Getting started docs are a bit sparse

[willingc]

Hi @ericmjl ☀️ I've been trying to test out llamabot and am getting bogged down with configuration issues. I'm trying to get the Feynman simple chat example with either the OpenAI API key or ollama locally with no success. Unsuccessful key validation for openai and missing dependencies with ollama (langchain-community, langchain-core, langsmith).

Any tips? I will be a happy to write docs (or have openai do so) once I am up and running.

P.S. Cool SciPy talk. We should all have a Paul Ivanov pun bot.

[amirmolavi]

Hi @willingc , I am new and trying to play/learn this cool bot :) I just tried it in a notebook. I use poetry for environment management. you can find the setup here. I also set up the environment variable like this:

As I recall, these are all I had to do.

[willingc]

Thanks @amirmolavi. Your config helped me troubleshoot my pixi environment. For some reason, pixi was resolving llamabot to 0.4.4.

Going to try a different package management tool now.

Update: Got it working with miniconda.

[willingc]

Thanks again @amirmolavi!

[ericmjl]

Thanks @amirmolavi. Your config helped me troubleshoot my pixi environment. For some reason, pixi was resolving llamabot to 0.4.4.

Going to try a different package management tool now.

Update: Got it working with miniconda.

@willingc I thought about what you might be encountering, and I think it may have to do with the Python version perhaps? If you were using pixi to install:

pixi install --pypi llamabot

llamabot 0.4.4 will probably be the version that's resolved if the Python version in your pixi config manifest is an older version (I think 0.4.4 still supported 3.9, but can't remember anymore). 3.10 is the new minimum version for llamabot, and I think I made the conscious choice to support a maximum of 2 minor versions behind the latest released version of Python.

[willingc]

@ericmjl It was definitely a pixi issue though I'm not sure exactly where. I was using either 3.12 or 3.13 and installed with the same command you mentioned. I'm not sure why the resolver wasn't picking up the latest llamabot. I love pixi so I may give it another go with pixi.

Overall, I'm up and running smoothly with OpenAI.

[willingc]

Feel free to close this issue.

[ericmjl]

Ahh, I see, @willingc! It may have been another dependency, then. I think it may be helpful for me to specify the version number of llamabot in README.md and keep it up-to-date using bumpversion.