Share feedback, ideas and get community help

Updated 3 months ago

Assistent Openai not loading

When entering credentials, the openAI assistant does not load
Attachment
image.png
7
C
E
s
49 comments
Assistent Openai not loading
I have the same problem
Same here in my Typebot account and in my self-hosted one, latest release.
I'm having the same issue. I hope Baptiste will fix it soon.
agreed not working on both
@Baptiste do check
The only way I made it working is saving the assistant ID in a variable and assigning it to the assistant I would have chosen. Try this as a workaround until it's fixed
@Baptiste sorry to follow up on this. Did you have a chance to have a look at this issue?
i have sme problem
I can see the values loading they are just mapped to the wrong element
I have the same issue.
Attachment
image.png
Do you have a screenshot of this because this appears not to work.
Anyway, I just rolled back my instance to 2.28.0 for now. I'll update again once the issue it's solved.
Ok thanks I wanted to make sure my instance was not messed up as the solution propsed is not working in the latest release.
The video I sent is from the latest release though.
This is how it works,
One question is it possible to pass a script to enable filesearch?
I think you can just add the files into your assistant and enable File Search. It should use them automatically. At least this is how it works with Vector Stores and works wonderfully.
In version 1 of the API it was this way, but now it is not possible to enable the use of a file in the vectorstory as a default.
I've literally just tried with API v2. Assistant gpt-4o with file search active, connected to a vector store and 1 pdf added into it. It worked.
For reference
Attachment
Screenshot_2024-09-05_at_20.35.13.png
yes, it's the same way but to use file search for all conversations, you now need to enable it
It only uses filesearch in the conversation if it is enabled
Attachment
image.png
but this version cannot always be kept enabled, it needs to be enabled for each new conversation
Yes you can. The problem is that you haven't enable it by default, but this can't be done in the Playground. You should go to this page first. Select your assistant, enable File search. Then go to the playground. It will be enabled by default.
it is configured like this
Attachment
image.png
I thought it should work like this by default
It does for me πŸ˜…
I created a new assintent the first time it worked, but the second time it no longer used filesearch
but thanks for the help
Does anyone know of any alternative until they resolve this?
If you create an assistant you can pass that ID to the AI block
I tried but I couldn't, thanks for trying! I'll wait for them to solve it, I hope they find a solution πŸ™πŸ»
@Baptiste im sure you are up to your eyeballs with stuff when I revert this things seem to work
at the same time it appears to not care if the base url is not api.openai.com
Not on the main release.
You can downgrade to 2.28.0 or if you compile your own you can add a list to the above commit and then it will work however you will be required to enter in your model endpoint. I don't suggest doing that unless you are familiar with this code and building your own viewer and builder form source. I am sure @Baptiste has a fix in the works this being an unforeseen bug it may not be a priority currently.
Sorry about that, I went into deep work yesterday and did not even open Discord.
This was just fixed
For self-hosters, you can upgrade to 2.28.2
THANKS ❀️
@Baptiste Do you know if it has been completely resolved?
Attachment
image.png
{
"message": "Could not create stream"
}
I would say so. I've upgraded to 2.82.2, and that error doesn't appear anymore.
Add a reply
Sign up and join the conversation on Discord