Streaming text chunks support for sales agent executor chain instead of intermediate steps

[aliabbas101]

I am building a low latency conversational agent and need to utilize the tool calling feature each after human_step() call, one thing I have noticed in the code below:

SalesGPT/salesgpt/agents.py

Line 420 in 466a441

if self.use_tools:

SalesGPT/salesgpt/agents.py

Line 421 in 466a441

ai_message = self.sales_agent_executor.invoke(inputs)

When the use_tools argument is set to true, we are using sales_agent_executor instead of the the utterance chain, which only streams the intermediate steps of the executor, is it possible to stream chunks of sentences as they are being generated to reduce the response latency ?

Thanks.