top of page

Ai Assistant

Developed to help me with my overall organization and problem-solving skills, I am developing an AI voice assistant which can receive voice input and output an audible output according to the given prompt.

Preliminary Testing

How does the GPT API respond to given commands?

Prior to starting this project, I wanted to see what I could employ the GPT-3.5 API to do. As such I used ChatGPT as a simulation to get a better understanding of what data will be returned. 

​

According to my initial tests, I have observed that ChatGPT will react in accordance with the initial prompt given. If asked to be an assistant, it will give reasonable advice according to the following prompts. As such, a more descriptive initial prompt is in requirement as to improve the responses produced by the API. 

​

Things GPT struggles with maintaining: 

  • Commands processed between long periods of time

    • I.E initial command given on the 1st, additional updates given on the 15th​

  • GPT will include unnecessary information and will give run-on responses if the response is not contained. 

  • Will unify time frames if not specifically provided with the information

A more improved initial command

After prior testing, I developed a better command that allows GPT to provide a more specific answer in accordance with my commands. Additional factors included in the command: 

  • Request to organize schedule:

  • Total schedule time frame restraint

  • Maximum word limit 

    • GPT struggles with following word limits​

  • Priority reference for tasks

    • High, medium, low priority tasks​

​

Additional factors noticed:

  • GPT will require continuous updates of the date and time. Struggles to understand that the date has changed unless given the information. 

    • Inputted data will need to be deleted periodically so as to not confuse GPT with unnecessary information.​

Libraries/API's

1 / Speech Recognition

This library's purpose is to recognize speech input coming from a microphone. I plan to use this so the assistant is always on the lookout for incoming commands. 

2 / OpenAI GPT3.5

This library was created by Open AI to allow for easy access to the GPT Api from any Python code. Will use the GPT API for generated schedules and responses based on the given information.

3 / TikTok Text to Speech API

I am using this API as a method of converting the GPT responses into an audible format. This API will allow for a proper output audio instead of a default robotic voice.

bottom of page