Title:	Interface for 'Google Gemini' API
Version:	0.16.0
Maintainer:	Jinhwan Kim <hwanistic@gmail.com>
Description:	Provides a comprehensive interface for Google Gemini API, enabling users to access and utilize Gemini Large Language Model (LLM) functionalities directly from R. This package facilitates seamless integration with Google Gemini, allowing for advanced language processing, text generation, and other AI-driven capabilities within the R environment. For more information, please visit https://ai.google.dev/docs/gemini_api_overview.
License:	MIT + file LICENSE
Depends:	R (≥ 4.1.0)
URL:	https://github.com/jhk0530/gemini.R
BugReports:	https://github.com/jhk0530/gemini.R/issues
Encoding:	UTF-8
Imports:	base64enc, cli, httr2, jsonlite, knitr, rstudioapi, tools
RoxygenNote:	7.3.2
Suggests:	testthat (≥ 3.0.0)
Config/testthat/edition:	3
Config/Needs/website:	rmarkdown
NeedsCompilation:	no
Packaged:	2025-07-22 13:31:08 UTC; KJH
Author:	Jinhwan Kim [aut, cre, cph], Maciej Nasinski [ctb]
Repository:	CRAN
Date/Publication:	2025-07-22 14:41:37 UTC

Add history for chating context

Description

Add history for chating context

Usage

addHistory(history, role = NULL, item = NULL)

Arguments

Value

The history of chat

Count Tokens for Gemini Content (Including Images)

Description

Calculates the token count for a given content, including text and image data, using the Vertex AI Gemini API.

Usage

countTokens(
  jsonkey = NULL,
  model_id = NULL,
  content = NULL,
  region = "us-central1"
)

Arguments

Value

A numeric value representing the token count of the content.

Examples

## Not run: 
library(gemini.R)

# For text content
key_file <- "YOURAPIKEY.json"
model <- "2.0-flash"
token_count_text <- countTokens(
  jsonkey = key_file,
  model_id = model,
  content = "Hello, world!"
)
print(token_count_text)

# For image content (assuming 'image.jpg' is in your working directory)
image_data <- base64enc::base64encode("image.jpg")
image_content <- list(data = image_data, mimeType = "image/jpeg")
token_count_image <- countTokens(
  jsonkey = key_file,
  model_id = model,
  content = image_content
)
print(token_count_image)

# For multiple content parts (text and image)
content_parts <- list(
  list(text = "This is the first part."),
  list(data = image_data, mimeType = "image/jpeg"),
  list(text = "This is the last part")
)
token_count_parts <- countTokens(
  jsonkey = key_file,
  model_id = model,
  content = content_parts
)
print(token_count_parts)

## End(Not run)

Generate text from text with Gemini

Description

Generate text from text with Gemini

Usage

gemini(
  prompt,
  model = "2.0-flash",
  temperature = 1,
  maxOutputTokens = 8192,
  topK = 40,
  topP = 0.95,
  seed = 1234,
  timeout = 60
)

Arguments

Value

Generated text or image

See Also

https://ai.google.dev/docs/gemini_api_overview#text_input

Examples

## Not run: 
library(gemini.R)
setAPI("YOUR_API_KEY")
gemini("Explain dplyr's mutate function")

## End(Not run)

Generate text from text with Gemini Vertex API

Description

Generate text from text with Gemini Vertex API

Usage

gemini.vertex(
  prompt = NULL,
  tokens = NULL,
  temperature = 1,
  maxOutputTokens = 8192,
  topK = 40,
  topP = 0.95,
  seed = 1234,
  timeout = 60
)

Arguments

Value

A character string containing the generated text.

See Also

https://ai.google.dev/docs/gemini_api_overview#text_input

Examples

## Not run: 
# token should be created before this. using the token.vertex() function
prompt <- "What is sachins Jersey number?"
gemini.vertex(prompt, tokens)

## End(Not run)

Analyze audio using Gemini

Description

This function sends audio to the Gemini API and returns a text description.

Usage

gemini_audio(
  audio = NULL,
  prompt = "Describe this audio",
  model = "2.0-flash",
  temperature = 1,
  maxOutputTokens = 8192,
  topK = 40,
  topP = 0.95,
  seed = 1234
)

Arguments

Details

The API key is now sent via the HTTP header x-goog-api-key instead of as a URL query parameter.

Value

A character vector containing the Gemini API's response.

Examples

## Not run: 
library(gemini.R)
setAPI("YOUR_API_KEY")
gemini_audio(audio = "YOUR_AUDIO_FILE")

## End(Not run)

Analyze Audio using Gemini Vertex API

Description

This function sends audio to the Gemini API and returns a text description.

Usage

gemini_audio.vertex(
  audio = NULL,
  prompt = "Describe this audio",
  tokens = NULL,
  temperature = 1,
  maxOutputTokens = 8192,
  topK = 40,
  topP = 0.95,
  seed = 1234
)

Arguments

Value

A character vector containing the Gemini API's description of the audio.

Multi-turn conversations (chat)

Description

Generate text from text with Gemini

Usage

gemini_chat(
  prompt,
  history = list(),
  model = "2.0-flash",
  temperature = 1,
  maxOutputTokens = 8192,
  topK = 40,
  topP = 0.95,
  seed = 1234
)

Arguments

Value

Generated text

See Also

https://ai.google.dev/docs/gemini_api_overview#chat

Examples

## Not run: 
library(gemini.R)
setAPI("YOUR_API_KEY")

chats <- gemini_chat("Pretend you're a snowman and stay in character for each")
print(chats$outputs)

chats <- gemini_chat("What's your favorite season of the year?", chats$history)
print(chats$outputs)

chats <- gemini_chat("How do you think about summer?", chats$history)
print(chats$outputs)

## End(Not run)

Summarize or analyze one or more local documents using Gemini API

Description

Summarize, compare, or analyze the content of one or more local documents (PDF, TXT, HTML, etc.) using the Gemini API.

Usage

gemini_docs(
  pdf_path,
  prompt,
  type = "PDF",
  api_key = Sys.getenv("GEMINI_API_KEY"),
  large = FALSE,
  local = FALSE
)

Arguments

Details

This function encodes one or more local files, sends them along with a prompt to the Gemini API, and returns the generated summary or response.

Value

The summary or response text from Gemini.

See Also

https://ai.google.dev/gemini-api/docs/document-processing?lang=rest

Examples

## Not run: 
gemini_docs(
  pdf_path = c("doc1.pdf", "doc2.pdf"),
  prompt = "Compare these documents",
  type = "PDF"
)

## End(Not run)

Summarize or analyze documents using Vertex AI Gemini

Description

Summarize, compare, or analyze the content of one or more documents (PDF, TXT, HTML, etc.) using Vertex AI Gemini.

Usage

gemini_docs.vertex(
  file_uri,
  prompt,
  mime_type = "application/pdf",
  tokens = NULL,
  temperature = 1,
  maxOutputTokens = 8192,
  topK = 40,
  topP = 0.95,
  seed = 1234
)

Arguments

Value

The summary or response text from Gemini Vertex.

See Also

https://cloud.google.com/vertex-ai/docs/generative-ai/multimodal/send-request-document

Examples

## Not run: 
tokens <- token.vertex()
gemini_docs.vertex(
  file_uri = "gs://cloud-samples-data/generative-ai/pdf/2403.05530.pdf",
  prompt = "Summarize this document.",
  mime_type = "application/pdf",
  tokens = tokens
)

## End(Not run)

Interact with Vertex AI Model Garden

Description

This function sends a PDF file to the Vertex AI Model Garden (Mistral model) for processing, such as OCR. The PDF is encoded as base64 and sent to the rawPredict endpoint. The function is designed for future extension to support other document types and tasks.

Usage

gemini_garden(token, project_id, pdf_path)

Arguments

Details

The PDF file is read and encoded as base64, then sent to the Vertex AI rawPredict endpoint for processing using a Mistral model. This function is structured for future extension to support other document types and model tasks available in Vertex AI Model Garden.

For more information about available models, endpoints, and supported tasks, see Vertex AI Model Garden documentation.

Value

A parsed list containing the results from the Vertex AI API (e.g., OCR results).

See Also

https://cloud.google.com/vertex-ai/generative-ai/docs/model-garden/explore-models

Examples

## Not run: 
# Issue a token using token.vertex() first
my_token <- token.vertex(
  jsonkey = "your-service-account.json",
  region = "us-central1",
  model_id = "mistral-ocr-2505"
)
result <- gemini_garden(
  token = my_token,
  project_id = "your-project-id",
  pdf_path = "sample.pdf"
)
print(result)

## End(Not run)

Generate text from text and image with Gemini

Description

Generate text from text and image with Gemini

Usage

gemini_image(
  image = NULL,
  prompt = "Explain this image",
  model = "2.0-flash",
  temperature = 1,
  maxOutputTokens = 8192,
  topK = 40,
  topP = 0.95,
  seed = 1234,
  type = "png"
)

Arguments

Details

The API key is now sent via the x-goog-api-key HTTP header instead of as a URL query parameter.

Value

Generated text

See Also

https://ai.google.dev/docs/gemini_api_overview#text_image_input

Examples

## Not run: 
library(gemini.R)
setAPI("YOUR_API_KEY")
gemini_image(image = system.file("docs/reference/figures/image.png", package = "gemini.R"))

## End(Not run)

Generate text from text and image with Gemini Vertex API

Description

Generate text from text and image with Gemini Vertex API

Usage

gemini_image.vertex(
  image = NULL,
  prompt = "Explain this image",
  type = "png",
  tokens = NULL,
  temperature = 1,
  maxOutputTokens = 8192,
  topK = 40,
  topP = 0.95,
  seed = 1234
)

Arguments

Value

A character vector containing Gemini's description of the image.

Generate narrative description for an input using Gemini

Description

Generate a narrative description for a given input (e.g., table, figure) by converting it to a suitable format and sending it to Gemini.

Usage

gemini_narrative(input, type = "table", prompt = NULL, ...)

Arguments

Value

Narrative description generated by Gemini

See Also

gemini

Examples

## Not run: 
# Example data.frame
table_data <- data.frame(
  sort = c(1, 1, 2, 2, 2, 3, 3, 3, 4, 4),
  category = c(
    "Gender", "Gender", "Age Group", "Age Group", "Age Group",
    "Race", "Race", "Race", "Ethnicity", "Ethnicity"
  ),
  type = c(
    "F", "M", "<65", ">80", "65-80",
    "AMERICAN INDIAN OR ALASKA NATIVE", "BLACK OR AFRICAN AMERICAN", "WHITE",
    "HISPANIC OR LATINO", "NOT HISPANIC OR LATINO"
  ),
  Placebo = c(53, 33, 14, 30, 42, NA, 8, 78, 3, 83),
  Xanomeline_High_Dose = c(40, 44, 11, 18, 55, 1, 9, 74, 3, 81),
  Xanomeline_Low_Dose = c(50, 34, 8, 29, 47, NA, 6, 78, 6, 78),
  stringsAsFactors = FALSE # Do not convert strings to factors
)
gemini_narrative(table_data)

## End(Not run)

Generate text with real-time information using Google Search (Grounding)

Description

Generate text responses that include up-to-date information from Google Search

Usage

gemini_search(
  prompt,
  temperature = 1,
  maxOutputTokens = 8192,
  topK = 40,
  topP = 0.95,
  seed = 1234
)

Arguments

Details

The API key is now sent via the HTTP header x-goog-api-key instead of as a URL query parameter.

Value

Generated text with real-time information from Google Search

See Also

https://ai.google.dev/docs/search_retrieval

Examples

## Not run: 
library(gemini.R)
setAPI("YOUR_API_KEY")
gemini_search("What is the current Google stock price?")

## End(Not run)

Generate text with real-time information using Gemini (Retrieval)

Description

Generate text responses with simplified access to Gemini models

Usage

gemini_searchR(
  prompt,
  model = "1.5-flash",
  temperature = 1,
  maxOutputTokens = 8192,
  topK = 40,
  topP = 0.95,
  seed = 1234
)

Arguments

Value

Generated text response from the Gemini model

See Also

https://ai.google.dev/docs/gemini_api_overview#text_input

Examples

## Not run: 
library(gemini.R)
setAPI("YOUR_API_KEY")
gemini_searchR("Who won the latest F1 grand prix?")

## End(Not run)

Generate structured response from Gemini

Description

Returns a structured (JSON) response from the Gemini API.

Usage

gemini_structured(
  prompt,
  schema,
  model = "2.5-flash",
  temperature = 1,
  maxOutputTokens = 8192,
  topK = 40,
  topP = 0.95,
  seed = 1234,
  timeout = 60
)

Arguments

Value

A structured list (parsed JSON).

Examples

## Not run: 
schema <- list(
  type = "ARRAY",
  items = list(
    type = "OBJECT",
    properties = list(
      recipeName = list(type = "STRING"),
      ingredients = list(
        type = "ARRAY",
        items = list(type = "STRING")
      )
    ),
    propertyOrdering = c("recipeName", "ingredients")
  )
)
gemini_structured(
  "List a few popular cookie recipes, and include the amounts of ingredients.",
  schema
)

## End(Not run)

Generate Roxygen Documentation

Description

Generates Roxygen2 documentation for an R function based on the currently selected code.

Usage

gen_docs(prompt = NULL)

Arguments

Value

Invisibly returns the generated documentation string, but primarily inserts the text into the RStudio console.

Examples

## Not run: 
# Select your function code in the editor, then run:
gen_docs()

# For custom instructions:
gen_docs("Generate minimal Roxygen docs for this function")

## End(Not run)

Generate and save image using Gemini

Description

Generate an image using Gemini's image generation capabilities and save it to a file

Usage

gen_image(
  prompt,
  filename = "gemini_image.png",
  overwrite = TRUE,
  model = "2.0-flash-exp-image-generation",
  temperature = 1,
  maxOutputTokens = 8192,
  topK = 40,
  topP = 0.95,
  seed = 1234
)

Arguments

Value

The path to the saved file or NULL if an error occurred

Examples

## Not run: 
library(gemini.R)
setAPI("YOUR_API_KEY")
gen_image("Create an image of a cat wearing sunglasses")

## End(Not run)

Generate Unit Tests for R Functions

Description

Generates unit test code for an R function using the Gemini AI model.

Usage

gen_tests(prompt = NULL)

Arguments

Value

Invisibly returns the generated test code, but primarily inserts it into the RStudio console.

Examples

## Not run: 
# Select your function code in the editor, then run:
gen_tests()

# For custom instructions:
gen_tests("Generate comprehensive testthat tests with edge cases")

## End(Not run)

Set Gemini API Key

Description

Sets the Gemini API key as an environment variable for use in API calls.

Usage

setAPI(api_key)

Arguments

Value

No return value, called for side effects.

Note

Please be aware you have to agree to the terms of service of the API provider. Any app that uses the API key is subject to the terms of service. Also, please be aware that the API key is a sensitive information.

See Also

https://makersuite.google.com/app/apikey

Examples

## Not run: 
setAPI("YOUR_API_KEY")

## End(Not run)

Store API key in local environment file

Description

Saves the API key to a local .Renviron file for persistent access across R sessions

Usage

setEnv(api_key, overwrite = TRUE, install_message = TRUE)

Arguments

Value

No return value, called for side effects.

See Also

setAPI which sets the API key for the current session only

Examples

## Not run: 
setEnv("your_api_key")

## End(Not run)

Generate Gemini Access Token and Endpoint URL

Description

Generates an access token for the Gemini model and constructs the corresponding endpoint URL.

Usage

token.vertex(
  jsonkey = NULL,
  model_id = NULL,
  expTime = 3600,
  region = "us-central1"
)

Arguments

Value

A list containing:

Examples

## Not run: 
library(gemini.R)
tokens <- token.vertex(jsonkey = "YOURAPIKEY.json", model_id = "1.5-flash")

# Specify a different region
tokens <- token.vertex(jsonkey = "YOURAPIKEY.json", model_id = "1.5-flash", region = "europe-west4")

## End(Not run)

Trim whitespace from string

Description

Removes leading and trailing whitespace from a string

Usage

trim(x)

Arguments

Value

Character string with leading and trailing whitespace removed

Validate Gemini API parameters

Description

Helper function to validate parameters for Gemini API calls

Usage

validate_params(
  prompt,
  model,
  temperature = 1,
  topP = 0.95,
  topK = 40,
  seed = 1234,
  api_key = TRUE,
  tokens = NULL
)

Arguments

Value

TRUE if all validations pass, otherwise the function stops execution with an error message