https://www.activepieces.com logo
thank for your answer!
# requests
t
thank for your answer!
m
I made a quick example flow if you wanted to generate a new image every hour for example with ChatGPT generated prompt, and InvokeAI with SDXL
t
Thank, I will search how to replace chatGPT with local LLM now
Is it possible to change API server of the chat GPT node?
m
Are you using Activepieces on cloud.activepieces.com or self hosting it?
You can use a piece called LocalAI to change the URL
t
self hosting
thank !
m
Make sure the url is without the HTTP/1.1 at the end, so:
Copy code
http://127.0.0.1:9090/api/v1/queue/default/enqueue_batch
If that still doesn't work, what do you get when you click test action on the send http request piece?
m
The URL with enqueue_batch at the end is for InvokeAI invoking, not LM Studio 😜
For LM studio just put http://localhost:1234 and try that
error again, the LM Studio console detect nothing =(
its only when I put the url in a web navigator
that the lm studio console detect something
Is really localAI node compatible with LM Studio?
m
Hmm try removing the slash at the end of the URL
t
(I read the doc, maybe something with ngrok)
I tried, no change
m
Are you getting any errors in the activepieces console?
t
jusrt that
but when I try a http request module
same error
(I run Activepiece on local Docker)
m
Then try changing the URL from localhost:1234 to the access url for the activepieces docker container
I think localhost points to the docker network for AP, which LM studio is outside of
t
oh I see
m
Try
Copy code
http://host.docker.internal:1234
t
yes!
finally something!
now I'm here
Activepiece detected my LLM model, good
but the test failed again =(
m
What version of LM studio are you running?
latest version
m
Actually the endpoints are the same nvm
Can you try toggling CORS in the server options on LM studio to on?
t
same error
m
Ohh my bad, the URL needs to be
Copy code
http://host.docker.internal:1234/v1
Weird that models are getting read but the prompting doesn't work 😂
t
yes!
the model load without /v1 in url
m
xD Okay I see the issue now
I'll deploy a fix for it by end of day, but for now, try setting the model value manually to the path of the model file here https://cdn.discordapp.com/attachments/1179561969322311750/1180103548277633044/image.png?ex=657c3413&is=6569bf13&hm=171875d30eeea03838a7a0af60fa26a7fd11b5aa401e17e36975ffbd57994dfe&
t
great!
It worked, thank you ❤️
m
Perfect!
t
where do you fin the template for this json ?
m
From the network tab 😜 I copied the request the invokeAI frontend sends when you click invoke
Unfortunately it isn't documented anywhere
t
thank, I finally get it work! do you have any idea how to retrieve the image on activepieces?