdf.on('end', () => {
            writer.close();
        });

        return new Response(readable, {
            status: 200,
            headers: {
                'Content-Type': 'application/pdf',
            },
        });

        const qoutePdf = await page.createPDFStream({
            printBackground: true,
            format: 'a4',
        });

        // Create web compatible readable stream
        const { readable, writable } = new TransformStream();

        const writer = writable.getWriter();

        qoutePdf.on('data', (chunk) => {
            writer.write(chunk);
        });

        qoutePdf.on('end', () => {
            writer.close();
        });

        return new Response(readable, {
            status: 200,
            headers: {
                'Content-Type': 'application/pdf',
            },
        });

        const qoutePdf = await page.createPDFStream({
            printBackground: true,
            format: 'a4',
        });

        // Create web compatible readable stream
        const { readable, writable } = new TransformStream();

        const writer = writable.getWriter();

        qoutePdf.on('data', (chunk) => {
            writer.write(chunk);
        });

        qoutePdf.on('end', () => {
            writer.close();
        });

        return new Response(readable, {
            status: 200,
            headers: {
                'Content-Type': 'application/pdf',
            },
        });

        const qoutePdf = await page.createPDFStream({
            printBackground: true,
            format: 'a4',
        });

        // Create web compatible readable stream
        const { readable, writable } = new TransformStream();

        const writer = writable.getWriter();

        qoutePdf.on('data', (chunk) => {
            writer.write(chunk);
        });

        qoutePdf.on('end', () => {
            writer.close();
        });

        return new Response(readable, {
            status: 200,
            headers: {
                'Content-Type': 'application/pdf',
            },
        });

Wouldn't it be more cost effective to just keep a DO with a session alive perpetually? It costs just

Similar Threads