| Title: | Computer Vision with Large Language Models |
| Version: | 0.1.0 |
| Description: | Make computer vision tasks approachable in R by leveraging Large Language Models. Providing fine-tuned prompts, boilerplate functions, and input/output helpers for common computer vision workflows, such as classifying and describing images. Functions are designed to take images as input and return structured data, helping users build practical applications with minimal code. |
| License: | MIT + file LICENSE |
| Encoding: | UTF-8 |
| RoxygenNote: | 7.3.2 |
| Depends: | R (≥ 4.1.0) |
| Imports: | dplyr, ellmer, graphics, gt, gtExtras, imager, jsonlite, magick, ollamar, rlang, stringr, tidyr, utils |
| Suggests: | knitr, quarto, shiny, bslib, rmarkdown, purrr, mirai, tibble, tictoc, httr |
| VignetteBuilder: | quarto, knitr |
| URL: | https://frankiethull.github.io/kuzco/, https://github.com/frankiethull/kuzco |
| BugReports: | https://github.com/frankiethull/kuzco/issues |
| NeedsCompilation: | no |
| Packaged: | 2026-01-21 20:55:30 UTC; frankiethull |
| Author: | Frank Hull [aut, cre, cph],
Johannes Breuer 
[ctb],
Jordi Rosell
[ctb] |
| Maintainer: | Frank Hull <frankiethull@gmail.com> |
| Repository: | CRAN |
| Date/Publication: | 2026-01-26 16:50:02 UTC |
kuzco: Computer Vision with Large Language Models
Description
Make computer vision tasks approachable in R by leveraging Large Language Models. Providing fine-tuned prompts, boilerplate functions, and input/output helpers for common computer vision workflows, such as classifying and describing images. Functions are designed to take images as input and return structured data, helping users build practical applications with minimal code.
Author(s)
Maintainer: Frank Hull frankiethull@gmail.com [copyright holder]
Other contributors:
Johannes Breuer johannes.breuer@gesis.org (ORCID) [contributor]
Jordi Rosell jroselln@gmail.com (ORCID) [contributor]
See Also
Useful links:
Report bugs at https://github.com/frankiethull/kuzco/issues
chat ellmer helper (predates ellmer::chat)
Description
a minimal wrapper function to switch which provider is used for each llm_image* function when ellmer backend is selected, ollamar only supports ollama
Usage
chat_ellmer(provider = "ollama")
Arguments
provider |
a provider, such as "ollama", or "claude", or "github" |
Value
which ellmer function (provider) to use for kuzco llm_image_* when backend is ellmer
edit prompt
Description
edit a listed prompt installed with kuzco
Usage
edit_prompt(prompt)
Arguments
prompt |
a prompt from list_prompts() |
Value
a prompt markdown file to edit
Examples
## Not run:
edit_prompt("system-prompt-alt-text.md")
## End(Not run)
shiny kuzco app
Description
a simple wrapper of kuzco to make computer vision for everyone. few-shot via frank hull and shiny assistant (https://gallery.shinyapps.io/assistant/)
Usage
kuzco_app()
Value
a shiny app instance as a playground for local llms
Examples
## Not run:
kuzco_app()
## End(Not run)
list prompts
Description
list prompts installed with kuzco
Usage
list_prompts()
Value
a list of prompts stored within kuzco
Examples
list_prompts()
Image Alt Text using LLMs
Description
Image Alt Text using LLMs
Usage
llm_image_alt_text(
llm_model = "qwen2.5vl",
image = system.file("img/test_img.jpg", package = "kuzco"),
backend = "ellmer",
additional_prompt = "",
provider = "ollama",
language = "English",
...
)
Arguments
llm_model |
a local LLM model either pulled from ollama or hosted |
image |
a local image path that has a jpeg, jpg, or png |
backend |
either 'ellmer' or 'ollamar', note that 'ollamar' suggests structured outputs while 'ellmer' enforces structured outputs |
additional_prompt |
text to append to the image prompt |
provider |
for |
language |
a language to guide the LLM model outputs |
... |
a pass through for other generate args and model args like temperature. set the temperature to 0 for more deterministic output |
Value
a df with text
Examples
llm_image_alt_text(
llm_model = "qwen2.5vl",
image = system.file("img/test_img.jpg", package = "kuzco"),
backend = 'ellmer',
additional_prompt = "",
provider = "ollama",
language = "English"
)
Image Classification using LLMs
Description
Image Classification using LLMs
Usage
llm_image_classification(
llm_model = "qwen2.5vl",
image = system.file("img/test_img.jpg", package = "kuzco"),
backend = "ellmer",
additional_prompt = "",
provider = "ollama",
language = "English",
...
)
Arguments
llm_model |
a local LLM model either pulled from ollama or hosted |
image |
a local image path that has a jpeg, jpg, or png |
backend |
either 'ollamar' or 'ellmer', note that 'ollamar' suggests structured outputs while 'ellmer' enforces structured outputs |
additional_prompt |
text to append to the image prompt |
provider |
for |
language |
a language to guide the LLM model outputs |
... |
a pass through for other generate args and model args like temperature |
Value
a df with image_classification, primary_object, secondary_object, image_description, image_colors, image_proba_names, image_proba_values
Examples
llm_image_classification(
llm_model = "qwen2.5vl",
image = system.file("img/test_img.jpg", package = "kuzco"),
backend = 'ellmer',
additional_prompt = "",
provider = "ollama",
language = "English"
)
Customized Vision using LLMs
Description
Customized Vision using LLMs
Usage
llm_image_custom(
llm_model = "qwen2.5vl",
image = system.file("img/test_img.jpg", package = "kuzco"),
backend = "ellmer",
system_prompt = "You are a terse assistant in computer vision sentiment.",
image_prompt = "return JSON describing image, do not include json or backticks",
example_df = NULL,
provider = "ollama",
...
)
Arguments
llm_model |
a local LLM model either pulled from ollama or hosted |
image |
a local image path that has a jpeg, jpg, or png |
backend |
either 'ollamar' or 'ellmer' |
system_prompt |
overarching assistant description, please note that the LLM should be told to return as JSON while kuzco will handle the conversions to and from JSON |
image_prompt |
anything you want to really remind the llm about. |
example_df |
an example data.frame to show the llm what you want returned note this will be converted to JSON for the LLM. |
provider |
for |
... |
a pass through for other generate args and model args like temperature |
Value
a customized return based on example_df for custom control
Examples
llm_image_custom(
llm_model = "qwen2.5vl",
image = system.file("img/test_img.jpg", package = "kuzco"),
backend = "ellmer",
system_prompt = "You are a terse assistant in computer vision sentiment.",
image_prompt = "return JSON describing image, do not include json or backticks",
example_df = NULL,
provider = "ollama"
)
Image OCR for Text Extraction using LLMs
Description
Image OCR for Text Extraction using LLMs
Usage
llm_image_extract_text(
llm_model = "qwen2.5vl",
image = system.file("img/text_img.jpg", package = "kuzco"),
backend = "ellmer",
additional_prompt = "",
provider = "ollama",
language = "English",
...
)
Arguments
llm_model |
a local LLM model either pulled from ollama or hosted |
image |
a local image path that has a jpeg, jpg, or png |
backend |
either 'ellmer' or 'ollamar', note that 'ollamar' suggests structured outputs while 'ellmer' enforces structured outputs |
additional_prompt |
text to append to the image prompt |
provider |
for |
language |
a language to guide the LLM model outputs |
... |
a pass through for other generate args and model args like temperature. set the temperature to 0 for more deterministic output |
Value
a df with text and a confidence score
Examples
llm_image_extract_text(
llm_model = "qwen2.5vl",
image = system.file("img/test_img.jpg", package = "kuzco"),
backend = 'ellmer',
additional_prompt = "",
provider = "ollama",
language = "English"
)
Image Recognition using LLMs
Description
Image Recognition using LLMs
Usage
llm_image_recognition(
llm_model = "qwen2.5vl",
image = system.file("img/test_img.jpg", package = "kuzco"),
recognize_object = "face",
backend = "ellmer",
additional_prompt = "",
provider = "ollama",
language = "English",
...
)
Arguments
llm_model |
a local LLM model either pulled from ollama or hosted |
image |
a local image path that has a jpeg, jpg, or png |
recognize_object |
an item you want to LLM to look for |
backend |
either 'ollamar' or 'ellmer', note that 'ollamar' suggests structured outputs while 'ellmer' enforces structured outputs |
additional_prompt |
text to append to the image prompt |
provider |
for |
language |
a language to guide the LLM model outputs |
... |
a pass through for other generate args and model args like temperature. set the temperature to 0 for more deterministic output |
Value
a df with object_recognized, object_count, object_description, object_location
Examples
llm_image_recognition(
llm_model = "qwen2.5vl",
image = system.file("img/test_img.jpg", package = "kuzco"),
recognize_object = "nose",
backend = 'ellmer',
additional_prompt = "",
provider = "ollama",
language = "English"
)
Image Sentiment using LLMs
Description
Image Sentiment using LLMs
Usage
llm_image_sentiment(
llm_model = "qwen2.5vl",
image = system.file("img/test_img.jpg", package = "kuzco"),
backend = "ellmer",
additional_prompt = "",
provider = "ollama",
language = "English",
...
)
Arguments
llm_model |
a local LLM model either pulled from ollama or hosted |
image |
a local image path that has a jpeg, jpg, or png |
backend |
either 'ollamar' or 'ellmer', note that 'ollamar' suggests structured outputs while 'ellmer' enforces structured outputs |
additional_prompt |
text to append to the image prompt |
provider |
for |
language |
a language to guide the LLM model outputs |
... |
a pass through for other generate args and model args like temperature. set the temperature to 0 for more deterministic output |
Value
a df with image_sentiment, image_score, sentiment_description, image_keywords
Examples
llm_image_sentiment(
llm_model = "qwen2.5vl",
image = system.file("img/test_img.jpg", package = "kuzco"),
backend = 'ellmer',
additional_prompt = "",
provider = "ollama",
language = "English"
)
View Images quickly and easily
Description
View Images quickly and easily
Usage
view_image(image = system.file("img/test_img.jpg", package = "kuzco"))
Arguments
image |
an image to view |
Value
a plot of the image in a Plots pane
Examples
view_image(image = system.file("img/test_img.jpg", package = "kuzco"))
view llm results as a tidy great table
Description
view llm results as a tidy great table
Usage
view_llm_results(llm_results)
Arguments
llm_results |
results from one of the llm_image_* functions |
Value
a great table to view the results neatly
Examples
## Not run:
view_llm_results(llm_image_alt_text())
## End(Not run)