DOC: How to add subgraphs？

[lee0v0]

Issue with current documentation:

I seem to have not found the related content on adding subgraphs in the documentation. For example, if there are three subgraphs that are three modules, how do I connect these subgraphs together?

Idea or request for content:

I currently have three subgraphs: the message summary subgraph, the self-rag subgraph, and the tool invocation subgraph. Each subgraph is a module. How do I piece together the subgraphs of the three modules?

[hinthornw]

We have an example of this here: https://github.com/langchain-ai/langgraph/blob/main/examples/multi_agent/hierarchical_agent_teams.ipynb

But will keep this issue open as we should improve our documentation of it

[russell-dot-js]

Some clear docs outside of a full blown example would be awesome! When reading through that full blown implementation, it can be hard to find the important parts.

Building on the example given above, it would be great to see some concise documentation showing how research_graph has its own ResearchTeamState, the authoring_graph has its DocWritingState, and how the super_graph orchestrates between the two of them, only passing the relevant state from its SuperState. To really illustrate how this works, it would be nice if the super_graph in this example had a more complex state than just "messages", for example:

SuperState:
  research_states: list[ResearchState]
  authoring_states: list[DocWritingState]
  # add messages, whatever else

Then it would be easy to illustrate how the supervisor_node adds to either the research_states or authoring_states and then how the super_graph delegates to the correct subgraph with only its state

[hinthornw]

Added an example #194

We will work to add better native support for composition, so I'll leave this ticket open for now

[arthurGrigo]

At the end of the example it looks like the 'Research team' is called multiple times and replies with the same response. Do I get something wrong or is this intended to show how suboptimal it works?