Struct AgentTool 

pub struct AgentTool { /* private fields */ }

A Tool that runs a child Agent on a task and returns its answer.

use autumn_plugin_agent::delegate::AgentTool;
use autumn_plugin_agent::tools::ToolEffect;

let researcher = Agent::new(Arc::clone(&client))
    .system_prompt("You research one question and answer in five lines.")
    .max_steps(6);
let tool = AgentTool::new(
    "research",
    "Research one question in depth and return a short answer.",
    researcher,
)
.effect(ToolEffect::ReadOnly);
let parent = Agent::new(client).tool(Arc::new(tool));

Implementations

impl AgentTool

pub fn new( name: impl Into<String>, description: impl Into<String>, agent: Agent, ) -> Self

Wrap agent as a tool. The model calls it with {"task": "..."}.

pub const fn effect(self, effect: ToolEffect) -> Self

Declare the strongest effect among the child’s tools. The default is ToolEffect::Write.

Trait Implementations

impl Clone for AgentTool

fn clone(&self) -> Self

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for AgentTool

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Tool for AgentTool

fn name(&self) -> &str

Tool name as the model calls it. Keep it snake_case.

fn description(&self) -> &str

What the tool does, shown to the model.

fn input_schema(&self) -> Value

JSON Schema for the tool’s input object.

fn effect(&self) -> ToolEffect

What the tool can change. Defaults to ToolEffect::Write: a tool is trusted with less only when it says so.

fn execute<'a>( &'a self, input: Value, _ctx: &'a ToolContext, ) -> Pin<Box<dyn Future<Output = Result<Value, AgentError>> + Send + 'a>>

Run the tool against decoded JSON arguments.

fn definition(&self) -> ToolDefinition

Bundle the metadata the model sees into a ToolDefinition.