Run a Recursive Language Model in one call

Convenience wrapper that creates a runner, module, and executes an RLM in a single call. Equivalent to:

runner <- r_code_runner(timeout = timeout)
mod <- rlm_module(signature, runner = runner, ...)
run(mod, ..., .llm = .llm)

For repeated use or optimization, prefer creating a module with rlm_module() and calling run() separately.

Usage

rlm(
  signature,
  ...,
  .llm = NULL,
  .timeout = 30,
  .max_iterations = 20L,
  .max_llm_calls = 50L,
  .sub_lm = NULL,
  .tools = list(),
  .verbose = FALSE
)

Arguments

signature

A Signature object or string notation defining inputs/outputs (e.g., "question -> answer")

...

Named arguments matching the signature's inputs. These are passed to run().

.llm

An ellmer Chat object. If NULL, uses the default Chat from get_default_chat().

.timeout

Numeric. Maximum execution time in seconds per code evaluation. Default 30.

.max_iterations

Integer. Maximum REPL iterations before fallback. Default 20.

.max_llm_calls

Integer. Maximum recursive LLM calls allowed. Default 50.

.sub_lm

Optional ellmer Chat for recursive llm_query() calls. NULL disables recursive queries.

.tools

Named list of user-defined R functions available in the REPL.

.verbose

Logical. Print execution progress. Default FALSE.

Value

The module output according to the signature.

See also

• rlm_module() for creating reusable RLM modules

• r_code_runner() for configuring the code execution backend

• run() for executing modules

• dsp() for simple one-shot LLM calls (no code execution)

Examples

if (FALSE) { # \dontrun{
# One-liner RLM call
result <- rlm(
  "document, question -> answer",
  document = readLines("big_file.txt") |> paste(collapse = "\n"),
  question = "What are the main themes?",
  .llm = ellmer::chat_openai()
)

# With recursive sub-queries
result <- rlm(
  "codebase, question -> answer",
  codebase = source_code,
  question = "How does auth work?",
  .llm = ellmer::chat_openai(),
  .sub_lm = ellmer::chat_openai(model = "gpt-4o-mini")
)
} # }