Help for package askgpt

Type:

Package

Title:

Asking GPT About R Stuff

Version:

0.1.3

Description:

A chat package connecting to API endpoints by 'OpenAI' (https://platform.openai.com/) to answer questions (about R).

Depends:

R (≥ 4.1.0)

Imports:

cli, callr, dplyr, glue, methods, rlang, httr2, rappdirs, jsonlite

Suggests:

covr, knitr, miniUI, rmarkdown, rstudioapi, shiny, shinycssloaders, spelling, testthat (≥ 3.0.0), withr

URL:

https://github.com/JBGruber/askgpt

BugReports:

https://github.com/JBGruber/askgpt/issues

License:

GPL (≥ 3)

Encoding:

UTF-8

RoxygenNote:

7.2.3

VignetteBuilder:

knitr

Config/testthat/edition:

Language:

en-GB

LazyData:

true

NeedsCompilation:

Packaged:

2023-09-08 08:14:55 UTC; johannes

Author:

Johannes Gruber [aut, cre]

Maintainer:

Johannes Gruber <johannesb.gruber@gmail.com>

Repository:

CRAN

Date/Publication:

2023-09-08 08:30:05 UTC

Annotate R code with inline comments

Description

Annotate R code with inline comments

Usage

annotate_code(code, ...)

Arguments

code

A character vector of R code. If missing the code currently selected in RStudio is documented (If RStudio is used).

...

passed on to askgpt.

Value

A character vector.

Ask openai's GPT models a question

Description

Ask openai's GPT models a question

Usage

askgpt(prompt, chat = TRUE, progress = TRUE, return_answer = FALSE, ...)

Arguments

prompt

What you want to ask

chat

whether to use the chat API (i.e., the same model as ChatGPT) or the completions API.

progress

Show a progress spinner while the request to the API has not been fulfilled.

return_answer

Should the answer be returned as an object instead of printing it to the screen?

...

additional options forwarded to chat_api or completions_api respectively.

Value

either an httr2 response from one of the APIs or a character vector (if return_answer).

Examples

## Not run: 
askgpt("What is an R function?")
askgpt("What is wrong with my last command?")
askgpt("Can you help me with the function aes() from ggplot2?")

## End(Not run)

Request answer from openai's chat API

Description

Request answer from openai's chat API

Usage

chat_api(
  prompt,
  model = NULL,
  config = NULL,
  max_tokens = NULL,
  api_key = NULL,
  ...
)

Arguments

prompt

character string of the prompt to be completed.

model

character string of the model to be used (defaults to "text-davinci-003").

config

a configuration prompt to tell the model how it should behave.

max_tokens

The maximum number of tokens to generate in the completion. 2048L is the maximum the models accept.

api_key

set the API key. If NULL, looks for the env OPENAI_API_KEY.

...

additional parameters to be passed to the API (see [the API documentation](https://platform.openai.com/docs/api-reference/completions)

Value

A tibble with available models

a httr2 response object

Examples

## Not run: 
chat_api("Hi, how are you?", config = "answer as a friendly chat bot")

## End(Not run)

Request answer from openai's completions API

Description

Mostly used under the hood for askgpt.

Usage

completions_api(
  prompt,
  model = NULL,
  temperature = NULL,
  max_tokens = NULL,
  api_key = NULL,
  ...
)

Arguments

prompt

character string of the prompt to be completed.

model

character string of the model to be used (defaults to "text-davinci-003").

temperature

numeric value between 0 and 1 to control the randomness of the output (defaults to 0.2; lower values like 0.2 will make answers more focused and deterministic).

max_tokens

The maximum number of tokens to generate in the completion. 2048L is the maximum the models accept.

api_key

set the API key. If NULL, looks for the env OPENAI_API_KEY.

...

additional parameters to be passed to the API (see [the API documentation](https://platform.openai.com/docs/api-reference/completions)

Details

Only a few parameters are implemented by name. Most can be sent through the .... For example, you could use the n parameter just like this completions_api("The quick brown fox", n = 2).

A couple of defaults are used by the package:

the model used by default is "text-davinci-003"
the default temperature is 0.2
the default for max_tokens is 2048L

You can configure how askgpt makes requests by setting options that start with askgpt_*. For example, to use a different model use options(askgpt_model = "text-curie-001"). It does not matter if the API parameter ist listed in the function or not. All are used.

Value

a httr2 response object

Examples

## Not run: 
completions_api("The quick brown fox")

## End(Not run)

Document R Code

Description

Document R Code

Usage

document_code(code, ...)

Arguments

code

A character vector of R code. If missing the code currently selected in RStudio is documented (If RStudio is used).

...

passed on to askgpt.

Value

A character vector.

Examples

## Not run: 
document_code()

## End(Not run)

Estimate token count

Description

Estimate token count

Usage

estimate_token(x, mult = 1.6)

Arguments

x

character vector

mult

the multiplier used

Details

This function estimates how many tokens the API will make of the input words. For the models 1 word is more than one token. The default multiplier value resulted from testing the API. See <https://help.openai.com/en/articles/4936856-what-are-tokens-and-how-to-count-them> for more information.

Value

a integer vector of token counts

Examples

estimate_token("this is a test")

Explain R code

Description

Explain R code

Usage

explain_code(code, ...)

Arguments

code

A character vector of R code. If missing the code currently selected in RStudio is documented (If RStudio is used).

...

passed on to askgpt.

Value

A character vector.

Improve code/documentation/writing using a prompt

Description

'tutorialise_addin()' opens an [RStudio gadget](https://shiny.rstudio.com/articles/gadgets.html) and [addin](http://rstudio.github.io/rstudioaddins/) that can be used to improve existing code, documentation, or writing.

Usage

improve_addin()

Value

No return value, opens a new file in RStudio

List Models

Description

List the models available in the API. You can refer to the [Models documentation](https://platform.openai.com/docs/models) to understand what models are available and the differences between them.

Usage

list_models(api_key = NULL)

Arguments

api_key

set the API key. If NULL, looks for the env OPENAI_API_KEY.

Value

A tibble with available models

Examples

## Not run: 
completions_api("The quick brown fox")

## End(Not run)

Initiate error logging

Description

Initiate error logging

Usage

log_init(...)

Arguments

...

forwarded to global_entrace

Details

Just an alias for rlang::global_entrace() with a more fitting name (for the purpose here).

Value

No return value, called to enable rlang error logging

Description

Usage

login(api_key, force_refresh = FALSE, cache_dir = NULL, no_cache = FALSE)

Arguments

api_key

API key to use for authentication. If not provided, the function look for a cached key or guide the user to obtain one.

force_refresh

cache_dir

dir location to save keys on disk. The default is to use rappdirs::user_cache_dir("askgpt").

no_cache

Don't cache the API key, only load it into the environment.

Value

a character vector with an API key

Start a new conversation

Description

Deletes the local prompt and response history to start a new conversation.

Usage

new_conversation()

Value

Does not return a value

Parse response from API functions

Description

Parse response from API functions

Usage

parse_response(response)

Arguments

response

a response object from chat_api or completions_api

Value

a character vector

Return the prompt/response history

Description

Return the prompt/response history

Usage

prompt_history(n = Inf)

Arguments

n

number of prompts/responses to return.

Value

a character vector

Return the prompt/response history

Description

Return the prompt/response history

Usage

response_history(n = Inf)

Arguments

n

number of prompts/responses to return.

Value

a character vector

Test R code

Description

Test R code

Usage

test_function(code, ...)

Arguments

code

A character vector of R code. If missing the code currently selected in RStudio is documented (If RStudio is used).

...

passed on to askgpt.

Value

A character vector.

Max tokens limits of the different models

Description

OpenAI's token limits for different models.

Usage

token_limits

Format

An object of class data.frame with 6 rows and 2 columns.

Source

<https://platform.openai.com/docs/models/overview>

Turn R code into a tutorial

Description

'tutorialise_addin()' opens an [RStudio gadget](https://shiny.rstudio.com/articles/gadgets.html) and [addin](http://rstudio.github.io/rstudioaddins/) that turns selected code into an R Markdown/Quarto Tutorial.

Usage

tutorialise_addin()

Value

No return value, opens a new file in RStudio

Annotate R code with inline comments

Description

Usage

Arguments

Value

Ask openai's GPT models a question

Description

Usage

Arguments

Value

Examples

Request answer from openai's chat API

Description

Usage

Arguments

Value

Examples

Request answer from openai's completions API

Description

Usage

Arguments

Details

Value

Examples

Document R Code

Description

Usage

Arguments

Value

Examples

Estimate token count

Description

Usage

Arguments

Details

Value

Examples

Explain R code

Description

Usage

Arguments

Value

Improve code/documentation/writing using a prompt

Description

Usage

Value

List Models

Description

Usage

Arguments

Value

Examples

Initiate error logging

Description

Usage

Arguments

Details

Value

Log in to OpenAI

Description

Usage

Arguments

Value

Start a new conversation

Description

Usage

Value

Parse response from API functions

Description

Usage

Arguments

Value

Return the prompt/response history

Description

Usage

Arguments

Value

Return the prompt/response history

Description

Usage