is this still happening? incident was resolved already
is this still happening? incident was resolved already
LNo, it has been resolved already. Thanks!
Cyour my hero
Bhello, im constantly getting "The browser context is undefined. This means the CDP connection to the browser failed". this is for stagehand. using the following code as taken from the docs..
const stagehand = new Stagehand({
env: "LOCAL",
localBrowserLaunchOptions: { cdpUrl: endpointURLString(env.BROWSER) },
llmClient: new WorkersAIClient(env.AI),
verbose: 1,
});
await stagehand.init(); //errors here
const page = stagehand.page;
await page.goto('https://demo.playwright.dev/movies');
anyone get this working.
its being called from a CF workflow for context.
i have used the CF browser rendering successfully with pupetteer from the same project.
const stagehand = new Stagehand({
env: "LOCAL",
localBrowserLaunchOptions: { cdpUrl: endpointURLString(env.BROWSER) },
llmClient: new WorkersAIClient(env.AI),
verbose: 1,
});
await stagehand.init(); //errors here
const page = stagehand.page;
await page.goto('https://demo.playwright.dev/movies');
anyone get this working.
its being called from a CF workflow for context.
i have used the CF browser rendering successfully with pupetteer from the same project.
WUnable to create new browser: code: 503: message: No browser availablegetting this on my Workers
MHi, hope that somebody can give me a hand: I want to develop a headless browser, where the user can browse like in a nowmal browser & can press a button to scrape the page that he sees? Is there something like that already out there that i might be capable to use or to build upon? Thx 4 the tipp
KI also keep getting this, even though I have way less browser instances than "workers paid" allow
KOn my browser rendering page I have "Average number of concurrent sessions - 0" and the history of browser sessions mostly have "Normal closure", but some sessions are missing "end time". Aren't they supposed to time out in 60s even if they weren't properly closed?
KI might be hitting this: "Rate limits are enforced with a fixed per-second fill rate. For example, a limit of 60 requests per minute translates to 1 request per second. This means you cannot send all 60 requests at once; the API expects them to be spread evenly over the minute. If your account has a custom higher limit, it will also be enforced as a per-second rate." source
KNo, that wasn't it. I even check my limits with puppeteer.limit(browser) first, and even though it says I have available browser - I get the same error anyway.
Kthanks for reporting. we made a change that should have resolved the 503s. can you confirm if working properly now?
cc: @kol3x as you had the same issue
cc: @kol3x as you had the same issue
Kthis is odd - can you dm me with account id and session ids so we can look into this please?
RHey @BlockMinister , are you running that code in the worker directly or is it in some Durable Object? Can you try the following code change?
(instead of passing
(instead of passing
env.BROWSER"BROWSER" string to endpointURLString)
SHow do I get
X-Browser-Ms-Used value with typescript sdk? I think it is impossible with typescript sdk right now.
Bthanks for the reply, im using it in a Cloudflare Workflow step. I get a type script error when using the string as you suggested...
i deployed it anyway but it didnt work ,it failed with the same error as I reported previously.
Argument of type 'string' is not assignable to parameter of type 'BrowserWorker'.ts i deployed it anyway but it didnt work ,it failed with the same error as I reported previously.
Rthat's strange, I have it running like this:
Can you check you're using
Can you check you're using
@cloudflare/playwright@1.0.0BROWSER
Khey WebGL should work now. Can yall confirm if working for your use cases?
@marek @katu @Jon M.
@marek @katu @Jon M.
KChecked with the team -
I think it should work with .asResponse() method. Something like this:
const contentRes = await client.browserRendering.content.create({ account_id: 'account_id' }).asResponse();
const browserMsUsed = parseInt(contentRes.headers.get('X-Browser-Ms-Used') || '');
I think it should work with .asResponse() method. Something like this:
const contentRes = await client.browserRendering.content.create({ account_id: 'account_id' }).asResponse();
const browserMsUsed = parseInt(contentRes.headers.get('X-Browser-Ms-Used') || '');
WAnyone else getting 503 errors when invoking from a worker binding?
GHello, can I cancel Browser Rendering calls with an abort signal?
Wand again...
Kyou're right we had an incident yesterday and today. it's our top priority to get to the bottom of. major apologies
E@JDoc I've had this happen a ton locally, but it seems better in production. Did you experience that Playwright(-MCP), too? (I have)
AThe issue concerns the parser start time.
EEThis may be more about #agents , but I'm surprised (or terrible at investigation) that https://github.com/cloudflare/playwright-mcp isn't
I would've expected it to internally do something like
acquireI would've expected it to internally do something like
routeAgentRequestenv.BROWSER
NHad a question regarding the Browser Rendering’s Rest API: In the docs it mentions that if a request is rate-limited, it’ll return a “Retry-After” header with the number of seconds until you can try again: https://developers.cloudflare.com/browser-rendering/platform/limits/#error-429-too-many-requests
From testing the api in curl it doesn’t seem to be including this header though. Even using the sample JS code shown ends up with a null retry count. When I was testing it last week I feel like it was including this header, so is this related to the outage that happened?
From testing the api in curl it doesn’t seem to be including this header though. Even using the sample JS code shown ends up with a null retry count. When I was testing it last week I feel like it was including this header, so is this related to the outage that happened?
AHiya, we're trying to port our PDF rendering from DocRaptor to the Browser Rendering API, but it seems like loading custom fonts via the REST API isn't working, and we'd like to avoid spinning up a worker just for this purpose. Is there a way to allow loading fonts via
style
WI've loaded custom fonts from Google fonts - but invoking via Worker, might be different via API
AYeah unfortunately we can't use Google fonts, we've got a custom font we self-host on our CDN. I know we can do this with a Worker, but we'd really rather not do that - our main app is hosted on GCP right now and building a GCP <> CF auth integration is more effort than it's worth to get fast PDF rendering
HDoes anyone else have this issue?
curl -X 'POST' 'https://api.cloudflare.com/client/v4/accounts/9357fcd5fdd482d8ff83604451aad279/browser-rendering/markdown' \
-H 'Content-Type: application/json' \
-H 'Authorization: Bearer INkf8LWxNWrE96x4Z-zDCiAgoxMG6KfUuYvUAvyg' \
-d '{
"url": "https://www.czone.com.pk/graphic-cards-pakistan-ppt.154.aspx"
}'
{"success":true,"result":"# !Icon for www.czone.com.pkwww.czone.com.pk\n\nVerifying you are human. This may take a few seconds.
\n\nwww.czone.com.pk needs to review the security of your connection before proceeding.\n\nVerification successful\n\nWaiting for www.czone.com.pk to respond...\n\nRay ID:
ce & security by Cloudflare"}
I'm simply trying to get prices from a website and get blocked by cloudflare itself. So its not possible to use cloudflare browser rendering for website that use cloudflare?
curl -X 'POST' 'https://api.cloudflare.com/client/v4/accounts/9357fcd5fdd482d8ff83604451aad279/browser-rendering/markdown' \
-H 'Content-Type: application/json' \
-H 'Authorization: Bearer INkf8LWxNWrE96x4Z-zDCiAgoxMG6KfUuYvUAvyg' \
-d '{
"url": "https://www.czone.com.pk/graphic-cards-pakistan-ppt.154.aspx"
}'
{"success":true,"result":"# !Icon for www.czone.com.pkwww.czone.com.pk\n\nVerifying you are human. This may take a few seconds.
\n\nwww.czone.com.pk needs to review the security of your connection before proceeding.\n\nVerification successful\n\nWaiting for www.czone.com.pk to respond...\n\nRay ID:
997a23bf1a81339ece & security by Cloudflare"}
I'm simply trying to get prices from a website and get blocked by cloudflare itself. So its not possible to use cloudflare browser rendering for website that use cloudflare?
Chey there ! using Stagehand in Browser Rendering I am wondering how to reuse sessions the same way its currently possible in Puppeteer as per these documentation : https://developers.cloudflare.com/browser-rendering/workers-bindings/reuse-sessions/
I am not sure I cant get to the browser sesh there and would love to not use DO for that
I am not sure I cant get to the browser sesh there and would love to not use DO for that
Cloudflare Docs
The best way to improve the performance of your browser rendering Worker is to reuse sessions. One way to do that is via Durable Objects, which allows you to keep a long running connection from a Worker to a browser. Another way is to keep the browser open after you've finished with it, and connect to that session each time you have a new request.
Cmaybe here ? https://docs.stagehand.dev/v3/best-practices/cost-optimization#3-optimize-browser-sessions
Stagehand Docs
Minimize costs while maintaining automation performance
KWhether browser rendering can access a site or not depends on the settings of that site. So seems the site you are trying to use is blocking bots
Browser rendering always identifies itself as a bot, so if the site owner chooses to block all bots, then browser rendering will be blocked too.
if it's a site you own, you can create a rule to allow for browser rendering
https://developers.cloudflare.com/browser-rendering/faq/#will-browser-rendering-bypass-cloudflares-bot-protection
Browser rendering always identifies itself as a bot, so if the site owner chooses to block all bots, then browser rendering will be blocked too.
if it's a site you own, you can create a rule to allow for browser rendering
https://developers.cloudflare.com/browser-rendering/faq/#will-browser-rendering-bypass-cloudflares-bot-protection
Cloudflare Docs
Below you will find answers to our most commonly asked questions about Browser Rendering.
DWhen trying to launch the browser inside a worker, it works 100% of the time. When launching from a workflow, it fails about half of the time with this error:
workerd/jsg/setup.c++:38: fatal: V8 fatal error; location = :0; message = Missing deoptimization information for OptimizedJSFrame::Summarize.
workerd/jsg/setup.c++:38: fatal: V8 fatal error; location = :0; message = Missing deoptimization information for OptimizedJSFrame::Summarize.
Kyou can
addStyleTagCloudflare Docs
The /screenshot endpoint renders the webpage by processing its HTML and JavaScript, then captures a screenshot of the fully rendered page.
Kchecked with the team:
They can use
Full example here https://developers.cloudflare.com/browser-rendering/workers-bindings/reuse-sessions/#4-code
They can use
keep_aliveconst browser = await puppeteer.launch(env.MYBROWSER, { keep_alive: 600000 });sessionIdhttps://developers.cloudflare.com/browser-rendering/platform/puppeteer/#list-open-sessionsFull example here https://developers.cloudflare.com/browser-rendering/workers-bindings/reuse-sessions/#4-code
Cloudflare Docs
The best way to improve the performance of your browser rendering Worker is to reuse sessions. One way to do that is via Durable Objects, which allows you to keep a long running connection from a Worker to a browser. Another way is to keep the browser open after you've finished with it, and connect to that session each time you have a new request.
CHey! Thanks for your reply, I have found this link already, but does this work for stagehand ? This seems to only work for puppeteer; I literally posted this link in my OG post haha
VThe /snapshot API documentation is outdated? or the response has been changed, I don't see
statussuccess: true/false
VFor example, I passed a bad link in url - which return 404 on browser, but there are no way to find out it retuned 404 using the /snapshot API. Which enforce me to send 2 requests (one for status, another for snapshot)
NHave anyone ran into an issue while trying to download a file via Playwright
The readable has no data in it. Doesn't throw an error, but just 0 chunks.
Download.createReadStream()The readable has no data in it. Doesn't throw an error, but just 0 chunks.
AYes we've tried both
addStyleTagfont
Mi made a worker for browser rendering and deployed it, the first request i made i got this error, why?
Error: Unable to create new browser: code: 429: message: Rate limit exceeded
Stack trace:
Error: Unable to create new browser: code: 429: message: Rate limit exceeded
at PuppeteerWorkers.acquire (index.js:18328:13)
at async PuppeteerWorkers.launch (index.js:18238:22)
at async Object.fetch (index.js:19976:23)
Error: Unable to create new browser: code: 429: message: Rate limit exceeded
Stack trace:
Error: Unable to create new browser: code: 429: message: Rate limit exceeded
at PuppeteerWorkers.acquire (index.js:18328:13)
at async PuppeteerWorkers.launch (index.js:18238:22)
at async Object.fetch (index.js:19976:23)
Manyone?
MHey team, what's going on here?
I'm trying to run the documented code snippet from https://developers.cloudflare.com/browser-rendering/rest-api/json-endpoint/#using-a-custom-model-byo-api-key
Simply replacing the ai provider from anthropic to google-ai-studio/gemini-2.5-flash, or google-ai-studio/gemini-flash-latest. I get the error {"success":false,"errors":[{"message":"AI error: undefined: undefined"}]}%. Looks like someone else is having the problem too: https://community.cloudflare.com/t/browser-rendering-json-with-custom-ai/854912
Am I missing something here, or is the snippet in the documentation incorrect?
I'm trying to run the documented code snippet from https://developers.cloudflare.com/browser-rendering/rest-api/json-endpoint/#using-a-custom-model-byo-api-key
Simply replacing the ai provider from anthropic to google-ai-studio/gemini-2.5-flash, or google-ai-studio/gemini-flash-latest. I get the error {"success":false,"errors":[{"message":"AI error: undefined: undefined"}]}%. Looks like someone else is having the problem too: https://community.cloudflare.com/t/browser-rendering-json-with-custom-ai/854912
Am I missing something here, or is the snippet in the documentation incorrect?
Zany chance anyone is making screenshot of an element? screenshot of playwright page works excelent, but when I try to run locator to find some element and then run screenshot - timeouts, tried like 5-6 different methods to get element
ZI wonder how much of time of browser rendering cpu time of workers is counted. Since browser rendering is paid, I hope majority of time spent on that will not be billed as worker cpu-time?
