Langgraph and Streamlit used to create human-in-the-loop news writing app

[tevslin]

https://meeting-reporter.streamlit.app/ mates Streamlit and Langgraph to create an app using both multiple agents and human-in-the-loop to generate news stories more reliably than AI can alone and more cheaply than humans can without AI. It's an example of how AI can help fill a gap in local news reporting.

Based on GPT4-turbo so you do need your own paid OpenAI API key to get past the first screen (cost a few pennies per run).

Code is open source at https://github.com/tevslin/meeting-reporter

Screenshots and transcript of a a session are here,

Most examples of Langgraph use are in Jupyter notebooks so not really suitable for deployment to a broad audience. Streamlit solves the UI problem but mating the Streamlit and Langgraph state machines is an interesting problem.

[Freezaa9]

nice job !

[skprasadu]

@tevslin ,

I am stuck on a problem, I am exploring an example in LangGraph @ https://langchain-ai.github.io/langgraph/tutorials/customer-support/customer-support/#part-3-conditional-interrupt and while following "Part 3: Conditional Interrupt", I am running into issue with below code using StreamLit as below, if I have to take in a user_input as y/n and pass it back to Langgraph from StreamLit, how would I do that. I tried several combination, without luck. Any help will be appreciated.

for question in tutorial_questions:
events = part_3_graph.stream(
{"messages": ("user", question)}, config, stream_mode="values"
)
for event in events:
_print_event(event, _printed)
snapshot = part_3_graph.get_state(config)
while snapshot.next:
# We have an interrupt! The agent is trying to use a tool, and the user can approve or deny it
# Note: This code is all outside of your graph. Typically, you would stream the output to a UI.
# Then, you would have the frontend trigger a new run via an API call when the user has provided input.
user_input = input(
"Do you approve of the above actions? Type 'y' to continue;"
" otherwise, explain your requested changed.\n\n"
)
if user_input.strip() == "y":
# Just continue
result = part_3_graph.invoke(
None,
config,
)
else:
# Satisfy the tool invocation by
# providing instructions on the requested changes / change of mind
result = part_3_graph.invoke(
{
"messages": [
ToolMessage(
tool_call_id=event["messages"][-1].tool_calls[0]["id"],
content=f"API call denied by user. Reasoning: '{user_input}'. Continue assisting, accounting for the user's input.",
)
]
},
config,
)
snapshot = part_3_graph.get_state(config)

[mraguth]

@tevslin Great job! Many organizations like to combine their internal data with external data and news. Could we possibly add an agent that can read a pandas dataframe and allow us to ask questions about both internal and external data? Just a thought... It's an excellent use case, and extending it this way could make it even more meaningful and widely accepted in the real world.