Syeah but to support streaming had to do a little custom stuff
Scan send if you need
Dbad, or lazy developers?
Sbad
Sthe whole thing
DI feel like I'd end up reinventing a fair amount of it
DI agree the code quality is questionable, since like.. as a new user, I really shouldn't be able to find mistakes and give a PR to address obvious bugs on my first use 
https://github.com/langchain-ai/langchainjs/pull/3189
https://github.com/langchain-ai/langchainjs/pull/3189Slike what exactly? nothing is too complicated and it's better to know what actually happening
DI like how there's an LLM class
Swe used to be friends
Dthis will probably be the postmark situation v2 in the end
Drelated question, if you don't actually "need" to stream your response back to the client, is there any real benefit to doing streaming instead of awaiting for the full response? (My guess is it's better to not stream if you don't need to, since that would cause more CPU usage on the Worker.)
Syeah probably better to not stream if you don't need it but with OpenAI we have seen weird issues and actually stream everything even if we just buffer in the worker
DYou’ve seen weird issues if you stream, or if you don’t stream?
Sdon't stream but it's possible it was happening with streaming too because by the time I added logging and pushed it the behavior went away but I was thinking there was some sort of network timeout happening between CF and Workers without streaming. it was strange
Salso there's a major outage right now with OpenAI
DPerfect timing, I was just heading to bed. 
Sit might have gotten me out of bed
BI can't open devtools when running "wrangler dev", any ideas?
HTry
http://internal/hash IIRC the URL has to be valid
Ddid you set up https://status.openai.com/ in your Slack? I did that a few days ago, it's really handy.
Welcome to OpenAI's home for real-time and historical data on system performance.
D
Sin our Discord yeah
SSlack 
Dsorry, I'm making this a bit too #off-topic lol
DDoes anyone have a public "ping" utility that uses a Worker to check the latency to a hostname?
DIs there a aws4fetch equivalent but for Google GCP?
KWhat's wrong with using it in GCP 
KIt's for anything that has the
fetch
DI want free authentication checks before getting billed for a request to my container(s).
Dhttps://github.com/aimoda/cloudflare-worker-to-aws-lambda-function-url-example <- I want this but with Google Cloud Run
KI still don't follow 
KThat just uses values from env. so what would stop you from doing that with a GCS Service Account & https://cloud.google.com/run/docs/authenticating/service-to-service
Doh https://cloud.google.com/run/docs/authenticating/service-to-service#acquire-token might be exactly what I wanted, thank you!
D
.Hey, I'm kinda new to workers. Why in the logs does it show that my post request was cancelled?
D@kian thanks a ton, got it working!
Dhmm, does Azure OpenAI have the same outages as OpenAI? I wonder if using both would fix a lot of issues.
Dyep, it would fix downtime issues. Azure OpenAI is hosted separately by Microsoft, see https://learn.microsoft.com/en-us/legal/cognitive-services/openai/data-privacy
The Azure OpenAI Service is fully controlled by Microsoft; Microsoft hosts the OpenAI models in Microsoft’s Azure environment and the Service does NOT interact with any services operated by OpenAI (e.g. ChatGPT, or the OpenAI API).
Syeah would fix it and so would dedicated instance
R
Show is this a "masked link"? just looks like some Markdown conversion fail
