Hi Bubblers! 
I just released OpenLLM Hub, a new plugin designed to give you the fastest and smartest AI backend for your Bubble apps, without the complexity of manual API calls.
We are leveraging the Groq API to deliver inference speeds of 500+ tokens/s (near instant), and yes, we support the new Llama 4 models which are natively multimodal (Text + Vision)!
Why use this instead of the standard OpenAI plugin?
1. 
Built-in Memory Logic (No more loops) Managing chat history in Bubble is painful. You usually have to manually stitch lists, calculate tokens, and format JSON. This plugin does it for you.
-
Includes a server-side action Prepare Chat Context.
-
You just pass your list of User/AI messages from the DB.
-
The plugin formats everything, cuts old messages to fit the context window, and prepares the JSON for the API.
2. 
Llama 4 Vision Support Give your app “eyes”.
-
Upload a receipt, a screenshot, or a photo.
-
Ask the AI to extract data or describe the scene.
-
Works seamlessly with the new llama-4-scout models.
3. 
Strict JSON Mode Perfect for SaaS apps that need to extract data. Force the AI to reply in valid JSON every time, so you can save the response directly to your Bubble Database fields.
4. 
Bring Your Own Key (Free Tier available) We don’t charge a markup per token. You use your own Groq API Key (which currently has a very generous free tier).
LINK:
- Version (Memory + Vision): Link
I’d love to hear your feedback and see what you build with Llama 4! Happy Bubbling! 
Also, we updated our plugins prices! If you want to check them out just CLICK HERE!