Since LangSmith Deployment literally only supports building and deploying LangGraph OSS graphs, I'd be fine with being more specific/precise here: Deployment focuses on how to build and deploy your LangGraph graphs as LangGraph Servers.

Not blocking.

Side bar: I'm really averse to using "agentic workflows" here because there's been so much discussion/controversy around naming and definitions of things (e.g. OpenAI Dev Day Agent Builder release, Harrison's follow-up blog post). Because of the quickly evolving nature of these things, my preference is that our documentation is dumb and simple and decoupled from this noise. For example, in 6 months, "agentic workflows" might mean something else and/or there will be a new term: "agentic work streams".

I think this is a good point. I removed this instance of "agentic workflows". It would be worthwhile to do a bit of a follow-up on this — I want the docs terminology to be as clear and relevant for as long as possible.

Suggestion: To be consistent with the new hosting.mdx page, LangGraph app --> LangGraph Server

title: Run a LangGraph Server locally
sidebarTitle: Run a LangGraph Server locally

There are other places where we use the term app, LangGraph app, and LangGraph application, and LangSmith application. For consistency, should we change all instances to LangGraph Server? I won't comment on all places in this PR. Also, this can probably be done as part of a different PR.

Side bar: I won't comment in all places, but we seem to be using LangGraph Server and LangSmith application interchangeably in the new docs. This is increasing the confusion IMHO. Can we only use 1 term?

Side, side bar: Do we need a Glossary of Terms page? This might be helpful.

Suggestion: Platform components --> Deployment components

The items under this section are only for LangSmith Deployment. Platform feels too wide/vague (for now).

Nit, but maybe disregard this comment: project & tenant IDs --> deployment (project_id) & workspace IDs (tenant_id)

A long time ago, we used to call deployments, "projects" and we used to call workspaces, "tenants". We never updated the GET /info endpoint. Source code:

async def meta_info(request: ApiRequest):
    plus = plus_features_enabled()
    return JSONResponse(
        {
            "version": __version__,
            "langgraph_py_version": langgraph.version.__version__,
            "flags": {
                "assistants": True,
                "crons": plus and config.FF_CRONS_ENABLED,
                "langsmith": bool(config.LANGSMITH_CONTROL_PLANE_API_KEY)
                and bool(config.TRACING),
                "langsmith_tracing_replicas": True,
            },
            "host": {
                "kind": metadata.HOST,
                "project_id": metadata.PROJECT_ID,
                "host_revision_id": metadata.HOST_REVISION_ID,
                "revision_id": metadata.REVISION,
                "tenant_id": metadata.TENANT_ID,
            },
        }
    )