No Code Chatbot (Post 1)

Using ChatGPT to build a chatbot from scratch

LinkedIn kickoff post

As a kid, I remember watching a Twilight Zone episode called “The Man In the Bottle”. An elderly couple unleashes a genie with the power to grant wishes. Except each wish brings trouble for the couple. I can say the same about prompting with any LLM. Vague prompts create unexpected answers. Drawing upon the project management experience and training, I tasked ChatGPT to meet the following requirements.

Goal: I am building an app to allow end users to create incident tickets with an AI agent. The incident tickets follow ITIL v4 best practices.
Project requirements:
1. The chat interface will use the Python Flask framework.
2. The chat agent (AI) will ask questions to generate the following information: a summary of the issue, a description, category, user name, user ID, call callback number.
3. The chat agent will repeat the information to the user to verify the ticket matches the user's problem.
4. The chat agent will generate a text file with the information.        

ChatGPT generated a high-level plan that provided the steps need to build a chat application.

- Set up the Python Flask environment.
- Create a chat interface using HTML, CSS, and JavaScript.
- Integrate the chat interface with the Flask backend.
- Implement the AI chat agent using OpenAI GPT or a similar library.
- Design the chat flow to gather necessary information for the incident ticket.
- Verify the information with the user.
- Generate a text file containing the ticket information.        

The high level view helped me break down the project into actionable pieces, which helps me understand developing an application and it also allows me to break down each part of the application to ensure ChatGPT doesn’t go over its character limitations.

ChatGPT generated a high-level plan that outlined a decent roadmap to complete my task. The outline broke down my tendency stall at “where do I begin” part of any new project. By breaking the tasks into manageable sizes, it becomes easier to comprehend the development of this application.

This has three major benefits:

1. Breaking the project down into actionable pieces reduces complexity and procrastination. By focusing on each step, I can better understand the process and achieve the goal. I can pause development and ask questions when a concept escapes my understanding. All of this helps me move forward toward the goal.
2. Staying within ChatGPT character limitation can be challenging. Seeing ChatGPT generate code only for it to stop half way can frustrate anyone; however, by using this process, I can prevent this frustration and increase the odds of a fully functional code for each part of the application.
3. With the help of the outline, I can test each part of the code and track my progress. I can catch errors in the code that is generated and re-prompt ChatGPT with better insights. This process further refines my prompting skill while leveraging ChatGPT’s strengths.

In the upcoming posts, I’ll share the code generated by ChatGPT, as well as images of the application.