ai_co_discuss 0.3.84 → 0.3.85

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "ai_co_discuss",
-  "version": "0.3.84",
+  "version": "0.3.85",
   "bin": {
     "codiscuss": "runner/codiscuss.mjs"
   },

package/runner/discussion_local.mjs

@@ -9,7 +9,7 @@ import { parseArgs } from 'node:util';
 
 import { atomic_json } from './files.mjs';
 import { acquire_lock } from './discussion_runner.mjs';
-import { assert_shared_home, main as service_main, service_home, service_pid, service_ready, start_service } from './discussion_service.mjs';
+import { assert_shared_home, main as service_main, service_home, service_pid, service_ready, service_stopped_by_user, start_service } from './discussion_service.mjs';
 import { detect_python, local_settings } from './discussion_setup.mjs';
 import { unlink_machine } from './remote_launch.mjs';
 import { tone } from './discussion_tui.mjs';
@@ -20,6 +20,8 @@ const VERSION = JSON.parse(readFileSync(new URL('../package.json', import.meta.u
 const START_WAIT_MS = 15000;
 const STOP_WAIT_MS = 25000;
 const POLL_MS = 200;
+// A service that exited without `service stop` (e.g. an old one adopted during an upgrade) is restarted at most this often.
+const SERVICE_REVIVE_MS = 10_000;
 const CONNECTION_TOKEN_BYTES = 32;
 const LOCAL_HELP = `用法：codiscuss local open|start|status|stop [--json]
       codiscuss local connect SITE [--setup-key KEY] [--no-open]
@@ -156,11 +158,17 @@ async function close_owned(child, owned_service, home) {
   }
 }
 
-/** Handle explicit restart requests in the supervisor so replacement service ownership stays local. */
+let revived_at = 0;
+
+/**
+ * Start the service on an explicit request, or again when it exited without `service stop`, so replacement
+ * ownership stays local and an adopted service that exits later does not leave every room unattended.
+ */
 async function requested_service(home, status, owned_service) {
   const path = join(home, 'local', 'start.request');
-  if (!existsSync(path)) return owned_service;
-  unlinkSync(path);
+  if (existsSync(path)) unlinkSync(path);
+  else if (service_pid(home) || service_stopped_by_user(home) || Date.now() - revived_at < SERVICE_REVIVE_MS) return owned_service;
+  revived_at = Date.now();
   const status_path = join(home, 'local', 'status.json');
   try {
     const service = await start_service(home);

package/runner/discussion_service.mjs

@@ -32,6 +32,8 @@ const REMOTE_SYNC_MS = 3000;
 const RELAY_RESTART_MIN_MS = 1000;
 const START_WAIT_MS = 5000;
 const STOP_WAIT_MS = 20_000;
+// Left by `service stop` and cleared by any start: the local supervisor restarts a service that exited only without it.
+const STOPPED_MARKER = 'service.stopped';
 const REGISTRY_WAIT_MS = 5000;
 // Housekeeping: logs above the limit keep only their newest part; a launch folder no room uses is removed once it is this old.
 const SWEEP_MS = 6 * 60 * 60 * 1000;
@@ -744,6 +746,7 @@ export async function run_service(home, output = () => {}) {
 /** Start the detached service once; an already running service is reused. */
 export async function start_service(home) {
   assert_shared_home(home);
+  rmSync(join(home, STOPPED_MARKER), { force: true });
   const running = service_pid(home);
   if (service_ready(home, running)) return { running: true, ready: true, pid: running, started: false };
   mkdirSync(home, { recursive: true });
@@ -793,6 +796,11 @@ function read_persisted_env() {
   } catch { return {}; }
 }
 
+/** Whether the user stopped the service on purpose, so the local supervisor must not bring it back. */
+export function service_stopped_by_user(home) {
+  return existsSync(join(home, STOPPED_MARKER));
+}
+
 /** Ask the service to pause its runs and exit, then wait for its lock to be released. */
 async function stop_service(home) {
   if (!service_pid(home)) return { running: false };
@@ -865,7 +873,9 @@ export async function main(args = process.argv.slice(2), print = value => proces
   switch (command) {
     case 'run': return run_service(home);
     case 'start': return print(await start_service(home));
-    case 'stop': return print(await stop_service(home));
+    case 'stop':
+      atomic_json(join(home, STOPPED_MARKER), { stopped_at: Math.floor(Date.now() / 1000) });
+      return print(await stop_service(home));
     case 'restart': await stop_service(home); return print(await start_service(home));
     case 'status': return print(service_status(home));
     case 'open': return print({ ...(await start_service(home)), opened: open_sites(home) });