feat: Claude Code Monitor — lanes, pipelines and a merged workspace
Internal SmartGift build of a Claude Code monitoring dashboard. Lanes: a durable unit of parallel agent work, one per working directory, tracked across session restarts. Managed lanes are git worktrees the dashboard provisions and can reset or remove behind a three-check destroy guard and a counted preflight; adopted lanes are directories you already own and are never destroyable. Pipelines: a lane moves through pipeline stages. A stage the agent declares with evidence renders green; a stage inferred from the tool-event stream renders dashed amber and never counts as done. Detection is forward-only within a 30-minute window, and never writes the declared stage. Workspace: one page at /run with a lane grid, the selected lane's pipeline, and a full Claude console behind a disclosure.
This commit is contained in:
@@ -0,0 +1,112 @@
|
||||
#!/usr/bin/env node
|
||||
|
||||
/**
|
||||
* Claude Code hook handler.
|
||||
* Receives hook event JSON on stdin and forwards it to every live Agent
|
||||
* Dashboard server. Designed to fail silently so it never blocks Claude
|
||||
* Code, and to fan out across multiple dashboards that use **different**
|
||||
* SQLite data directories (e.g. the macOS desktop app alongside `npm run dev`
|
||||
* when each has its own DB). Servers sharing one database receive hooks through
|
||||
* a single ingest port so events are never duplicated.
|
||||
*
|
||||
* Delivery is fire-and-forget: we exit as soon as the request body is on the
|
||||
* wire, WITHOUT waiting for the dashboard's HTTP response. The hook only needs
|
||||
* to *deliver* the event — on loopback the local server reads the buffered
|
||||
* request and processes it even after this short-lived process exits. Waiting
|
||||
* for the response is what made Claude Code sit at "running hooks" for seconds
|
||||
* whenever a dashboard was busy, slow, or wedged.
|
||||
*
|
||||
* @author Nguyễn Ngọc Trí Vĩ <vinnt@smartgift.vn>
|
||||
*/
|
||||
|
||||
const http = require("http");
|
||||
|
||||
const hookType = process.argv[2] || "unknown";
|
||||
|
||||
/**
|
||||
* Resolve every live dashboard server's port via the discovery file. Falls
|
||||
* back to the `CLAUDE_DASHBOARD_PORT` override or the conventional 4820 if
|
||||
* the discovery module can't load for any reason. Never throws.
|
||||
*/
|
||||
function resolvePorts() {
|
||||
try {
|
||||
return require("../server/lib/server-info").resolveHookIngestPorts();
|
||||
} catch {
|
||||
const envPort = parseInt(process.env.CLAUDE_DASHBOARD_PORT || "", 10);
|
||||
return [Number.isInteger(envPort) && envPort > 0 ? envPort : 4820];
|
||||
}
|
||||
}
|
||||
|
||||
const ports = resolvePorts();
|
||||
|
||||
let input = "";
|
||||
|
||||
process.stdin.setEncoding("utf8");
|
||||
process.stdin.on("data", (chunk) => (input += chunk));
|
||||
process.stdin.on("end", () => {
|
||||
let parsedData;
|
||||
try {
|
||||
parsedData = JSON.parse(input);
|
||||
} catch {
|
||||
parsedData = { raw: input };
|
||||
}
|
||||
|
||||
const payload = JSON.stringify({
|
||||
hook_type: hookType,
|
||||
data: parsedData,
|
||||
});
|
||||
const contentLength = Buffer.byteLength(payload);
|
||||
|
||||
// Fan out one POST per live server. Each per-target promise resolves the
|
||||
// moment the request body has been flushed — NOT when the dashboard replies
|
||||
// — so a busy, slow, or wedged dashboard can't stall the hook. Each promise
|
||||
// always resolves (never rejects), so one dead listener can't starve the
|
||||
// others and Promise.all can't be left hanging by a single failure.
|
||||
const sends = ports.map(
|
||||
(port) =>
|
||||
new Promise((resolve) => {
|
||||
let settled = false;
|
||||
const done = () => {
|
||||
if (settled) return;
|
||||
settled = true;
|
||||
resolve();
|
||||
};
|
||||
|
||||
const req = http.request(
|
||||
{
|
||||
hostname: "127.0.0.1",
|
||||
port,
|
||||
path: "/api/hooks/event",
|
||||
method: "POST",
|
||||
headers: {
|
||||
"Content-Type": "application/json",
|
||||
"Content-Length": contentLength,
|
||||
},
|
||||
timeout: 2000,
|
||||
},
|
||||
// Drain any response so the socket closes cleanly if the server does
|
||||
// reply before we exit. We never block on it.
|
||||
(res) => res.resume()
|
||||
);
|
||||
|
||||
req.on("error", done); // dead listener (ECONNREFUSED) — nothing to deliver
|
||||
req.on("timeout", () => {
|
||||
req.destroy();
|
||||
done();
|
||||
});
|
||||
req.write(payload);
|
||||
// The 'end' callback fires once the body is on the wire: delivery is
|
||||
// done and the local server will process it on its own schedule.
|
||||
req.end(done);
|
||||
})
|
||||
);
|
||||
|
||||
// Give the kernel one tick to hand the buffered request bytes to the local
|
||||
// server before our sockets close, then exit. The hook returns in ms.
|
||||
Promise.all(sends).finally(() => setImmediate(() => process.exit(0)));
|
||||
});
|
||||
|
||||
// Safety net — guarantees the hook never blocks Claude Code even if a send
|
||||
// somehow never settles. Shorter than the old 5s wait because we no longer
|
||||
// block on the dashboard's response, only on the request flush.
|
||||
setTimeout(() => process.exit(0), 2500);
|
||||
Reference in New Issue
Block a user