How can I use Multimodal API?

Ejb503 · 2024-12-21 06:05:57.854429

Googles Multimodal API relies on a websocket session. The docs explicitly state this is server to server only, requiring an API key. I therefore need a proxy server as middleware to stream the session to the client. I'm going crazy trying to get this working inside NextJS with Vercel hosting. The only solution i can see is running my own node server and deploying that as a proxy. Is there really no solution to being able to proxy websocket streams that is compatible with vercel hosting?

Dont think so.

My implementation is all in the client, I just need to be able to connect a websocket to the Multimodal API, hiding my API key... adding an extra provider won't help me here.

Ive read quite a lot about the issue

https://github.com/vercel/next.js/discussions/38057

GitHub

proxy websocket · vercel/next.js · Discussion #38057

Describe the feature you'd like to request websocket proxy is needed for many people Describe the solution you'd like websocket proxy is needed for many people Describe alternatives you've consider...

It seems a common problem but as far as i can tell my only solution is to self host and run a proxy server...

Means an awful lot of work (and cant use vercel hosting...)