The models are always getting faster and faster… streaming probably won’t be needed in 6 months time
Good point, maybe you’re right about that. Would you extend that to audio and video as well? I can’t help but feel like understanding the stream is going to be essential. Maybe I’m wrong about that.
Yeah, of course… none of my new client apps use streaming anymore because there’s so much value to be had in having everything entirely in the backend and being highly modular… the response times aren’t too long and you can always improve the UX with loaders.
Was this streaming plugin launched?
Yes
They do have an entire division for education now, so it could be a task that those folks pick up as it doesn’t require engineering to code anything, just maybe assist the education team with understanding the technical components the education team might not.
They may actually be working on things like that, given the direction they moved with AI providing the instructions in a modular fashion on how to build certain functionalities into an app.
Or this for actual streaming
Does it use a third party server?
Yes it does, as indicated in the description.
Streaming is so analog.
is there any plugin that support Azure OpenAI streaming?
I think Bubble is now working on it, as their Product Manager contacted me which is :
That last question is the right one… third party plugins already resolve it just as well as a native one.
- Use a third party plugin
- Use a third party plugin that allows you to self host
- Make your own third party plugin
Yeah I know the fact it’s possible with the community doesn’t mean Bubble doesn’t need to do anything, but on the list of things to do it’s not up there Making specific integrations for some specific APIs isn’t the way to go down.
I got the same email. And responded the exact same way I did on that thread. Third party plugins use third party servers or run it from the client. I can’t use either of those solutions where I’m working.
You laugh but it’s not about a specific API. It’s about whether or not bubble can handle streaming. They can’t. Not without using a third party server or sharing your api key in the browser. They aren’t fit for the future of AI.
What are the blocking points, especially given the 3rd party services Bubble is already using at its core (AWS, CloudFlare, Imgix, Unpkg…)?
These are my unadulterated concerns:
- What if the person hosting the third party server dies… and their credit card expires etc?
- What about the security of the api keys… where do they go once they leave the bubble ecosystem?
- Redundancies?
- It’s another point for functionality to break down - how big is the team to get it back up and running? will it come back up?
If it does go down is it bubble? Is it me, is it the api? Now, is it something else I have no control over? Debugging in the event of an issue and relaying that to our users and my boss.
The uncertainties are a non starter for our team.
You can’t have built your app on Bubble and be worried about being locked into a third party vendor! You can just switch if their server goes down!
Again, this is an argument for building it and owning the server yourself rather than getting Bubble to do it.
to your first response: I can be and I am. Maybe they aren’t the solution.
to your second response. My point entirely, another no-code solution would be better suited.
Given the above questions raised, which are completely valid. I would lean on the same side of @georgecollier .
That level of requirements tends to a custom build with an enterprise-level QoS wall to wall.
Using a SaaS rely on 3rd party services and it means to accept inherent risks associated with. Including dependencies you identify and try to minimize (tip of the iceberg) and those you don’t (bottom).
A fair warning though, the proposed questions you asked to qualify the 3rd party refer to trust, not facts.
I run services having uptime of magnitudes higher than Bubble. Does it mean that you should trust me more than Bubble? Of course not, but factually my users had less downtime than Bubble’s ones.
As a closing note:
That’s a pretty good image! haha