Smithers

An LLM-powered CLI that simulates job interviews based on your resume.

Example

pipx run smithers-llm [RESUME_PATH] --role=[ROLE]

Installation

With pip:

pip install -g smithers-llm

With pipx (automatically adds smithers to PATH):

pipx install smithers-llm

Please notice that the user is expected to run the model separately as well.

So far the only model supported is Llama3.1

You can easily self-host it with Ollama. After installing it, pull the model and keep it running on the background with the command below:

ollama run llama3.1

Usage

With smithers installed on the current env:

python -m smithers-llm [RESUME_PATH] --role=[ROLE]

With smithers installed globally and defined on PATH:

smithers-llm [RESUME_PATH] --role=[ROLE]

With pipx:

pipx run smithers-llm [RESUME_PATH] --role=[ROLE]

Learnings

• LangChain
  • Structured Outputs
• LangGraph
  • Human in the Loop
  • Checkpointer
• Python packaging with SetupTools and Build
  • Differences between project configurations files such as the modern pyproject.toml and the battle-tested setup.py and setup.cfg
• Publishing to PyPI and TestPyPI with Twine
• Creating CLIs with Click
  • Handling different kinds of options: parameters and arguments
• Experimenting with Jupyter Notebooks
  • Research whether it was worth commiting the notebooks, although they might be reasonable for documentation purposes, they weren't for the way I used them in this project, to scratch the general ideas up right in the beginning
• Self-reflection if RAG was necessary here, the answer I came to was NO. The files are too small.
• Creating a GitHub Pages workflow to publish project to PyPI on pushing tagged commits to GH.

Fun fact

The repo was named after a character from the show "The Simpsons".

Smithers conducts a job interview with Homer on the episode "I Married Marge".