llama-agents 通过retry 可以进行异常的处理，同时还提供了一个全局的catch_error 装饰器

参考玩法

• retry 定义

from workflows import Workflow, Context, step
from workflows.events import StartEvent, StopEvent
from workflows.retry_policy import (
    retry_policy,
    stop_after_attempt,
    wait_fixed,
)


class MyWorkflow(Workflow):
    @step(
        retry_policy=retry_policy(
            wait=wait_fixed(5),
            stop=stop_after_attempt(10),
        )
    )
    async def flaky_step(self, ctx: Context, ev: StartEvent) -> StopEvent:
        result = flaky_call()  # this might raise
        return StopEvent(result=result)

• 异常捕捉处理

通过catch_error

class Pipeline(Workflow):
    @step(retry_policy=...)
    async def fetch(self, ev: StartEvent) -> FetchedEvent: ...

    @step(retry_policy=...)
    async def parse(self, ev: FetchedEvent) -> StopEvent: ...

    @catch_error(for_steps=["fetch"])
    async def on_fetch_failure(
        self, ctx: Context, ev: StepFailedEvent
    ) -> StopEvent:
        return StopEvent(result={"fallback": True})

    @catch_error  # wildcard — covers `parse` and anything else
    async def on_any_failure(
        self, ctx: Context, ev: StepFailedEvent
    ) -> StopEvent:
        return StopEvent(result={"aborted": ev.step_name})

参考资料

https://developers.llamaindex.ai/python/llamaagents/workflows/retry_steps/