sendPrompt: Send a prompt to a specified language model agent and return...

In mergen: AI-Driven Code Generation, Explanation and Execution for Data Analysis

Send a prompt to a specified language model agent and return the response.

Description

Send a prompt to a specified language model agent and return the response.

Usage

sendPrompt(
  agent,
  prompt,
  context = promptContext(type = "simple"),
  return.type = c("text", "object"),
  previous.msgs = NULL,
  ...
)

Arguments

agent	An object containing the agent's information (e.g., type and model etc.).
prompt	The prompt text to send to the language model.
context	Optional context to provide alongside the prompt (default is promptContext(type = "simple")).
return.type	The type of output to return, either the text response ("text") or the entire response object ("object").
previous.msgs	a list of lists for previous prompts and responses. Useful to add context of the previous messages for the API. this argument works with openai and generic models, but not with replicate API. Default: NULL
...	Additional arguments to be passed to the LLM API. Such as maximum tokens, ("max_tokens"), to be returned. Users can also also provide other arguments in openai API-like arguments documented on [the official documentation](https://platform.openai.com/docs/api-reference/chat/create). Other APIs are also following similar argument naming patterns.

Value

The text response or the entire response object, based on the specified return type.

See Also

promptContext for predefined contexts to use.

Examples

## Not run: 
agent <- setupAgent(name="openai",type="chat",model="gpt-4",
                    ai_api_key=Sys.getenv("OPENAI_API_KEY"))
prompt <- "tell me a joke"
response <- sendPrompt(agent, prompt,context="")

# increase tokens, it is important for getting longer responses
response <- sendPrompt(agent,prompt,context="",return.type="text", max_tokens = 500)

# get previous messages into the context
prompt="what about 2010?"
response <- sendPrompt(agent,prompt,context="",
                       return.type="text",
                       previous.msgs=list(
                                          list(
                                               "role" = "user",
                                                "content" = "Who won the world
                                                 series in 2020?"
                                                 ),
                                          list(
                                                "role" = "assistant",
                                              "content" = "The Los Angeles Dodgers"
                                              )
                                          )
                       )



## End(Not run)

mergen documentation built on June 27, 2024, 5:10 p.m.