Describe text for the LLM to generate — prompt

Use prompt_llm() for text in add_question() to make an adaptive question, or for intro to put generated content before a question.

Usage

prompt_llm(prompt, format = NULL)

Arguments

prompt: The prompt for the LLM. It can use {id} placeholders for the answers to earlier questions.
format: For intro, or a generated completion in set_messages(): a template that places the generated text. It must contain {content} and can use {id} placeholders. For an intro, the survey puts the question after it, with a blank line between.

Value

A list of class surveychat_prompt.

Details

An intro gets only the answers that its placeholders name. An adaptive question and a generated completion in set_messages() also get every answer so far, as does the LLM validation of each reply. Each answer is one line, - <question> (id): value, with the question text that the user saw.

Examples

prompt_llm("Share a short fun fact about {flavor} ice cream.")
#> $prompt
#> [1] "Share a short fun fact about {flavor} ice cream."
#> 
#> $format
#> NULL
#> 
#> attr(,"class")
#> [1] "surveychat_prompt"
prompt_llm("Share a fun fact about {flavor}.", format = "Oh! {content}")
#> $prompt
#> [1] "Share a fun fact about {flavor}."
#> 
#> $format
#> [1] "Oh! {content}"
#> 
#> attr(,"class")
#> [1] "surveychat_prompt"