Quick and simple way to have AI based Chat in PeopleSoft Application
LLM Model/AI Chat Assistant in PeopleSoft
Introduction:
For a long time I was having a question "What would be the perfect business case to implement the AI and ML models on PeopleSoft". But recently I felt that "Any big picture starts with a stroke", so wanted to have that stroke by implementing AI Chat on PeopleSoft Application for a general tasks like content summarization, content filtering or searching latest PeopleSoft news without going out of the PIA screen.
How it was done?
Surprisingly the steps are very simple and anyone who is having a basic understanding of HTML, Scripts, and REST Services can do it within a day.
- Step 1: Get your API-key (or token) and service details from the service provider. I used Groq for POC as they offer free plan (check out: Groq Free Plan - Rate Limits)
- Step 2: Create a simple AI Prompt to create a JavaScript Code for you to have the Chat Window overlay on PeopleSoft Application. And save the code as HTML Object in Application Designer.
- Step 3: Include the HTML on any existing HTML/Script object to load during the site loading. I used PT_SAVEWARNINGSCRIPT, as it is present in both Classic and Fluid components.
What you should be doing next?
Depending of your interest and curiosity level you could choose any or both ways.
Quick way: Get the project and JS code from my GitHub Sample page and import it into your development or demo
Developer way: Follow the above steps to have your own Chat Assistant.
You could even choose to set the model's behaviour by setting the SYSTEM role content on requests. Or use Postman or similar tool to validate the content before putting them on PeopleSoft PIA.
Things to Note:
- Though it may look like a simple plugin but it gives confident to the management and perspective to future plans.
- Play around different LLM models to scope your business cases, like voice to text, audio processing etc.,
- I have provided an option to feed the API Key during the PIA Load. In practical you would need the same encrypted and stored in server and retrieve it according to the session.
- You could include the similar service on a classic/fluid component, and set the context of a model to serve specific use case.
- Ensure the compare the project with your DB before importing as it has a customized HTML (PT_SAVEWARNINGSCRIPT)
I hope this would give you perspective and interest in making such improvements in the PeopleSoft application. And the Big Picture is coming soon.
Here is the Demo Video for your reference:
Thanks and Happy Learning. 😊
Comments
Post a Comment