Package 'gptstudio'

Title: Use Large Language Models Directly in your Development Environment
Description: Large language models are readily accessible via API. This package lowers the barrier to use the API inside of your development environment. For more on the API, see <https://platform.openai.com/docs/introduction>.
Authors: Michel Nivard [aut, cph], James Wade [aut, cre, cph] (ORCID: <https://orcid.org/0000-0002-9740-1905>), Samuel Calderon [aut] (ORCID: <https://orcid.org/0000-0001-6847-1210>)
Maintainer: James Wade <[email protected]>
License: MIT + file LICENSE
Version: 0.4.0.9018
Built: 2026-06-04 08:10:18 UTC
Source: https://github.com/michelnivard/gptstudio

Help Index

Message Buffer

Description

A message buffer

Public fields

value

Placeholder for the message

Methods

Public methods

Method add_chunk()

Add a message chunk

Usage
Buffer$add_chunk(chunk)
Arguments
chunk

A message chunk

Method clone()

The objects of this class are cloneable with this method.

Usage
Buffer$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Chat Interface for gptstudio

Description

This function provides a high-level interface for communicating with various services and models supported by gptstudio. It orchestrates the creation, configuration, and execution of a request based on user inputs and options set for gptstudio. The function supports a range of tasks from text generation to code synthesis and can be customized according to skill level and coding style preferences.

Usage

chat(
  prompt,
  service = getOption("gptstudio.service"),
  history = list(list(role = "system", content = "You are an R chat assistant")),
  stream = FALSE,
  model = getOption("gptstudio.model"),
  skill = getOption("gptstudio.skill"),
  style = getOption("gptstudio.code_style", "no preference"),
  task = getOption("gptstudio.task", "coding"),
  custom_prompt = NULL,
  process_response = FALSE,
  session = NULL,
  ...
)

Arguments

prompt

A string containing the initial prompt or question to be sent to the model. This is a required parameter.
service

The AI service to be used for the request. If not explicitly provided, this defaults to the value set in getOption("gptstudio.service"). If the option is not set, make sure to provide this parameter to avoid errors.
history

An optional parameter that can be used to include previous interactions or context for the current session. Defaults to a system message indicating "You are an R chat assistant".
stream

A logical value indicating whether the interaction should be treated as a stream for continuous interactions. If not explicitly provided, this defaults to the value set in getOption("gptstudio.stream").
model

The specific model to use for the request. If not explicitly provided, this defaults to the value set in getOption("gptstudio.model").
skill

A character string indicating the skill or capability level of the user. This parameter allows for customizing the behavior of the model to the user. If not explicitly provided, this defaults to the value set in getOption("gptstudio.skill").
style

The coding style preferred by the user for code generation tasks. This parameter is particularly useful when the task involves generating code snippets or scripts. If not explicitly provided, this defaults to the value set in getOption("gptstudio.code_style").
task

The specific type of task to be performed, ranging from text generation to code synthesis, depending on the capabilities of the model. If not explicitly provided, this defaults to the value set in getOption("gptstudio.task").
custom_prompt

An optional parameter that provides a way to extend or customize the initial prompt with additional instructions or context.
process_response

A logical indicating whether to process the model's response.
session

An optional parameter for a shiny session object.
...

Reserved for future use.

Value

Depending on the task and processing, the function returns the response from the model, which could be text, code, or any other structured output defined by the task and model capabilities. The precise format and content of the output depend on the specified options and the capabilities of the selected model.

Examples

## Not run: 
# Basic usage with a text prompt:
result <- chat("What is the weather like today?")

# Advanced usage with custom settings, assuming appropriate global options are set:
result <- chat(
  prompt = "Write a simple function in R",
  skill = "advanced",
  style = "tidyverse",
  task = "coding"
)

# Usage with explicit service and model specification:
result <- chat(
  prompt = "Explain the concept of tidy data in R",
  service = "openai",
  model = "gpt-4-turbo-preview",
  skill = "intermediate",
  task = "general"
)

## End(Not run)

Create system prompt

Description

This function creates a customizable system prompt based on user-defined parameters such as coding style, skill level, and task. It supports customization for specific use cases through a custom prompt option.

Usage

chat_create_system_prompt(
  style = getOption("gptstudio.code_style"),
  skill = getOption("gptstudio.skill"),
  task = getOption("gptstudio.task"),
  custom_prompt = getOption("gptstudio.custom_prompt"),
  in_source = FALSE
)

Arguments

style

A character string indicating the preferred coding style. Valid values are "tidyverse", "base", "no preference". Defaults to getOption(gptstudio.code_style).
skill

The self-described skill level of the programmer. Valid values are "beginner", "intermediate", "advanced", "genius". Defaults to getOption(gptstudio.skill).
task

The specific task to be performed: "coding", "general", "advanced developer", or "custom". This influences the generated system prompt. Defaults to "coding".
custom_prompt

An optional custom prompt string to be utilized when task is set to "custom". Default is NULL.
in_source

A logical indicating whether the instructions are intended for use in a source script. This parameter is required and must be explicitly set to TRUE or FALSE. Default is FALSE.

Value

Returns a character string that forms a system prompt tailored to the specified parameters. The string provides guidance or instructions based on the user's coding style, skill level, and task.

Examples

## Not run: 
chat_create_system_prompt(in_source = TRUE)
chat_create_system_prompt(
  style = "tidyverse",
  skill = "advanced",
  task = "coding",
  in_source = FALSE
)

## End(Not run)

Append to chat history

Description

This appends a new response to the chat history

Usage

chat_history_append(history, role, content, name = NULL)

Arguments

history

List containing previous responses.
role

Author of the message. One of c("user", "assistant")
content

Content of the message. If it is from the user most probably comes from an interactive input.
name

Name for the author of the message. Currently used to support rendering of help pages

Value

list of chat messages

Default chat message

Description

Default chat message

Usage

chat_message_default(translator = create_translator())

Arguments

translator

A Translator from shiny.i18n::Translator

Value

A default chat message for welcoming users.

Check API Connection

Description

This generic function checks the API connection for a specified service by dispatching to related methods.

Usage

check_api_connection(service, api_key = "", model = NULL)

Arguments

service

The name of the API service for which the connection is being checked.
api_key

The API key used for authentication.
model

The service's model to check

Value

A logical value indicating whether the connection was successful.

Chat App Theme

Description

Create a bslib theme that matches the user's RStudio IDE theme.

Usage

create_chat_app_theme(ide_colors = get_ide_theme_info())

Arguments

ide_colors

List containing the colors of the IDE theme.

Value

A bslib theme

Generate text using Azure OpenAI's API

Description

Use this function to generate text completions using OpenAI's API.

Usage

create_completion_azure_openai(
  prompt,
  task = Sys.getenv("AZURE_OPENAI_TASK"),
  base_url = Sys.getenv("AZURE_OPENAI_ENDPOINT"),
  deployment_name = Sys.getenv("AZURE_OPENAI_DEPLOYMENT_NAME"),
  api_key = Sys.getenv("AZURE_OPENAI_API_KEY"),
  api_version = Sys.getenv("AZURE_OPENAI_API_VERSION")
)

Arguments

prompt

a list to use as the prompt for generating completions
task

a character string for the API task (e.g. "completions"). Defaults to the Azure OpenAI task from environment variables if not specified.
base_url

a character string for the base url. It defaults to the Azure OpenAI endpoint from environment variables if not specified.
deployment_name

a character string for the deployment name. It will default to the Azure OpenAI deployment name from environment variables if not specified.
api_key

a character string for the API key. It will default to the Azure OpenAI API key from your environment variables if not specified.
api_version

a character string for the API version. It will default to the Azure OpenAI API version from your environment variables if not specified.

Value

a list with the generated completions and other information returned by the API

Chat message colors in RStudio

Description

This returns a list of color properties for a chat message

Usage

create_ide_matching_colors(
  role = c("user", "assistant"),
  ide_colors = get_ide_theme_info()
)

Arguments

role

The role of the message author
ide_colors

List containing the colors of the IDE theme.

Value

list

Internationalization for the ChatGPT addin

Description

The language can be set via options("gptstudio.language" = "<language>") (defaults to "en").

Usage

create_translator(language = getOption("gptstudio.language"))

Arguments

language

The language to be found in the translation JSON file.

Value

A Translator from shiny.i18n::Translator

Create Chat Client for Different API Providers

Description

This function provides a generic interface for creating chat clients for different API providers (e.g., OpenAI, HuggingFace, Google AI Studio). It dispatches the actual client creation to the relevant method based on the class of the skeleton argument.

Usage

ellmer_chat(skeleton, all_turns)

Arguments

skeleton

A gptstudio_request_skeleton object containing API configuration
all_turns

A list of conversation turns formatted for the ellmer package

Value

An ellmer chat client object for the specific API provider

List supported service providers

Description

List the names of all providers/vendors supported by the gptstudio chat app.

Usage

get_all_available_services()

Value

A character vector

Examples

get_all_available_services()

List supported models

Description

Get a list of the models supported by the selected service.

Usage

get_available_models(service)

Arguments

service

The API service

Value

A character vector

See Also

get_all_available_services()

Examples

## Not run: 
get_available_models()

## End(Not run)

Get IDE Theme Information

Description

Retrieves the current RStudio IDE theme information including whether it is a dark theme, and the background and foreground colors in hexadecimal format.

Usage

get_ide_theme_info()

Value

A list with the following components:

is_dark

A logical indicating whether the current IDE theme is dark.
bg

A character string representing the background color of the IDE theme in hex format.
fg

A character string representing the foreground color of the IDE theme in hex format.

If RStudio is unavailable, returns the fallback theme details.

Examples

theme_info <- get_ide_theme_info()
print(theme_info)

a function that determines the appropriate directory to cache a token

Description

a function that determines the appropriate directory to cache a token

Usage

gptstudio_cache_directory()

Run GPTStudio Chat App

Description

This function initializes and runs the Chat GPT Shiny App as a background job in RStudio and opens it in the viewer pane or browser window.

Usage

gptstudio_chat(host = getOption("shiny.host", "127.0.0.1"))

Arguments

host

A character string specifying the host on which to run the app. Defaults to the value of getOption("shiny.host", "127.0.0.1").

Details

The function performs the following steps:

  1. Verifies that RStudio API is available.

  2. Finds an available port for the Shiny app.

  3. Creates a temporary directory for the app files.

  4. Runs the app as a background job in RStudio.

  5. Opens the app in the RStudio viewer pane or browser window.

Value

This function does not return a value. It runs the Shiny app as a side effect.

Note

This function is designed to work within the RStudio IDE and requires the rstudioapi package.

Examples

## Not run: 
gptstudio_chat()

## End(Not run)

ChatGPT in Source

Description

Call this function as a Rstudio addin to ask GPT to improve spelling and grammar of selected text.

Usage

gptstudio_chat_in_source_addin()

Value

This function has no return value.

Examples

# Select some text in a source file
# Then call the function as an RStudio addin
## Not run: 
gptstudio_chat_in_source()

## End(Not run)

Comment Code Addin

Description

Call this function as a Rstudio addin to ask GPT to add comments to your code

Usage

gptstudio_comment_code()

Value

This function has no return value.

Examples

# Open a R file in Rstudio
# Then call the function as an RStudio addin
## Not run: 
gptstudio_comment_code()

## End(Not run)

Create a Request Skeleton

Description

This function dynamically creates a request skeleton for different AI text generation services.

Usage

gptstudio_create_skeleton(
  service = "openai",
  prompt = "Name the top 5 packages in R.",
  history = list(list(role = "system", content = "You are an R chat assistant")),
  stream = TRUE,
  model = "gpt-4o-mini",
  ...
)

Arguments

service

The text generation service to use. Currently supports "openai", "huggingface", "anthropic", "google", "azure_openai", "ollama", and "perplexity".
prompt

The initial prompt or question to pass to the text generation service.
history

A list indicating the conversation history, where each element is a list with elements "role" (who is speaking; e.g., "system", "user") and "content" (what was said).
stream

Logical; indicates if streaming responses should be used. Currently, this option is not supported across all services.
model

The specific model to use for generating responses. Defaults to "gpt-3.5-turbo".
...

Additional arguments passed to the service-specific skeleton creation function.

Value

Depending on the selected service, returns a list that represents the configured request ready to be passed to the corresponding API.

Examples

## Not run: 
request_skeleton <- gptstudio_create_skeleton(
  service = "openai",
  prompt = "Name the top 5 packages in R.",
  history = list(list(role = "system", content = "You are an R assistant")),
  stream = TRUE,
  model = "gpt-3.5-turbo"
)

## End(Not run)

Perform API Request

Description

This function provides a wrapper for calling different APIs (e.g., OpenAI, HuggingFace, Google AI Studio). It dispatches the actual API calls to the relevant ellmer chat.

Usage

gptstudio_request_perform(skeleton, shiny_session = NULL)

Arguments

skeleton

A gptstudio_request_skeleton object
shiny_session

Shiny session to send messages to. Only relevant if skeleton$stream is TRUE.

Value

A list with a skeleton and and the last response

Run the ChatGPT app

Description

This starts the chatgpt app. It is exported to be able to run it from an R script.

Usage

gptstudio_run_chat_app(
  ide_colors = get_ide_theme_info(),
  code_theme_url = get_highlightjs_theme(),
  host = getOption("shiny.host", "127.0.0.1"),
  port = getOption("shiny.port")
)

Arguments

ide_colors

List containing the colors of the IDE theme.
code_theme_url

URL to the highlight.js theme
host

The IPv4 address that the application should listen on. Defaults to the shiny.host option, if set, or "127.0.0.1" if not. See Details.
port

The TCP port that the application should listen on. If the port is not specified, and the shiny.port option is set (with options(shiny.port = XX)), then that port will be used. Otherwise, use a random port between 3000:8000, excluding ports that are blocked by Google Chrome for being considered unsafe: 3659, 4045, 5060, 5061, 6000, 6566, 6665:6669 and 6697. Up to twenty random ports will be tried.

Value

Nothing.

Current Configuration for gptstudio

Description

This function prints out the current configuration settings for gptstudio and checks API connections if verbose is TRUE.

Usage

gptstudio_sitrep(verbose = TRUE)

Arguments

verbose

Logical value indicating whether to output additional information, such as API connection checks. Defaults to TRUE.

Value

Invisibly returns NULL, as the primary purpose of this function is to print to the console.

Examples

## Not run: 
gptstudio_sitrep(verbose = FALSE) # Print basic settings, no API checks
gptstudio_sitrep() # Print settings and check API connections

## End(Not run)

Spelling and Grammar Addin

Description

Call this function as a Rstudio addin to ask GPT to improve spelling and grammar of selected text.

Usage

gptstudio_spelling_grammar()

Value

This function has no return value.

Examples

# Select some text in Rstudio
# Then call the function as an RStudio addin
## Not run: 
gptstudio_spelling_grammar()

## End(Not run)

An audio clip input control that records short audio clips from the microphone

Description

An audio clip input control that records short audio clips from the microphone

Usage

input_audio_clip(
  id,
  record_label = "Record",
  stop_label = "Stop",
  reset_on_record = TRUE,
  mime_type = NULL,
  audio_bits_per_second = NULL,
  show_mic_settings = TRUE,
  ...
)

Arguments

id

The input slot that will be used to access the value.
record_label

Display label for the "record" control, or NULL for no label. Default is 'Record'.
stop_label

Display label for the "stop" control, or NULL for no label. Default is 'Record'.
reset_on_record

Whether to reset the audio clip input value when recording starts. If TRUE, the audio clip input value will become NULL at the moment the Record button is pressed; if FALSE, the value will not change until the user stops recording. Default is TRUE.
mime_type

The MIME type of the audio clip to record. By default, this is NULL, which means the browser will choose a suitable MIME type for audio recording. Common MIME types include 'audio/webm' and 'audio/mp4'.
audio_bits_per_second

The target audio bitrate in bits per second. By default, this is NULL, which means the browser will choose a suitable bitrate for audio recording. This is only a suggestion; the browser may choose a different bitrate.
show_mic_settings

Whether to show the microphone settings in the settings menu. Default is TRUE.
...

Additional parameters to pass to the underlying HTML tag.

Value

An audio clip input control that can be added to a UI definition.

App Server

Description

App Server

Usage

mod_app_server(id, ide_colors = get_ide_theme_info())

Arguments

id

id of the module
ide_colors

List containing the colors of the IDE theme.

App UI

Description

App UI

Usage

mod_app_ui(
  id,
  ide_colors = get_ide_theme_info(),
  code_theme_url = get_highlightjs_theme()
)

Arguments

id

id of the module
ide_colors

List containing the colors of the IDE theme.
code_theme_url

URL to the highlight.js theme

Chat server

Description

Chat server

Usage

mod_chat_server(
  id,
  ide_colors = get_ide_theme_info(),
  translator = create_translator(),
  settings,
  history
)

Arguments

id

id of the module
ide_colors

List containing the colors of the IDE theme.
translator

Translator from shiny.i18n::Translator
settings, history

Reactive values from the settings and history module

Chat UI

Description

Chat UI

Usage

mod_chat_ui(
  id,
  translator = create_translator(),
  code_theme_url = get_highlightjs_theme()
)

Arguments

id

id of the module
translator

A Translator from shiny.i18n::Translator
code_theme_url

URL to the highlight.js theme

Create HTML dependency for multimodal component

Description

Create HTML dependency for multimodal component

Usage

multimodal_dep()

Parse a Data URI

Description

This function parses a data URI and returns the MIME type and decoded data.

Usage

parse_data_uri(data_uri)

Arguments

data_uri

A string. The data URI to parse.

Value

A list with two elements: 'mime_type' and 'data'.

Prepare chat completion prompt

Description

This function prepares the chat completion prompt to be sent to the OpenAI API. It also generates a system message according to the given parameters and inserts it at the beginning of the conversation.

Usage

prepare_chat_history(
  history = list(list(role = "system", content = "You are an R chat assistant")),
  style = getOption("gptstudio.code_style"),
  skill = getOption("gptstudio.skill"),
  task = getOption("gptstudio.task", "coding"),
  custom_prompt = NULL
)

Arguments

history

A list of previous messages in the conversation. This can include roles such as 'system', 'user', or 'assistant'. System messages are discarded. Default is NULL.i
style

The style of code to use. Applicable styles can be retrieved from the "gptstudio.code_style" option. Default is the "gptstudio.code_style" option. Options are "base", "tidyverse", or "no preference".
skill

The skill level of the user for the chat conversation. This can be set through the "gptstudio.skill" option. Default is the "gptstudio.skill" option. Options are "beginner", "intermediate", "advanced", and "genius".
task

Specifies the task that the assistant will help with. Default is "coding". Others are "general", "advanced developer", and "custom".
custom_prompt

This is a custom prompt that may be used to guide the AI in its responses. Default is NULL. It will be the only content provided to the system prompt.

Value

A list where the first entry is an initial system message followed by any non-system entries from the chat history.

RGB str to hex

Description

RGB str to hex

Usage

rgb_str_to_hex(rgb_string)

Arguments

rgb_string

The RGB string as returned by rstudioapi::getThemeInfo()

Value

hex color

Set allowed models by provider

Description

[Experimental]

Sometimes it is useful to restrict the list of options in the model selection dropdown of the chat app. This function will check against get_available_models() to restrict the list to models that are actually available.

Usage

set_allowed_models(service, models = NULL)

Arguments

service

The API service
models

A character vector containing the list of allowed models that should be shown in the dropdown selector. If NULL (default), all models will be available.

Streaming message

Description

Places an invisible empty chat message that will hold a streaming message. It can be reset dynamically inside a shiny app

Usage

streamingMessage(
  ide_colors = get_ide_theme_info(),
  width = NULL,
  height = NULL,
  element_id = NULL
)

Arguments

ide_colors

List containing the colors of the IDE theme.
width, height

Must be a valid CSS unit (like '100%', '400px', 'auto') or a number, which will be coerced to a string and have 'px' appended.
element_id

The element's id

Shiny bindings for streamingMessage

Description

Output and render functions for using streamingMessage within Shiny applications and interactive Rmd documents.

Usage

streamingMessageOutput(outputId, width = "100%", height = NULL)

renderStreamingMessage(expr, env = parent.frame(), quoted = FALSE)

Arguments

outputId

output variable to read from
width, height

Must be a valid CSS unit (like '100%', '400px', 'auto') or a number, which will be coerced to a string and have 'px' appended.
expr

An expression that generates a streamingMessage
env

The environment in which to evaluate expr.
quoted

Is expr a quoted expression (with quote())? This is useful if you want to save an expression in a variable.

Style Chat History

Description

This function processes the chat history, filters out system messages, and formats the remaining messages with appropriate styling.

Usage

style_chat_history(history, ide_colors = get_ide_theme_info())

Arguments

history

A list of chat messages with elements containing 'role' and 'content'.
ide_colors

List containing the colors of the IDE theme.

Value

A list of formatted chat messages with styling applied, excluding system messages.

Examples

chat_history_example <- list(
  list(role = "user", content = "Hello, World!"),
  list(role = "system", content = "System message"),
  list(role = "assistant", content = "Hi, how can I help?")
)

## Not run: 
style_chat_history(chat_history_example)

## End(Not run)

Style chat message

Description

Style a message based on the role of its author.

Usage

style_chat_message(message, ide_colors = get_ide_theme_info())

Arguments

message

A chat message.
ide_colors

List containing the colors of the IDE theme.

Value

An HTML element.

Custom textAreaInput

Description

Modified version of textAreaInput() that removes the label container. It's used in mod_prompt_ui()

Usage

text_area_input_wrapper(
  inputId,
  label,
  value = "",
  width = NULL,
  height = NULL,
  cols = NULL,
  rows = NULL,
  placeholder = NULL,
  resize = NULL,
  textarea_class = NULL
)

Arguments

inputId

The input slot that will be used to access the value.
label

Display label for the control, or NULL for no label.
value

Initial value.
width

The width of the input, e.g. '400px', or '100%'; see validateCssUnit().
height

The height of the input, e.g. '400px', or '100%'; see validateCssUnit().
cols

Value of the visible character columns of the input, e.g. 80. This argument will only take effect if there is not a CSS width rule defined for this element; such a rule could come from the width argument of this function or from a containing page layout such as fluidPage().
rows

The value of the visible character rows of the input, e.g. 6. If the height argument is specified, height will take precedence in the browser's rendering.
placeholder

A character string giving the user a hint as to what can be entered into the control. Internet Explorer 8 and 9 do not support this option.
resize

Which directions the textarea box can be resized. Can be one of "both", "none", "vertical", and "horizontal". The default, NULL, will use the client browser's default setting for resizing textareas.
textarea_class

Class to be applied to the textarea element

Value

A modified textAreaInput

Transcribe Audio from Data URI Using OpenAI's Whisper Model

Description

This function takes an audio file in data URI format, converts it to WAV, and sends it to OpenAI's transcription API to get the transcribed text.

Usage

transcribe_audio(audio_input, api_key = Sys.getenv("OPENAI_API_KEY"))

Arguments

audio_input

A string. The audio data in data URI format.
api_key

A string. Your OpenAI API key. Defaults to the OPENAI_API_KEY environment variable.

Value

A string containing the transcribed text.

Examples

## Not run: 
audio_uri <- "data:audio/webm;base64,SGVsbG8gV29ybGQ=" # Example data URI
transcription <- transcribe_audio(audio_uri)
print(transcription)

## End(Not run)

Welcome message

Description

HTML widget for showing a welcome message in the chat app. This has been created to be able to bind the message to a shiny event to trigger a new render.

Usage

welcomeMessage(
  ide_colors = get_ide_theme_info(),
  translator = create_translator(),
  width = NULL,
  height = NULL,
  element_id = NULL
)

Arguments

ide_colors

List containing the colors of the IDE theme.
translator

A Translator from shiny.i18n::Translator
width, height

Must be a valid CSS unit (like '100%', '400px', 'auto') or a number, which will be coerced to a string and have 'px' appended.
element_id

The element's id

Shiny bindings for welcomeMessage

Description

Output and render functions for using welcomeMessage within Shiny applications and interactive Rmd documents.

Usage

welcomeMessageOutput(outputId, width = "100%", height = NULL)

renderWelcomeMessage(expr, env = parent.frame(), quoted = FALSE)

Arguments

outputId

output variable to read from
width, height

Must be a valid CSS unit (like '100%', '400px', 'auto') or a number, which will be coerced to a string and have 'px' appended.
expr

An expression that generates a welcomeMessage
env

The environment in which to evaluate expr.
quoted

Is expr a quoted expression (with quote())? This is useful if you want to save an expression in a variable.