Trait Tool

Source

pub trait Tool: Send + Sync {
    // Required methods
    fn definition(&self) -> &ToolDefinition;
    fn execute<'life0, 'async_trait>(
        &'life0 self,
        args: Value,
    ) -> Pin<Box<dyn Future<Output = Result<Value, ToolError>> + Send + 'async_trait>>
       where Self: 'async_trait,
             'life0: 'async_trait;

    // Provided methods
    fn validate(&self, _args: &Value) -> Result<(), ToolError> { ... }
    fn capabilities(&self) -> Vec<Capability> { ... }
    fn timeout(&self) -> Duration { ... }
    fn is_available(&self) -> bool { ... }
}

Expand description

The core Tool trait — interface all tools must implement.

Tools are the bridge between LLM decision-making and real-world actions. Every tool must:

Provide its definition (name, description, schema)
Implement async execution

§Security

validate() is called before execute() — reject bad input early
capabilities() declares what the tool needs for sandboxing
timeout() prevents DoS from hanging operations

§Example

use vex_llm::{Tool, ToolDefinition, ToolError, Capability};
use async_trait::async_trait;

pub struct MyTool {
    definition: ToolDefinition,
}

#[async_trait]
impl Tool for MyTool {
    fn definition(&self) -> &ToolDefinition {
        &self.definition
    }

    async fn execute(&self, args: serde_json::Value) -> Result<serde_json::Value, ToolError> {
        // Your implementation here
        Ok(serde_json::json!({"status": "done"}))
    }
}