One of my many projects using AI. As a CP volounteer I often need to take care of people in stress situations and the same happens of course to all the CP people.
It's not easy, so I create this training app where you can choose the simulation a speak to AI people in distress. The simulation is virtually never ending, but of course context lenght is a physical limit. At the end of the simulation, the volounteer will receive a result of the interaction.
Here are some screenshots (only italian language, sorry).
The project actually hides other two projects linked to the Tiiny:
To make it portable, I connected Tiiny with a Raspberry Pi using the LAN cable. The RPi also acts as dhcp server to assign an ip address to the Tiiny. The interface is accessible connecting a PC to the wifi created by the RPi, or connecting the RPi directly to a display and a keyboard.
To make Tiiny easier to use in my Python projects I created a gateway to intercept all the requests from my apps and a Python library. The gateway has a timed queue to avoid errors during concurrent requests. With the library (OpenAI compatible) I can just call every api using simple syntax and for apps that were configured for using ChatGPT I only need to change the creation of the chat object. Everything else does not change.
Several other projects are on the way to use my Tiiny.
#UseCase
That’s fantastic! Is there a plan to share information about open-source projects? I have an idea for a salesperson training system.
There are few things that need to be changed and fixed, but I think I could release the sources. I do not have much time, but when I manage to make the changes I will surely make a post about this.
@carlomascellani: expect!
@ecolee1126: I made several changes (also translated into english). You can find the sources here: https://www.carlomascellani.it/tiiny/trainer.zip Look at the README.md for installation and configuration. I tested on both Linux (Ubuntu) and Windows and works fine.
This is really inspiring! And I got a quick question. What dimensions does the final evaluation score (e.g. empathy, communication, procedure)?
Currently it evaluates 6 variables: joy, sadness, anger, surprise, fear, disgust and the chat history by sending everything to the ai model with a specific system message.
@carlomascellani: Good job!!
You're full of ideas.
Thanks! :)
Wow, this is a meaningful and practical use case
Great idea! Thanks for sharing!