Thanks for the added details. I suppose I’ve had it drilled into me that it is never a good idea to expose an API key on the client side. Is that an issue here or now?
Yeh you’re absolutely right and that’s on me to resolve. I’ll make the changes and let you know once done.
Update v1.20.0
Groq/whisper model for real time transcriptions + API key security.
The whisper model can now be used for real time transcriptions as an alternative to using Deepgram. It has some advantages, the main ones being that’s it free and handles multiple languages better. The settings around this have changed within the plugin element.
Both of these options require API keys to be set, the keys have been moved out of the plugin and are now stored within the tokenAuth page, accessed here
https://tokenauth.bubbleapps.io/ for added security.
The Groq key is used when using the whisper model for transcriptions. The Deepgram key is used when using Deepgram’s speech to text service for transcriptions, and the Hume AI key is used when you want to make use of the expression management service.
The docs page has been updated to reflect these changes.
https://paul-testing.bubbleapps.io/version-test/livekit_instructions?tab=1
https://paul-testing.bubbleapps.io/version-test/livekit_instructions?tab=10
https://paul-testing.bubbleapps.io/version-test/livekit_instructions?tab=12
In the plugin options is a new token field that you’ll need to enter after successfully generating one.
Agents
A quick update on the agent…
I’m having some issues trying to find a suitable location to host the agent at. It’s a separate program which needs to run in an environment so it can connect to a room. Unfortunately, it can’t be run in Bubble through any type of client/server side code. LiveKit have a deployment guide on this but since I’m attempting to get the agent running somewhere where I can control it through the plugin and have it dynamically configurable so you can pass your own WS URL and API keys to it, is proving a bit of a challenge.
I’m able to spin it up locally and connect it to a room, it works really well and it’s fast to respond when using OpenAI’s speech to speech service. There are some bugs that will probably get resolved over time since this JS agent has only been out for a few months.
I’m still working on it anyway and watching for updates from LK.
Paul
Update v1.22.0
Fixes and changes made to the camera/mic start-up options to support usage for online events, conferences and webinars plus fully customizable setup to configure the plugin to match the theme of your site.
There have been 2 additional demo pages setup for this. One is the presenter page where a dedicated participant can host a session. The other page is the viewer where viewers can connect to view that session.
Demo pages
Presenter: presenter
Viewer: viewer
Paul