Code Execution

Let agents write and run code

Enable agents to write and execute code to solve problems.

Quick Start

1. Create Code Execution Agent

   use praisonai::{Agent, tool};
   
   // Define a code execution tool
   #[tool(description = "Execute Python code and return the result")]
   async fn run_python(code: String) -> String {
       // Execute code in sandbox
       execute_python_sandboxed(&code).await
   }
   
   // Create agent with code execution tool
   let agent = Agent::new()
       .name("Coder")
       .instructions("Write and execute code to solve problems. Use the run_python tool.")
       .tool(run_python)
       .build()?;
   
   let response = agent.chat("Calculate the first 10 Fibonacci numbers").await?;
   // Agent writes Python code and executes it via the tool

2. Safe Code Sandbox

   use praisonai::tool;
   
   #[tool(description = "Execute JavaScript code safely")]
   async fn run_js(code: String) -> String {
       // Use sandboxed environment with timeout
       sandbox::execute_with_timeout(
           &code,
           std::time::Duration::from_secs(10)
       ).await
   }

Supported Languages

Security

[!CAUTION] Code execution allows running arbitrary code. Use in trusted environments only.

Related