Analyze artwork images with AI (Unified Interface)

AI-powered functions for analyzing artwork images with provider abstraction. These functions work with both Google Gemini and OpenAI, allowing you to switch providers via the provider parameter or environment variable.

• art_about_ai(): Generate natural language descriptions for artwork profiles

• art_style_ai(): Extract 5-10 stylistic features as structured data

• classify_styles_ai(): Standardize free-form tags into consistent categories

Usage

art_about_ai(img_path, provider = NULL, ml = NULL, temp = 0.7, timeout = 60)

art_style_ai(img_path, provider = NULL, ml = NULL, temp = 0, timeout = 60)

classify_styles_ai(
  tags,
  exist_cats = NULL,
  provider = NULL,
  ml = NULL,
  temp = 0,
  timeout = 120
)

Arguments

img_path

Character. Path to local image file or URL. Supports common formats (PNG, JPEG, WebP). Images are resized to 1000px width for transmission.

provider

Character. AI provider to use: "gemini" or "openai". If NULL (default), uses ARTCURATOR_PROVIDER env var or defaults to gemini.

ml

Character. Model ID to use. Provider-specific (e.g., "gemini-2.5-flash", "gpt-5.1"). If NULL, uses provider default from env var.

temp

Numeric. Temperature setting (0-2). Lower values produce deterministic output; higher values produce varied responses.

timeout

Numeric. Request timeout in seconds. Default 60.

tags

Character vector. Style tags to classify (e.g., from art_style_ai() output or user-submitted tags). Each tag is mapped to a standardized category.

exist_cats

Character vector. Existing categories to prefer when classifying, ensuring consistency with previously categorized tags in the database. If NULL, the model creates appropriate categories.

Value

Character string with artwork description. Includes prompt_versions, model/modelVersion, and usage/usageMetadata attributes.

data.table with columns tag, tag_norm, and desc. Includes prompt_versions, model/modelVersion, and usage/usageMetadata attributes.

data.table with columns tag (original input) and category (standardized). Includes prompt_versions attribute for auditing. Row order matches input order.

Details

Provider Selection Logic:

1. If provider parameter is specified, use that provider

2. Else if ARTCURATOR_PROVIDER env var is set, use that

3. Else default to "gemini"

Note: artgemini and artopenai previously exported these functions but they have been removed. Use artcurator for all artwork analysis.

Functions

• art_about_ai(): Generate natural language artwork description

• art_style_ai(): Extract 5-10 key stylistic features from artwork

• classify_styles_ai(): Standardize style tags into consistent categories

Examples

if (FALSE) { # \dontrun{
# Auto-detect provider (defaults to gemini)
desc <- art_about_ai("path/to/artwork.png")

# Use Gemini explicitly
desc <- art_about_ai("path/to/artwork.png", provider = "gemini", temp = 0.7)

# Use OpenAI
desc <- art_about_ai("path/to/artwork.png", provider = "openai", ml = "gpt-5.1")
} # }
if (FALSE) { # \dontrun{
# Auto-detect provider
styles <- art_style_ai("path/to/artwork.png")

# Use Gemini with deterministic output
styles <- art_style_ai("path/to/artwork.png", provider = "gemini", temp = 0)
} # }
if (FALSE) { # \dontrun{
tags <- c("Photorealistic", "Detailed Rendering", "Vibrant Colors")

# Auto-detect provider
classified <- classify_styles_ai(tags)

# Use Gemini with existing categories
classified <- classify_styles_ai(tags,
  exist_cats = c("Realism", "Color Theory"),
  provider = "gemini"
)
} # }