- Cerum AI
- Posts
- Custom Chat with PDF + more w/ Bubble.io Plugin
Custom Chat with PDF + more w/ Bubble.io Plugin
Integrate custom LLMs in just a clicks
Howdy fellow builders & entrepreneurs!
A lot of you guys have been asking how to build and implement your own custom LLMs using your data into your applications such as with bubble.io and may not have the technical know-how or budget to create your own solution.
At Cerum.ai (formerly known as No Code Blackbox) we’ve created a plug n’ play solution to achieve just this. It’s not only ready out of the box, but it’s built to be integrated with your existing bubble workflows and secure enough to handle personal/company sensitive information.
Introducing Cerum Flow 🎉
How does it work?
Upload any relevant data you’d like to be referenced for your language model

Upload all types of file formats (PDF, CSV, JSON, DOCX, etc)
Create your prompt template, this will serve as a set of instructions for your language model to perform on top of

3. Install the bubble.io plugin to have access to our workflows (direct integration) and past in the API key generated from us.
That’s really it…

Aside from simplicity, we built this with one thing in mind, to be abstract as possible meaning it can provide a lot more than just chatting with as far as it’s implementation goes. As a workflow this can directly be integrated within your other bubble workflows, making it communicate with other data and existing workflows within your application.
For example you could summarize data coming in from other API’s that your app communicates with.
All of this is powered by a framework called Langchain, essentially allows you to really bypass many of the limitations that current LLMs such as GPT-3 & 4.
If you’re interested in checking out a demo of our product, I’ve uploaded an in-depth video on my youtube channel here showcasing this, and I have a signup link for early access (beta) that’s releasing this week!
I know their are some other Open source apps out there that provide very similar functionality however the issue with these was they just aren’t consumer/production ready just yet. Such as with authentication, security, abstractness, etc. Along with this their is a lot of set-up / configuration to get something like this working. Our solution is simplistic and ready, and we’ll be continually adding on features based off the feedback we receive.
That’s all I got, stay hungry & keep creating my friends!