Automatically switch provider based on envirnment variables, Ollama support: closes #13

[Le09]

Fairly straightforward. In the spirit of the repo, I've kept things simple rather than also automatically add everything (like optional Anthropic libraries or others).

I think asking beginners to modify directly call_llm poses a risk of them putting their private keys on Github, so it's better to have them learn how to use environment variables to avoid taking chances.

[zachary62]

Thank you! I'm a bit worried that env will add friction to semi-technical people, but will add notes about security issue

[eliliam]

Nice and clean, very extensible too if anyone wanted to add new providers down the line.

[redliu312]

hihi @Le09

do you think it is required to add

from dotenv import load_dotenv
load_dotenv()

in the top section of the call_llm.py

I tried your branch and it seems this is needed for load the .env file when the script executed

[Le09]

@redliu312 Oh, I see.
Environment variables are variables that are set in you terminal. So a standard development workflow is to do

workon Tutorial-Codebase-Knowledge-venv  # activate your virtual environment with the correct set of python dependencies
source .env  # load the API keys
python file.py

If done this way, the variables are available in everything that you do in this terminal. So python main.py and python utils.call_llm.py would both get variables.
If you do the call from main, then main loads the variable first so call_llm should get them.
So, the issue is that calling utils/call_llm directly does not get them from main; since this is for testing purpose, the load_dotenv() should indeed be added to this file, but under the if __name__ == "__main__": line.