Skip to main content
Shows how to use With*ExecutionConfig options to give each agent loop operation its own timeout and max attempts instead of relying on the SDK defaults. Source: examples/agent_with_execution_config/

​
What it demonstrates

  • WithLLMExecutionConfig — tighter LLM timeout, up to 3 attempts
  • WithToolExecutionConfig — short tool deadline, single attempt
  • WithSubAgentExecutionConfig — explicit sub-agent ceiling of agent run timeout
  • Zero fields fall back to SDK defaults — only set what you want to change

​
Run

From examples/: 
go run ./agent_with_execution_config
go run ./agent_with_execution_config "What is 99 times 77?"
AGENT_RUNTIME=temporal go run ./agent_with_execution_config

​
Key code

a, err := agent.NewAgent(
    agent.WithName("exec-config-agent"),
    agent.WithLLMClient(llmClient),
    agent.WithTools(calculator.New(), currenttime.New()),
    agent.WithToolApprovalPolicy(agent.AutoToolApprovalPolicy()),

    // LLM: tighter timeout for a chat agent; up to 3 attempts for transient errors.
    agent.WithLLMExecutionConfig(agent.ExecutionConfig{
        Timeout:     10 * time.Minute,
        MaxAttempts: 3,
    }),

    // ToolExecute: fast tools — short deadline, one attempt.
    agent.WithToolExecutionConfig(agent.ExecutionConfig{
        Timeout:     2 * time.Minute,
        MaxAttempts: 1,
    }),

    // SubAgent: explicit ceiling — delegates must finish within 5 minutes.
    agent.WithSubAgentExecutionConfig(agent.ExecutionConfig{
        Timeout: 5 * time.Minute,
    }),
)

​
Expected output

user: What is 123 multiplied by 456? Also, what time is it right now?

assistant: 123 × 456 = 56,088. The current time is 14:32 UTC.

​
Learn more

Execution Config

Full feature docs — defaults table, retry backoff, runtime behavior

Timeouts & Modes

Agent run timeout, approval timeout, and agent modes