Set the config of a survey spec — set

Each argument is optional. The call changes only the values that it names.

Usage

set_config(
  spec,
  tries = NULL,
  response_delay = NULL,
  character_delay = NULL,
  delay_variance = NULL,
  version = NULL,
  valid = NULL,
  skip_answered = NULL,
  check_model = NULL
)

Arguments

spec: A survey spec from survey_spec().
tries: The maximum number of retries for an answer that is not valid. After the last retry, the survey keeps the answer and continues.
response_delay: The delay in seconds before the bot starts a message.
character_delay: The delay in seconds between characters of the simulated typing. Use 0 for no delay.
delay_variance: The random variation in seconds of character_delay.
version: The version of the question set.
valid: The default condition for a valid answer. It applies to each add_question() call after this one that has no valid of its own.
skip_answered: Whether a reply can answer later questions. If TRUE, the LLM also extracts clear answers to later fixed questions, and the survey records them and does not ask those questions. Adaptive questions are always asked. In the database, such an answer has no question_text, and answer_raw is the reply that gave it.
check_model: Whether the server sends the model a short test prompt when the chat opens. If TRUE, a spinner covers the chat until the model answers, and a failed request locks the survey with the locked message of set_messages(). The request has one try. A success serves every session that opens in the next 30 seconds, so most sessions add no request. A failure serves the sessions of the next 5 seconds. If FALSE, the survey starts at once with no extra request.

Value

spec with the new config.

Examples

survey_spec() |>
  set_config(tries = 3, character_delay = 0)
#> <surveychat_spec> version "1.0", 0 questions