openai custom assistant integration
# support
b
I was able to add the openai key and that was checked but now how do i access my own custom assitant, i'm confused on what url to put in the custom url box for cusotm model and when i click on openai model i don't see my custom assistannt i made with openai playground
z
Do you want to use the assistant or custom model?
We don't have openai Assistant yet. What is your use case?
b
yes it would be using my custom openai assistant
z
yes, as mentioned by Jay we don't support it yet. Can u tell us your exact usecase for which u would prefer OpenAI assistant API?
y
Hello, Is it possible to use predefined Assistants in OpenAI? https://platform.openai.com/docs/assistants/overview Because is there a usage example to manage "Retrieval" and "Functions" actions through the custom gpt assistant? Currently, only functions can be defined via Vapi, but it seems there is no possibility to answer questions from certain documents.
As a scenario, I want to be able to answer questions through the documents we upload to our own assistant.
j
following this
I want to be able to provide data/files as context for assistant like you can with with openAI assistants/GPTs
b
Same
o
me too
a
same. This would also allow me to use the "required" properties section within the function call (which does not seem possible within the Vapi software).
m
I also would like to see some native vapi feature that can be used as a Knowledge base for the assistant
f
Me too
a
ditto
d
Me too
I am trying to go around this by creating my server that streams responses from my OpenAI assistant and then adding this server to the "use your own server" option for vapi
Is anyone going down this route too?
n
Can you explain, sounds interesting. How is the latency ?
n
yes knowledge base is a huge request. we have it on the docket for next week
@Abe required is possible but you have to use the API to specify the functions
d
I am currently still building it out. I am expecting mad latency so I am thinking of adding some phrases for my call assistant to yap about so that that can buy me 9-20 seconds to cover up for the latency.