[Question]: why the second reponse is empty?

[codeh2o]

Question Validation

• I have searched both the documentation and discord for an answer.

Question

workflow = FunctionAgent(
    llm=llm,
    system_prompt="You are an  assistant"
)

ctx = Context(workflow)


async def main():
    response = await workflow.run(user_msg="Hi, my name is Laurie!", ctx=ctx)
    print("first response:", response)
    response2 = await workflow.run(user_msg="what's my name?", ctx=ctx)
    print("second response:", response2)


if __name__ == '__main__':
    asyncio.run(main())

the response is

first response: Hello Laurie! It's nice to meet you. How can I assist you today? Is there anything specific you'd like to chat about or any questions you have?
second response: 

Process finished with exit code 0

https://developers.llamaindex.ai/python/framework/understanding/agent/state/

as the documenation shown,the second response what I expect is

Your name is Laurie!

but it is empty

[dosubot]

The second response is empty because the FunctionAgent does not automatically remember or store information like your name between runs, even when using a Context object. The agent only retains state if its logic or tools explicitly read from and write to the Context—for example, by extracting the user's name and saving it in the context during the first run, then recalling it in the second run. This behavior is not built into the default FunctionAgent; you need to add custom logic or use tools that interact with Context to achieve the expected memory "Your name is Laurie!".

If you want the agent to remember and recall the user's name, you can define a tool or function that updates and retrieves the name from Context. Here’s a minimal example:

from llama_index.core.workflow import Context

async def set_name(ctx: Context, name: str) -> str:
    state = await ctx.get("state") or {}
    state["name"] = name
    await ctx.set("state", state)
    return f"Name set to {name}"

async def get_name(ctx: Context) -> str:
    state = await ctx.get("state") or {}
    name = state.get("name", "unknown")
    return f"Your name is {name}!"

# Then create your agent/workflow using these tools and pass the same Context object to each run.

If you’re using streaming output, make sure to reset ctx.is_running = False before the next run, as leaving it True can block further responses "source".

Let me know if you want a full example or help adapting this to your workflow!

To reply, just mention @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}    

[logan-markewich]

@codeh2o I can't reproduce
https://colab.research.google.com/drive/16D7If2GnsyZAA9QElaTwBFbEVnwE95GZ?usp=sharing

What LLM are you using? If you can reproduce in colab that'd be extremely helpful