fix: enable Laya language detection and harden live analysis (#18 #19 #20)

Ensure Docker images keep their container-built native dependencies, wait for the Laya sidecar healthcheck, expose accurate Laya readiness, and treat empty AI language batches as fallback failures.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

docker-compose.yml

@@ -14,11 +14,14 @@ services:
       - NODE_ENV=production
       - PORT=3000 # Internal container port
       - LAYA_BASE_URL=http://laya:8000
+      - LAYA_BATCH_TIMEOUT_MS=120000
+    depends_on:
+      laya:
+        condition: service_healthy
     devices:
       - /dev/dri:/dev/dri # Required for VAAPI/QuickSync/AMF (Linux)
 
   laya:
-    profiles: ["laya"]
     build:
       context: .
       dockerfile: Dockerfile.laya
@@ -29,4 +32,10 @@ services:
       - LAYA_PRELOAD=0
       - LAYA_HOST=0.0.0.0
       - LAYA_PORT=8000
+    healthcheck:
+      test: ["CMD", "python", "-c", "import urllib.request; urllib.request.urlopen('http://127.0.0.1:8000/health', timeout=3)"]
+      interval: 10s
+      timeout: 5s
+      retries: 12
+      start_period: 30s
     restart: unless-stopped