Docs · Quick start

LangGraph

Your graph on the board while it runs: each node updates its card as it finishes.

Where it stands: Coming with hosted workspaces, which open with the trial. Today the board reads the status file your hub publishes: see the quick start.

Steps

Make a write key. In your workspace, open Keys and make a write key for this bot. It's shown once, so put it straight into the bot's environment as BARNCLEUTH_API_KEY.

Run the graph through this. It streams the compiled graph's updates, the same ones you'd get yourself.

import json, os, urllib.request

def barncleuth(state, note=""):
    body = json.dumps({"bot_id": "my-graph", "state": state, "note": note}).encode()
    request = urllib.request.Request("https://barncleuth.com/api/heartbeat", data=body, headers={
        "Authorization": "Bearer " + os.environ["BARNCLEUTH_API_KEY"],
        "Content-Type": "application/json",
    })
    urllib.request.urlopen(request, timeout=10).read()

def run(app, inputs):
    barncleuth("LIVE", "Graph started")
    for update in app.stream(inputs, stream_mode="updates"):
        for node in update:
            barncleuth("LIVE", f"Finished {node}")
    barncleuth("DONE", "Graph finished")

Use it. Compile as usual, then run.

app = builder.compile()
run(app, {"topic": "Mars"})

What it sends

The bot's id, its state and node names. Never the graph's state.

How it's tested

Tested in Python against a stand-in shaped like a compiled graph's stream. Not against a live LangGraph install yet.