Biological analysis with Claude code

This repo contains simple command-line-interface (CLI) tools that help research scientists at UPSIDE Foods with local analyses on their laptops. These tools will enable them to use Claude Code and an ETL pipeline to automate their current data pre-processing, visualization, and local analysis efforts.

Tool summary

• Tool 1 (claude_model.py): A simple CLI tool to switch between different Claude models configured in aws bedrock.
• Tool 2 (name_TBA): A (set of) simple CLI tool(s) used to extract, transform, and load (ETL) data that is uploaded by the research scientist as (format) file for local analysis. This tool helps satisfy the following needs by Yulong as mentioned in Slack:
  • Script grabs semi-structured data from a designated folder -> model API call to organize variables into a standardized format -> then uses hard coded script to graph based on a few predetermined templates.
  • This could reduce the most time-consuming part of organizing data, but could still provide consistently formatted graphs.

Prerequisites and dependencies

1. Ensure you have Python version >3.8.

2. Clone repo and cd to folder

3. Set up virtual environment and install dependencies:

python -m venv venv
pip install -r requirements.txt

How to use Tool 1 (claude_model.py)

1. Open terminal
2. cd to this repo location
3. Run the script in the terminal to select 1 of the # models:

• python3 claude_model.py fast
• python3 claude_model.py balanced
• python3 claude_model.py accurate

How to use Tool 2 (name_TBA.py)

Notes

01/11/2025:

• For tool 1, need more info from IT team about:
  • AWS bedrock model IDs
• For tool 2, need more info from Yulong about:
  • File type that his data is stored in.
  • Variables, column names, and data format in the file.
  • His current pre-process and analysis he's doing.
  • Graphs he wants to see.