main - Give workers stable names instead of raw pids

Pids change on every restart, making it hard to tell workers apart across
deploys or to notice a specific one crash-looping. Each worker now claims
a name ("Worker A", "Worker B", ...) from a small shared file the first
time it starts, keeps it for its whole lifetime, and never reuses a name
for a different pid — so a restart shows up as a new name appearing while
the old one's alert fades out. Alerts and worker cards use the name; the
pid is still available as a tooltip on the card.

main.py

@@ -113,7 +113,7 @@ function render(d){
   document.getElementById('grid').innerHTML=w.map(function(x){
     var c=workerCol(x.rss_mb),p=(x.rss_mb/maxRss*100).toFixed(0);
     return '<div class="card" style="--c:'+c+'">'+
-      '<div class="card-top"><span class="pid">pid '+x.pid+'</span><span class="cn">'+x.cache_size+'/'+MAX+'</span></div>'+
+      '<div class="card-top"><span class="pid" title="pid '+x.pid+'">'+(x.name||('pid '+x.pid))+'</span><span class="cn">'+x.cache_size+'/'+MAX+'</span></div>'+
       '<div class="rss">'+x.rss_mb+'</div>'+
       '<div class="card-bot"><span class="unit">MB</span><span class="req">'+((x.requests||0)+' req')+'</span></div>'+
       '<div class="track"><div class="fill" style="width:'+p+'%;background:'+c+'"></div></div>'+

similarity.py

@@ -163,6 +163,42 @@ def _update_cpu_percent():
 _MEM_STATS_DIR = "/tmp/similarity_mem_stats"
 os.makedirs(_MEM_STATS_DIR, exist_ok=True)
 
+_NAMES_FILE = os.path.join(_MEM_STATS_DIR, "_names.json")
+_WORKER_NAME_ALPHABET = "ABCDEFGHIJKLMNOPQRSTUVWXYZ"
+
+
+def _assign_worker_name():
+    try:
+        with open(_NAMES_FILE) as f:
+            names = json.load(f)
+    except (OSError, json.JSONDecodeError):
+        names = {}
+
+    pid_key = str(os.getpid())
+    if pid_key in names:
+        return names[pid_key]
+
+    used = set(names.values())
+    for letter in _WORKER_NAME_ALPHABET:
+        candidate = f"Worker {letter}"
+        if candidate not in used:
+            break
+    else:
+        candidate = f"Worker {len(names) + 1}"
+
+    names[pid_key] = candidate
+    try:
+        with open(_NAMES_FILE, "w") as f:
+            json.dump(names, f)
+    except OSError:
+        pass
+    return candidate
+
+
+# Assigned once at import time and kept for the process's lifetime — a name is
+# never reused for a different pid, so a restart is visible as a new name.
+_worker_name = _assign_worker_name()
+
 
 def _read_cgroup_memory():
     # cgroup v2 (modern Docker)
@@ -219,6 +255,7 @@ def get_memory_stats():
 
     stats = {
         "pid": os.getpid(),
+        "name": _worker_name,
         "rss_mb": rss_mb,
         "cache_size": cache_size,
         "requests": _request_count,
@@ -259,6 +296,8 @@ def start_memory_trimmer(interval_seconds=300):
 def _read_all_worker_files():
     entries = []
     for filename in os.listdir(_MEM_STATS_DIR):
+        if filename.startswith("_"):
+            continue  # internal files (e.g. _names.json), not per-worker stats
         fpath = os.path.join(_MEM_STATS_DIR, filename)
         try:
             with open(fpath) as f:
@@ -284,18 +323,20 @@ def get_dashboard_data():
                 pass
             continue
 
+        name = stats.get("name") or f"pid {stats['pid']}"
+
         if age > _STALE_AFTER_SECONDS:
-            alerts.append(f"worker {stats['pid']} sem reportar há {int(age)}s")
+            alerts.append(f"{name} sem reportar há {int(age)}s")
             continue
 
         alive.append(stats)
 
         if (stats.get("uptime_seconds", 0) > _RSS_SPIKE_MIN_UPTIME_SECONDS
                 and stats.get("rss_delta_2min_mb", 0) > _RSS_SPIKE_MB):
-            alerts.append(f"worker {stats['pid']} RAM subiu {stats['rss_delta_2min_mb']:.0f} MB em 2 min")
+            alerts.append(f"{name} RAM subiu {stats['rss_delta_2min_mb']:.0f} MB em 2 min")
 
         if stats.get("evictions_per_min", 0) > _CACHE_THRASH_EVICTIONS_PER_MIN:
-            alerts.append(f"worker {stats['pid']} cache a saturar (~{stats['evictions_per_min']:.0f} evicções/min)")
+            alerts.append(f"{name} cache a saturar (~{stats['evictions_per_min']:.0f} evicções/min)")
 
     alive.sort(key=lambda s: s["pid"])