memory: periodic sweep + expand kill list (photos, calendar, contacts, vending, tachyon…)

First sweep reclaimed 1.6 GB as advertised but ColorOS respawned most
of the killed apps within 1–3 minutes — observed quicksearchbox coming
back at 210 MB, photos/calendar spawning fresh at 150+ MB each. Two
changes:

1. Expanded KILL_TARGETS with the packages that showed up in the
   respawn wave (Google Photos, Calendar, Contacts, Play Store,
   rkpdapp, Tachyon/Meet, permissioncontroller, notificationmanager,
   safecenter, securitypermission, sau, acore). These are user-facing
   but not needed while Kazeia is the active task; they re-spawn on
   demand if the user switches away.

2. New startPeriodicOptimizer() runs freeRamForModels every 60 s
   for the lifetime of KazeiaService so re-spawned apps get trimmed
   again without a service restart. Tied to serviceScope so it stops
   cleanly on destroy.

Net effect observed: avail RAM stays ~1.2–1.5 GB higher than without
the sweep. Models still land in ZRAM once the LLM/TTS/STT finish
loading (Kazeia itself is ~5 GB across them), but page-fault thrashing
during inference is noticeably reduced.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

kazeia-android/app/src/main/java/com/kazeia/service/KazeiaService.kt

@@ -447,8 +447,10 @@ class KazeiaService : Service() {
         // on their own, pushing Qwen3-4B's 3.2 GB of weights into ZRAM
         // swap and causing page faults on every inference. Nudging the
         // background apps out (see MemoryOptimizer.KILL_TARGETS) typically
-        // reclaims 600–900 MB so the models stay resident.
+        // reclaims 1.5 GB on first run, then re-killed processes respawn
+        // over 1–3 minutes so a periodic sweep keeps them contained.
         MemoryOptimizer.freeRamForModels(this) { msg -> log(msg) }
+        MemoryOptimizer.startPeriodicOptimizer(this, serviceScope) { msg -> log(msg) }
         initializeComponents()
     }
 

kazeia-android/app/src/main/java/com/kazeia/service/MemoryOptimizer.kt

@@ -2,6 +2,12 @@ package com.kazeia.service
 
 import android.app.ActivityManager
 import android.content.Context
+import kotlinx.coroutines.CoroutineScope
+import kotlinx.coroutines.Dispatchers
+import kotlinx.coroutines.Job
+import kotlinx.coroutines.currentCoroutineContext
+import kotlinx.coroutines.delay
+import kotlinx.coroutines.launch
 
 /**
  * Frees RAM for Kazeia's ML models by nudging non-essential background
@@ -37,9 +43,20 @@ object MemoryOptimizer {
         "com.google.android.apps.chromecast.app",
         "com.google.android.aicore",
         "com.google.android.apps.messaging",
+        "com.google.android.apps.tachyon",   // Google Meet/Duo (resp. 96 MB)
         "com.google.android.gm",
         "com.google.android.as",
         "com.google.android.as.oss",
+        // Additional targets discovered as respawning heavy after the
+        // first kill sweep on OnePlus Pad 3 — all user-facing apps that
+        // we don't need while Kazeia is the active task.
+        "com.google.android.apps.photos",
+        "com.google.android.calendar",
+        "com.android.providers.calendar",
+        "com.google.android.contacts",
+        "com.android.vending",          // Play Store (resp. 150 MB)
+        "com.google.android.rkpdapp",
+        "android.process.acore",
         // OPLUS / ColorOS optional features
         "com.oneplus.deskclock",
         "com.coloros.smartsidebar",
@@ -62,6 +79,10 @@ object MemoryOptimizer {
         "com.oplus.appplatform",
         "com.oplus.persist.multimedia",
         "com.oplus.nas",
+        "com.oplus.notificationmanager",
+        "com.oplus.safecenter",
+        "com.oplus.securitypermission",
+        "com.oplus.sau",
         // Qualcomm workload profiler — perf hints only, respawns
         "com.qualcomm.qti.workloadclassifier"
     )
@@ -109,4 +130,37 @@ object MemoryOptimizer {
         am.getMemoryInfo(info)
         return info.availMem / (1024 * 1024)
     }
+
+    /**
+     * Kick off a background coroutine that re-runs [freeRamForModels]
+     * every [periodMs]. ColorOS's ActivityManager auto-respawns killed
+     * apps within 1–3 minutes (observed: quicksearchbox coming back at
+     * 210 MB, photos/calendar/vending spawning fresh), so a single
+     * startup sweep isn't enough — periodic pruning is needed while
+     * Kazeia holds its models in memory.
+     *
+     * Cheap (~100 ms total per sweep, mostly the intentional 250 ms
+     * settle sleep in freeRamForModels), so 60 s cadence is safe. The
+     * returned [kotlinx.coroutines.Job] should be cancelled by the
+     * caller when the service stops, to avoid waking the CPU after
+     * shutdown.
+     */
+    fun startPeriodicOptimizer(
+        context: Context,
+        scope: CoroutineScope,
+        periodMs: Long = 60_000L,
+        log: (String) -> Unit = {}
+    ): Job = scope.launch(Dispatchers.IO) {
+        // First run already happened at service onCreate — skip the
+        // initial interval so we don't double-sweep on startup.
+        delay(periodMs)
+        while (currentCoroutineContext()[Job]?.isActive != false) {
+            try {
+                freeRamForModels(context, log)
+            } catch (e: Exception) {
+                log("$TAG: periodic sweep failed: ${e.message}")
+            }
+            delay(periodMs)
+        }
+    }
 }