Skip to main content

Tool

Trait Tool 

Source
pub trait Tool: Send + Sync {
    // Required methods
    fn name(&self) -> &str;
    fn description(&self) -> &str;
    fn parameters_schema(&self) -> Value;
    fn execute<'life0, 'life1, 'async_trait>(
        &'life0 self,
        args: Value,
        ctx: &'life1 mut AgentContext,
    ) -> Pin<Box<dyn Future<Output = Result<ToolOutput, ToolError>> + Send + 'async_trait>>
       where Self: 'async_trait,
             'life0: 'async_trait,
             'life1: 'async_trait;

    // Provided methods
    fn is_system(&self) -> bool { ... }
    fn is_read_only(&self) -> bool { ... }
    fn execute_readonly<'life0, 'life1, 'async_trait>(
        &'life0 self,
        args: Value,
        ctx: &'life1 AgentContext,
    ) -> Pin<Box<dyn Future<Output = Result<ToolOutput, ToolError>> + Send + 'async_trait>>
       where Self: 'async_trait,
             'life0: 'async_trait,
             'life1: 'async_trait { ... }
    fn to_def(&self) -> ToolDef { ... }
}
Expand description

A tool that an agent can invoke.

Implement this trait for each capability you want to expose to the LLM agent. Tools are registered in a ToolRegistry and dispatched by the agent loop.

Read-only tools (is_read_only() -> true) can execute in parallel. Write tools execute sequentially with exclusive &mut AgentContext.

Required Methods§

Source

fn name(&self) -> &str

Unique tool name (used as discriminator in LLM function calling).

Source

fn description(&self) -> &str

Human-readable description shown to the LLM.

Source

fn parameters_schema(&self) -> Value

JSON Schema for the tool’s parameters (generated via json_schema_for::<Args>()).

Source

fn execute<'life0, 'life1, 'async_trait>( &'life0 self, args: Value, ctx: &'life1 mut AgentContext, ) -> Pin<Box<dyn Future<Output = Result<ToolOutput, ToolError>> + Send + 'async_trait>>
where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait,

Provided Methods§

Source

fn is_system(&self) -> bool

System tools are always visible (not subject to progressive discovery).

Source

fn is_read_only(&self) -> bool

Read-only tools can execute in parallel via execute_readonly.

Source

fn execute_readonly<'life0, 'life1, 'async_trait>( &'life0 self, args: Value, ctx: &'life1 AgentContext, ) -> Pin<Box<dyn Future<Output = Result<ToolOutput, ToolError>> + Send + 'async_trait>>
where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait,

Execute without mutable context (for parallel read-only dispatch). Default: delegates to execute with a cloned context. Override for true read-only tools to avoid the clone.

Source

fn to_def(&self) -> ToolDef

Implementors§