Generates a neuroscientific plausibility table by directly querying an LLM. — generate_plausibility_table

This function constructs a detailed prompt with specific IDPs, Outcomes, Domains, scoring principles, and desired output format, then queries an LLM (Groq or OpenRouter) to generate a complete plausibility table.

generate_plausibility_table_direct(
  backend = c("groq", "openrouter"),
  model = NULL,
  api_key_env = NULL,
  max_retries = 5,
  retry_delay_base = 1,
  temperature = 0.01,
  verbose = TRUE,
  return_prompt = FALSE,
  custom_system_prompt = NULL,
  custom_idps = NULL,
  custom_outcomes = NULL,
  custom_domains = NULL
)

Arguments

backend: Character string, specifies which API backend to use: "groq" or "openrouter".
model: Character string, model name for the chosen backend. Defaults to "llama3-70b-8192" for Groq and "openai/gpt-4o" for OpenRouter.
api_key_env: Character string, name of the environment variable storing the API key.
max_retries: Integer, maximum number of retries in case of API errors or malformed responses.
retry_delay_base: Numeric, base delay in seconds for exponential backoff between retries.
temperature: Numeric, sampling temperature for the LLM. A very low value (e.g., 0.01) is recommended for deterministic, instruction-following behavior.
verbose: Logical, if TRUE, prints progress and debugging information.
return_prompt: Logical, if TRUE, returns the constructed system and user prompts as a list without calling the LLM. Useful for debugging prompts.
custom_system_prompt: Character string, an optional custom system prompt to use. If NULL, the default detailed system prompt (acting as a neuroscientist) is used.
custom_idps: Character vector, an optional custom list of IDPs to use. If NULL, the globally defined 'idps' list is used.
custom_outcomes: Character vector, an optional custom list of Outcomes to use. If NULL, the globally defined 'outcomes' list is used.
custom_domains: Character vector, an optional custom list of Domains to use. If NULL, the globally defined 'domains' list is used.

Value

A list containing:

dataframe: A data.frame with IDPs as rows, Outcomes as columns, and plausibility scores as values.
raw_json: The raw JSON string received from the LLM.

If return_prompt is TRUE, returns a list with system_prompt and user_prompt.