paper geode
For variaos reasons we have to run LLM requests through our own backend. So we cannot rely on @ai-sdk to make the connection. Is there a way to have something custom here? Like instead of providing the ai-sdk we hit our endpoint directly for the requests?
I would like to have the whole state machine and agent machine run on the client.
I hope I made our usecase clear. Happy to clarify further
thorn wasp
You can write a custom provider to work with the ai-sdk https://sdk.vercel.ai/providers/community-providers/custom-providers
they also have providers built that handle self hosted models
rotund moon
I strongly recommend working with the ai-sdk, which is a really robust and battle-tested library - if we dropped the dependency, we'd have to basically reinvent it 😨
paper geode
Alright thx, i thought there might be a way to use ai-sdk with our custom API. I ll give this a go.
amber rivet
Hey @paper geode I have a similar use case. Were you able to create a custom provider that works as "proxy" to forward the requests to your own backend and use it with the agent?
paper geode
We havent started implementing yet. Will let you know if we have a solution.