A hybrid agent module that combines tool calling with R code execution. The model can choose between calling registered tools or generating R code to solve problems. This enables flexible agentic workflows that leverage both external tools and computational capabilities.
Details
CodeAct extends the ReAct pattern by adding an execute_r_code tool that
allows the agent to write and run R code. The execution flow is:
Agent receives the task and available tools (including code execution)
Agent iteratively calls tools or executes code until it has enough info
Agent produces final structured answer
Security: Code execution requires explicit opt-in via a runner parameter. The runner provides subprocess isolation but is NOT a security sandbox. For untrusted inputs, use OS-level sandboxing.
Examples
if (FALSE) { # \dontrun{
# Create a runner for code execution
runner <- r_code_runner(timeout = 30)
# Create a CodeAct agent with custom tools
search_tool <- ellmer::tool(
function(query) "Search results...",
description = "Search for information"
)
agent <- code_act(
signature = "question -> answer",
tools = list(search = search_tool),
runner = runner
)
# The agent can now search AND compute
result <- run(agent,
question = "What is 10% of France's population?",
.llm = llm
)
} # }