echoclaw-relay-agent 0.17.2 → 0.18.0

package/dist/cli.js

@@ -255,6 +255,30 @@ async function runSetup(code, relay, bridgePort) {
             }
         }
         catch { /* no service manager on this platform — fine */ }
+        // Kill any remaining relay-agent processes not managed by service manager.
+        // Use narrow match: only our own binary name, exclude our own PID.
+        try {
+            const { execSync } = await import('child_process');
+            const myPid = process.pid.toString();
+            try {
+                execSync(`pgrep -f "echoclaw-relay-agent.*--gateway" | grep -v "^${myPid}$" | xargs kill 2>/dev/null`, { stdio: 'ignore' });
+            }
+            catch { /* no matching processes — fine */ }
+            // Poll until processes are gone (max 5s)
+            const deadline = Date.now() + 5000;
+            while (Date.now() < deadline) {
+                try {
+                    const out = execSync(`pgrep -f "echoclaw-relay-agent.*--gateway" | grep -v "^${myPid}$"`, { stdio: 'pipe' }).toString().trim();
+                    if (!out)
+                        break;
+                }
+                catch {
+                    break; /* pgrep returns non-zero = no matches */
+                }
+                await new Promise(r => setTimeout(r, 300));
+            }
+        }
+        catch { /* pgrep not available — fine */ }
         const sid = existingSession?.relaySessionId?.slice?.(0, 8) ?? 'unknown';
         console.log(`  ${DIM}Clearing previous session (${sid}...)${RESET}`);
         try {
@@ -645,32 +669,35 @@ async function runDaemon(relay, code, gateway, bridgePort) {
     agent.on('connected', () => {
         updater.startPeriodicCheck();
     });
-    // Retry logic — transient failures (network, server restart) shouldn't kill the service
-    const MAX_RETRIES = 5;
-    const RETRY_DELAY = 3000;
-    let lastErr;
-    for (let attempt = 1; attempt <= MAX_RETRIES; attempt++) {
+    // V2: Never-give-up retry logic — daemon must stay alive forever.
+    // Only exits on: explicit UNPAIR, SIGINT/SIGTERM, or no saved session + no code.
+    const BASE_DELAY = 3000;
+    const MAX_DELAY = 60000;
+    let attempt = 0;
+    while (!stopping) {
         try {
+            attempt++;
             await agent.start(code);
-            return; // success — stay alive
+            // Connected — reset attempt counter
+            attempt = 0;
+            // agent.start() resolved means connection dropped, loop to reconnect
         }
         catch (err) {
-            lastErr = err;
             const msg = err instanceof Error ? err.message : String(err);
-            console.error(`  [start-failed] attempt=${attempt}/${MAX_RETRIES} error=${msg}`);
-            // Permanent errors — don't retry
-            if (msg.includes('No saved session found') ||
-                msg.includes('DEVICE_TOKEN_MISMATCH') ||
-                msg.includes('SESSION_NOT_FOUND')) {
+            console.error(`  [start-failed] attempt=${attempt} error=${msg}`);
+            // Truly permanent: no session AND no code to pair with
+            if (msg.includes('No saved session found') && !code) {
+                console.error(`  [fatal] No session and no pairing code. Run: echoclaw-relay setup <CODE>`);
                 break;
             }
-            if (attempt < MAX_RETRIES) {
-                console.log(`  [retry] waiting ${RETRY_DELAY / 1000}s...`);
-                await new Promise(r => setTimeout(r, RETRY_DELAY));
-            }
+            // SESSION_NOT_FOUND is NOT permanent in V2 — identity reconnect will handle it
+            // Just retry with backoff
+            // Exponential backoff: 3s, 6s, 12s, 24s, 48s, 60s (capped)
+            const delay = Math.min(BASE_DELAY * Math.pow(2, attempt - 1), MAX_DELAY);
+            console.log(`  [retry] waiting ${(delay / 1000).toFixed(0)}s...`);
+            await new Promise(r => setTimeout(r, delay));
         }
     }
-    throw lastErr;
 }
 // ── Main ─────────────────────────────────────────────────────
 async function main() {
@@ -710,4 +737,16 @@ async function main() {
         process.exit(1);
     }
 }
+// V2: Global error handlers — log and exit, let supervisor (launchd/systemd) restart.
+// Staying alive after uncaughtException risks corrupted state (half-open sockets,
+// partial writes). Clean crash + restart is safer than zombie process.
+process.on('uncaughtException', (err) => {
+    console.error(`  [FATAL] uncaughtException: ${err.message}`);
+    console.error(err.stack || '');
+    process.exit(1);
+});
+process.on('unhandledRejection', (reason) => {
+    console.error(`  [FATAL] unhandledRejection: ${reason}`);
+    process.exit(1);
+});
 main();

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "echoclaw-relay-agent",
-  "version": "0.17.2",
+  "version": "0.18.0",
   "description": "EchoClaw Relay Connection — E2E encrypted relay transport, pairing, and session management",
   "type": "module",
   "main": "./dist/index.js",