TerminalGPT

• Console app built with C# for the sake of learning.
• Current objective is to build a Mininal "LLM-Shell" desktop application.

a minimal data-flow diagram

Learnt Concepts

1. Vibe Coding:

• How to leverage the power of chatGPT to learn a new programming language and build a usable app.

2. OO Programming:

• How to model the entire app as interactions between entities with distinct roles.

3. C# Features:

• Pattern matching:
  • switch expression
  • => (arrow function) for lambda functions.
• Indeces and Range
  • https://learn.microsoft.com/en-us/dotnet/csharp/tutorials/ranges-indexes#type-support-for-indices-and-ranges

4. Adaptor Pattern:

• Essentially "I have an existing class with desired functionalities, let me do something around it so this class fits into the entire system".
• Typcially use interface (or abstract base class in C++) as wrapper around a service(dependency), so the service could easily interact with the rest of the program through the client.
• The client is implemented based on the contract defined in the interface/abstract base class.
• Adaptors implement the client and wraps around the service.

Project Status/Demo

Resources

Csharp Tutorial

1. https://learn.microsoft.com/en-us/dotnet/csharp/tour-of-csharp/tutorials/

OpenAI API

1. https://platform.openai.com/docs/api-reference/authentication?utm_source=chatgpt.com
2. https://platform.openai.com/docs/guides/streaming-responses
3. https://github.com/openai/openai-dotnet/tree/main
4. https://platform.openai.com/docs/guides/text

Design Notes

• context, the user-assistant chat history list, is passed into the LLM as a serialized JSON list.