Is it possible to combine two graphs?

[mraguth]

In my example, I like to combine app and graph graphs. Here my app output is input string to 'graph' workflow. Is that possible?
I am using supervisor workflow from this link:
https://langchain-ai.github.io/langgraph/tutorials/multi_agent/agent_supervisor/#construct-graph

please hlep:

# Compile both workflows
app = workflow.compile()
output = app.invoke(question)
graph = supervisor_workflow.compile()

a supervisor code sample is like below:

import functools
import operator
from typing import Sequence
from typing_extensions import TypedDict

from langchain_core.messages import BaseMessage

from langgraph.graph import END, StateGraph, START
from langgraph.prebuilt import create_react_agent


# The agent state is the input to each node in the graph
class AgentState(TypedDict):
    # The annotation tells the graph that new messages will always
    # be added to the current states
    messages: Annotated[Sequence[BaseMessage], operator.add]
    # The 'next' field indicates where to route to next
    next: str


research_agent = create_react_agent(llm, tools=[tavily_tool])
research_node = functools.partial(agent_node, agent=research_agent, name="Researcher")

# NOTE: THIS PERFORMS ARBITRARY CODE EXECUTION. PROCEED WITH CAUTION
code_agent = create_react_agent(llm, tools=[python_repl_tool])
code_node = functools.partial(agent_node, agent=code_agent, name="Coder")

workflow = StateGraph(AgentState)
workflow.add_node("Researcher", research_node)
workflow.add_node("Coder", code_node)
workflow.add_node("supervisor", supervisor_agent)

for member in members:
    # We want our workers to ALWAYS "report back" to the supervisor when done
    workflow.add_edge(member, "supervisor")
# The supervisor populates the "next" field in the graph state
# which routes to a node or finishes
conditional_map = {k: k for k in members}
conditional_map["FINISH"] = END
workflow.add_conditional_edges("supervisor", lambda x: x["next"], conditional_map)
# Finally, add entrypoint
workflow.add_edge(START, "supervisor")

graph = workflow.compile()

[dosubot[bot]]

Yes, it is possible to combine two graphs, specifically using an app's output as input to a 'graph' workflow, within the LangGraph multi-agent supervisor workflow. You can define a workflow using StateGraph where nodes and edges are set up to process data through a series of steps. The workflow is compiled into an app, which can be invoked with input messages. The output from one step can be used as input for the next step in the workflow, allowing you to chain processes and use outputs as inputs within the LangGraph framework ^[1].

To continue talking to Dosu, mention @dosu.

[mraguth]

@dosu. can you give me a example code please?

[dosubot[bot]]

Certainly! Here's an example of how you can combine two graphs using the LangGraph framework. This example assumes you have two workflows: app_workflow and graph_workflow. The output of app_workflow is used as the input to graph_workflow.

import functools
from langchain_core.messages import BaseMessage
from langgraph.graph import StateGraph, START, END
from langgraph.prebuilt import create_react_agent

# Define the agent state
class AgentState(TypedDict):
    messages: Sequence[BaseMessage]
    next: str

# Create agents
llm = ...  # Your language model
tavily_tool = ...  # Define your tool
python_repl_tool = ...  # Define your tool

research_agent = create_react_agent(llm, tools=[tavily_tool])
research_node = functools.partial(agent_node, agent=research_agent, name="Researcher")

code_agent = create_react_agent(llm, tools=[python_repl_tool])
code_node = functools.partial(agent_node, agent=code_agent, name="Coder")

# Define the first workflow (app_workflow)
app_workflow = StateGraph(AgentState)
app_workflow.add_node("Researcher", research_node)
app_workflow.add_edge(START, "Researcher")
app_workflow.add_edge("Researcher", END)

# Compile the first workflow
app = app_workflow.compile()

# Define the second workflow (graph_workflow)
graph_workflow = StateGraph(AgentState)
graph_workflow.add_node("Coder", code_node)
graph_workflow.add_edge(START, "Coder")
graph_workflow.add_edge("Coder", END)

# Compile the second workflow
graph = graph_workflow.compile()

# Invoke the first workflow and use its output as input to the second
app_output = app.invoke({"messages": []})  # Assuming the input is a dictionary with messages
graph_input = {"messages": app_output["messages"]}  # Use the output messages as input
graph_output = graph.invoke(graph_input)

print(graph_output)

This code sets up two separate workflows and demonstrates how to pass the output from the first (app_workflow) as input to the second (graph_workflow). Adjust the tools and language model (llm) as needed for your specific use case.

To continue talking to Dosu, mention @dosu.

[mraguth]

@dosu.... how I input my 'question' to app.invoke?

app_output = app.invoke({"messages": []}) # Assuming the input is a dictionary with messages

[dosubot[bot]]

To input your 'question' into app.invoke, you should structure your input as a dictionary with the key "keys" containing another dictionary with the key "question" and your question as its value. Here's how you can do it:

inputs = {"keys": {"question": "Your question here"}}
app_output = app.invoke(inputs)

Replace "Your question here" with your actual question. This format is used to process the question through the defined nodes and edges in the LangGraph framework ^[1].

To continue talking to Dosu, mention @dosu.