How waughb/vision-streamlit works
Project type and purpose
• From the top‐level file tree and README, this is a small full-stack prototype: a Streamlit web application (“vision-streamlit”) that exercises the llama3.2-vision model via Ollama, combined with a local data-processing submodule (“vessel-information”).
File structure
• Top level:
• app.py – the Streamlit entry point (frontend + API in one)
• .gitignore, .gitattributes, README.md
• Subdirectory “vessel-information/”:
• data ingestion & processing scripts (download-ais.py, extract-all-zip.py, move-data.py, data-exploration.py)
• static CSV lookups (cargo-lookup.csv, vessel-type-lookup.csv)
• its own requirements.txt
• a README.md describing how to run the data pipeline
README for additional insights
• The app uses Ollama to pull and run llama3.2-vision locally
• Setup steps: create a venv, install Streamlit and Ollama, then streamlit run app.py
Detail
• File paths under the Data Store: /vessel-information/*.csv, /vessel-information/raw_zips/
• Call out that the data pipeline scripts can be run standalone or invoked on-demand by the Streamlit app