cat.cog

Cognitive assessment scoring with LLMs. A domain wrapper around cat.stack that scores CERAD constructional praxis (drawn shape) assessments using vision-capable large language models.

cat.cog wraps the Python cat-cog package via reticulate.

Installation

# From R-universe (recommended)
install.packages("cat.cog",
                 repos = c("https://chrissoria.r-universe.dev",
                          "https://cloud.r-project.org"))

# Or from a local clone
devtools::install("path/to/cat.stack")
devtools::install("path/to/cat.cog")

# Install the Python backend (one-time setup)
# pip install cat-cog

Quick Start

Score a single drawing

library(cat.cog)

result <- cerad_drawn_score(
  shape       = "circle",
  image_input = "path/to/circle_drawing.png",
  api_key     = Sys.getenv("OPENAI_API_KEY")
)

Score a directory of drawings

results <- cerad_drawn_score(
  shape       = "cube",
  image_input = "path/to/cube_drawings/",
  api_key     = Sys.getenv("OPENAI_API_KEY"),
  user_model  = "claude-sonnet-4-5-20250929",
  model_source = "anthropic"
)

Functions

Function	Description
cerad_drawn_score()	Score CERAD constructional praxis drawings

Supported Shapes

• "circle" - Circle drawing assessment
• "diamond" - Diamond drawing assessment
• "rectangles" - Overlapping rectangles assessment
• "cube" - 3D cube drawing assessment

License

MIT