Hi.

from llama_index.core.agent import AgentRunner, ReActAgentWorker, ReActAgent
prompt="An interesting question?"
task = agent.create_task(prompt)
step_output = agent.stream_step(task.task_id)
for token in step_output.output.response_gen:
    print(token, end="")

llm = OpenAI(model="gpt-3.5-turbo")
agent = OpenAIAgent.from_tools(
    [multiply_tool, add_tool], llm=llm, verbose=True
)

task = agent.create_task("What is 2123 * 215123")

while True:
    task_result = agent.stream_step(task.task_id)
    for token in task_result.output.response_gen:
        print(token, end="")
        time.sleep(0.2) # sleeping, just to emphasize the streaming
    
    if task_result.is_last:
        break

Cell In[18], line 4
      2 while True:
      3     task_result = agent.stream_step(task.task_id)
----> 4     for token in task_result.output.response_gen:
      5         print(token, end="")
      6         time.sleep(0.2) # sleeping, just to emphasize the streaming

File ~/Desktop/project/venv/lib/python3.10/site-packages/llama_index/core/chat_engine/types.py:73, in AgentChatResponse.response_gen(self)
     71 """Used for fake streaming, i.e. with tool outputs."""
     72 if not self.is_dummy_stream:
---> 73     raise ValueError(
     74         "response_gen is only available for streaming responses. "
     75         "Set is_dummy_stream=True if you still want a generator."
     76     )
     78 for token in self.response.split(" "):
     79     yield token + " "

ValueError: response_gen is only available for streaming responses. Set is_dummy_stream=True if you still want a generator.