diff --git a/app/api/dashboard.py b/app/api/dashboard.py
index 8263b7c..94bdbc8 100644
--- a/app/api/dashboard.py
+++ b/app/api/dashboard.py
@@ -35,6 +35,20 @@ class PipelineStatus(BaseModel):
     count: int
 
 
+class QueueLag(BaseModel):
+    """파이프라인 stage 별 처리 지연 — 운영 카드용.
+
+    pipeline_status 는 24h 누적 통계라 현재 적체 신호로 부족.
+    queue_lag 는 현재 시점 pending/processing/failed + oldest pending age 로
+    "지금 막힌 게 있는가" 를 보여준다.
+    """
+    stage: str
+    pending: int
+    processing: int
+    failed: int
+    oldest_pending_age_sec: int | None  # 가장 오래된 pending 의 created_at 기준 경과 (초)
+
+
 class DashboardResponse(BaseModel):
     today_added: int
     today_by_domain: list[DomainCount]
@@ -48,6 +62,10 @@ class DashboardResponse(BaseModel):
     documents_count: int = 0
     memos_count: int = 0
     news_count: int = 0
+    # §4 — category 기반 카드 + 승인 pending + queue lag
+    category_counts: dict[str, int] = {}
+    library_pending_suggestions: int = 0
+    queue_lag: list[QueueLag] = []
 
 
 @router.get("/", response_model=DashboardResponse)
@@ -130,6 +148,56 @@ async def get_dashboard(
     memos_count = counts[2]
     news_count = counts[3]
 
+    # §4 — 카테고리별 count (§1 documents.category enum)
+    cat_result = await session.execute(
+        text("""
+            SELECT category, COUNT(*)
+            FROM documents
+            WHERE deleted_at IS NULL AND category IS NOT NULL
+            GROUP BY category
+        """)
+    )
+    category_counts = {row[0]: row[1] for row in cat_result.all()}
+
+    # §4 — 승인 대기 (library 제안)
+    pending_result = await session.execute(
+        text("""
+            SELECT COUNT(*)
+            FROM documents
+            WHERE deleted_at IS NULL
+              AND ai_suggestion IS NOT NULL
+              AND ai_suggestion->>'proposed_category' = 'library'
+        """)
+    )
+    library_pending_suggestions = pending_result.scalar() or 0
+
+    # §4 — queue lag (현재 시점 stage 별 적체 신호)
+    # extract/classify/embed 외에 stt/thumbnail (§3) 도 자동 포함.
+    lag_result = await session.execute(
+        text("""
+            SELECT
+                stage,
+                COUNT(*) FILTER (WHERE status='pending')        AS pending,
+                COUNT(*) FILTER (WHERE status='processing')     AS processing,
+                COUNT(*) FILTER (WHERE status='failed')         AS failed,
+                EXTRACT(EPOCH FROM (NOW() - MIN(created_at) FILTER (WHERE status='pending')))::int
+                                                                AS oldest_pending_age_sec
+            FROM processing_queue
+            GROUP BY stage
+            ORDER BY stage
+        """)
+    )
+    queue_lag = [
+        QueueLag(
+            stage=row[0],
+            pending=row[1] or 0,
+            processing=row[2] or 0,
+            failed=row[3] or 0,
+            oldest_pending_age_sec=row[4],
+        )
+        for row in lag_result.all()
+    ]
+
     return DashboardResponse(
         today_added=today_added,
         today_by_domain=[
@@ -156,4 +224,7 @@ async def get_dashboard(
         documents_count=documents_count,
         memos_count=memos_count,
         news_count=news_count,
+        category_counts=category_counts,
+        library_pending_suggestions=library_pending_suggestions,
+        queue_lag=queue_lag,
     )
diff --git a/app/models/queue.py b/app/models/queue.py
index 0110f65..9fcd94d 100644
--- a/app/models/queue.py
+++ b/app/models/queue.py
@@ -16,7 +16,15 @@ class ProcessingQueue(Base):
     id: Mapped[int] = mapped_column(BigInteger, primary_key=True)
     document_id: Mapped[int] = mapped_column(BigInteger, ForeignKey("documents.id"), nullable=False)
     stage: Mapped[str] = mapped_column(
-        Enum("extract", "classify", "summarize", "embed", "chunk", "preview", name="process_stage"), nullable=False
+        # 'stt' (audio): migration 150 / 'thumbnail' (video): queue_consumer 가 enqueue.
+        # DB enum 변경은 마이그레이션이 처리하므로 create_type=False.
+        Enum(
+            "extract", "classify", "summarize", "embed", "chunk", "preview",
+            "stt", "thumbnail",
+            name="process_stage",
+            create_type=False,
+        ),
+        nullable=False,
     )
     status: Mapped[str] = mapped_column(
         Enum("pending", "processing", "completed", "failed", name="process_status"),
diff --git a/docker-compose.yml b/docker-compose.yml
index 91904ab..d78b833 100644
--- a/docker-compose.yml
+++ b/docker-compose.yml
@@ -73,7 +73,9 @@ services:
               count: 1
               capabilities: [gpu]
     healthcheck:
-      test: ["CMD", "python3", "-c", "import urllib.request; urllib.request.urlopen('http://localhost:3300/health')"]
+      # /ready: CUDA 디바이스 + 모델 적재 둘 다 확인. ready=true 만 healthy 처리.
+      # /health 는 단순 liveness 라 모델 미적재 상태도 healthy 로 잡혀 운영 신호로 부적합.
+      test: ["CMD", "python3", "-c", "import json,urllib.request,sys; r=urllib.request.urlopen('http://localhost:3300/ready'); sys.exit(0 if json.load(r).get('ready') else 1)"]
       interval: 30s
       timeout: 10s
       retries: 3
diff --git a/frontend/src/lib/stores/system.ts b/frontend/src/lib/stores/system.ts
index 3cfbb5b..d442de4 100644
--- a/frontend/src/lib/stores/system.ts
+++ b/frontend/src/lib/stores/system.ts
@@ -26,6 +26,14 @@ export interface PipelineStatus {
   count: number;
 }
 
+export interface QueueLag {
+  stage: string;
+  pending: number;
+  processing: number;
+  failed: number;
+  oldest_pending_age_sec: number | null;
+}
+
 export interface DashboardSummary {
   today_added: number;
   today_by_domain: DomainCount[];
@@ -38,6 +46,10 @@ export interface DashboardSummary {
   documents_count: number;
   memos_count: number;
   news_count: number;
+  // §4 — category 카드 + 승인 pending + queue lag
+  category_counts: Record<string, number>;
+  library_pending_suggestions: number;
+  queue_lag: QueueLag[];
 }
 
 const POLL_INTERVAL_MS = 60_000;
diff --git a/services/stt/server.py b/services/stt/server.py
index 6258b34..decc778 100644
--- a/services/stt/server.py
+++ b/services/stt/server.py
@@ -1,16 +1,37 @@
 """STT 마이크로서비스 — faster-whisper (GPU) 기반 음성 전사.
 
-filePath → {text, segments:[{start,end,text}]}. 모델은 첫 요청 시 lazy loading.
+filePath → {text, segments:[{start,end,text}]}.
+모델은 startup 에서 eager preload (Docker /ready healthcheck 가 모델 적재까지 검증).
 기본 모델 large-v3 (VRAM ~3GB, float16). 환경변수로 교체 가능.
+
+환경변수 `STT_PRELOAD=0` 으로 lazy 로 강제 가능 (개발/테스트용).
 """
 
+import logging
 import os
 import unicodedata
+from contextlib import asynccontextmanager
 from pathlib import Path
 
 from fastapi import FastAPI
 
-app = FastAPI()
+logger = logging.getLogger("stt")
+
+
+@asynccontextmanager
+async def lifespan(_app: FastAPI):
+    # startup: 모델 eager preload 시도. 실패해도 프로세스는 살아 있고
+    # /ready 가 false 로 남아 healthcheck 가 unhealthy 처리.
+    if os.getenv("STT_PRELOAD", "1") != "0":
+        try:
+            _load_model()
+            logger.info("stt model preloaded: %s (%s, %s)", _MODEL_NAME, _DEVICE, _COMPUTE_TYPE)
+        except Exception as e:
+            logger.exception("stt model preload failed: %s", e)
+    yield
+
+
+app = FastAPI(lifespan=lifespan)
 
 _model = None
 _MODEL_NAME = os.getenv("WHISPER_MODEL", "large-v3")