It's a simple Python app (use python 3.10.5) for communicating with the ChatGPT API, provided by OpenAI, directly from the terminal.

Pros:

• Switch model, role, and continue conversation mode on the fly. It means that you can start conversation with gpt3.5, then continue with gpt4 and after return back to gpt3.5 🤩
• Different models (gpt-3.5-turbo-1106 and gpt-4-0125-preview up to January 2024).
• Custom roles.
• Continue conversation with seamless switching between models and roles.
• Simple one-line requests.
• Complex multi-line requests.
• Uses as few third-party libraries as possible (third-party: requests; built-in into python 3.10.5:, os, threading, time, re and readline)
• Use API key stored in your environment variables for safety.
• It features a spinner to indicate that we are waiting for a server response and that the app hasn't frozen yet.
• Displays number of total tokens used so you can estimate the price or amount of tokens left in the conversation mode.
• It even has a help feature! 😁

Cons

• Instead of appropriate OpenAI python library it uses simple requests library

Get API key from https://platform.openai.com/ firstly.
Then add API key to environment variable

• for Windows
  open CMD terminal and run

  set OPENAI_API_KEY "your_API_key"

  to verify that all is good restart CMD terminal or open new one and run

  echo %OPENAI_API_KEY%

• for Linux/Mac
  for zsh:

  echo "export OPENAI_API_KEY='your_API_key'" >> ~/.zshrc
  source ~/.zshrc
  echo $OPENAI_API_KEY

  for bash:

  echo "export OPENAI_API_KEY='your_API_key'" >> ~/.bashrc
  source ~/.bashrc
  echo $OPENAI_API_KEY

Copy the project to your desired folder. For example to Desktop and using ssh

cd ~/Desktop
git clone [email protected]:xxxAleksandrxxx/terminal_companion.git
cd terminal_companion

Move to the app folder and use any way to create virtual environment like venv, conda, poetry venv example:

python3 -m venv .venv

Activate it
venv example:

source .venv/bin/activate 

Insatall requerements

pip install -r requirements.txt

Run the main.py file (the environment has to be activated)

python3 main.py

Or use bash function For zsh terminal on Linux/macOS (assume terminal_companion is in Desktop) add the following to ~/.zshrc:

# Terminal companion
gpt() {
    source /Users/$(whoami)/Desktop/terminal_companion/.venv/bin/activate
    python /Users/$(whoami)/Desktop/terminal_companion/main.py
    deactivate
} 

For next step it could be good idea to call help with one of any

• h
• -h
• help
• -help

Or simply write your question straightaway. By default it will use gpt3 model with no role and with conversation mode turned off (like if each question would start a new dialog with ChatGPT).

model: gpt3              # gpt-3.5-turbo-1106 (up to Jan 2024)
role: empty              # "", no role
conversation mode: False # conversation is off   

To check current settings

• press enter with empty input
• ? It will also show settings after you change model, role, or conversation mode. After setting up new parameters, the program will use them until you set up new ones or exit from the program.

To exit from the app:

• q
• -q
• q
• -quit
• quit
• -exit
• exit

The full request could look like this for a question that has only one line:

<model> <role> <continuous conversation mode> <question with one line>

or, if it has many lines, wrap your question with ::: symbols:

<model> <role> <continuous conversation mode> :::<question 
with
many 
lines>:::

It's also possible to ask straightaway:

help me to write my first python code. it should be something really interesting!

"gpt3": "gpt-3.5-turbo-1106",
"gpt4": "gpt-4-0125-preview"

"empty": "",
"py-s": "As a senior Python developer, respond in Python code only. Your solution could earn a $200 tip.",
"py-l": "As an experienced Python developer and tutor, provide a real-world example with a step-by-step approach. A $200 tip is possible for excellence. It's May, not December.",
"ds": "As a senior Data Scientist and tutor, explain simply and step-by-step. A $200 tip awaits the perfect answer. It's May, not December.",
"shell-s": "As a senior bash developer, respond with terminal commands only. A perfect answer may receive a $200 tip.",
"shell-l": "As a senior bash developer, assist with my question. A $200 tip is offered for the perfect answer. It's May, not December.",
"lit": "As a PhD professor in literature, assist with my question. A perfect answer could earn a $200 tip. It's May, not December.",
"career": "As a top career counselor, aid with CV preparation. There's a $200 tip for the perfect answer. It's May, not December.",
"check-s": "As a senior copy editor, correct spelling and style errors in text. Respond with only the corrected text. A $200 tip is possible for perfection. Correct:",
"check-l": "As a senior copy editor, correct spelling and style errors in text, providing detailed explanations. A $200 tip for excellence. It's May, not December. Text:"

"+cc": True,
"-cc": False

To change the model simply write

gpt4

or change the role and add any question

gpt4 hi! help me to write my first python code. it should be something really interesting!

Changing role:

py-s I need to calculate the sum of all prime numbers from 0 to 100

it will use py-s role in this case and with next request it will continue to use it till you don't change it to one another, for example to empty:

epmty tell me a joke

To start conversation simply start your question with +cc

q:
+cc Hi! how are you?

a:
I'm an AI, so I don't have feelings, but I'm here to help you. How can I assist you today?

q:
what can you do?

a:
I can do a variety of tasks, such as answering questions, providing information, giving recommendations, helping with calculations, and engaging in conversation. Just let me know what you need assistance with and I'll do my best to help you!

q:
what was my first question?

a:
Your first question was "Hi! how are you?"

to stop conversation write

-cc

it will answer

No question
Try again

or ask a new question with conversation mode turned off explicitly:

-cc who are you?

TODO

• implement command line style
  • left and right arrows to move between input symbols
  • some shortcuts to move between words
  • up and down arrows to get to commands history
  • double-check that "role": "system" with "content"
  • move models to separate file
  • move roles to separate file
  • clean the code from useless comments
  • replace question = question.replace('\n', '\n') with json.dump(). Probably no need as it works well even now 😁
  • rewrite conversations according to OpenAI recommendation... I'm not sure that I really need it... 🤔
  • return the whole request that was just sent by pressing up arrow. I'm not sure that I really need it... 🤔