Ask a query using a ragcat RAG store — ask_rag

Creates a retrieval-enabled ellmer chat, asks the query, parses structured output, and optionally writes answer artifacts.

Usage

ask_rag_store(
  query,
  store_location = RAG_STORE_LOCATION,
  prompt_file = NULL,
  output_instructions_file = NULL,
  topic = "custom_rag",
  top_k = 12L,
  save_outputs = TRUE,
  output_dir = RESULTS_DIR,
  output_prefix = paste0(make_store_identifier(topic), "_answer"),
  system_prompt = NULL,
  system_prompt_file = NULL,
  use_saved_system_prompt = TRUE,
  output_instructions = NULL,
  echo = FALSE,
  ...
)

Arguments

query: User query or task.
store_location: Path to the ragnar store.
prompt_file: Optional path to a Markdown/text system prompt file.
output_instructions_file: Optional path to Markdown/text output instructions.
topic: Topic label used in the retrieval tool description and output prefix.
top_k: Number of chunks retrieved by the RAG retrieval tool.
save_outputs: Logical; save outputs to disk.
output_dir: Output directory.
output_prefix: Output file prefix.
system_prompt: Optional character scalar system prompt. Ignored when prompt_file is supplied.
system_prompt_file: Deprecated alias for prompt_file.
use_saved_system_prompt: Logical; whether to use system_prompt.txt beside the store when neither prompt_file nor system_prompt is supplied.
output_instructions: Optional character scalar output instructions. Ignored when output_instructions_file is supplied.
echo: Logical; print query, prompt-file paths, saved-output paths, and the answer to the console.
...: Additional arguments passed to make_rag_chat(), such as model, endpoint, or params.

Value

A list with query, answer, summary, structured, saved, prompt_file, and output_instructions_file.

Details

ask_rag_store() is the preferred user-facing API for querying an existing ragcat/ragnar DuckDB store. Use prompt_file for the system prompt and output_instructions_file for the expected response format.

Examples

if (FALSE) { # \dontrun{
answer <- ask_rag_store(
  query = "What evidence is available for the project question?",
  store_location = file.path("data", "example_rag_store", "rag_store.duckdb"),
  prompt_file = file.path("prompts", "system_prompt.md"),
  output_instructions_file = file.path("prompts", "output_instructions.md"),
  top_k = 12
)
} # }