feat(infra): Phase 1.5 진단 도구 3개 + trace 정리

scheduler_status, queue_status, run_verify 추가.
MCP 10개 도구 + NanoClaude wrapper + pre-route 키워드.
worker.py trace print 제거.

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

nanoclaude/services/worker.py

@@ -190,7 +190,7 @@ def _pre_route(message: str) -> dict | None:
             return {"action": "tools", "tool": "document", "operation": "search", "params": {"query": query}}
 
     # 인프라 도구 키워드
-    infra_keywords = ["docker", "컨테이너", "디스크", "용량", "헬스체크", "tailscale", "ollama 모델", "mlx 모델"]
+    infra_keywords = ["docker", "컨테이너", "디스크", "용량", "헬스체크", "tailscale", "ollama 모델", "mlx 모델", "스케줄러", "scheduler", "큐 상태", "queue", "처리 큐", "verify", "검증"]
     if any(k in msg for k in infra_keywords):
         # docker/컨테이너 상태
         if any(k in msg for k in ["docker", "컨테이너"]):
@@ -212,6 +212,15 @@ def _pre_route(message: str) -> dict | None:
         if any(k in msg for k in ["ollama 모델", "mlx 모델"]):
             host = "mlx" if "mlx" in msg else "gpu"
             return {"action": "tools", "tool": "infra", "operation": "models", "params": {"host": host}}
+        # 스케줄러
+        if any(k in msg for k in ["스케줄러", "scheduler"]):
+            return {"action": "tools", "tool": "infra", "operation": "scheduler", "params": {}}
+        # 큐
+        if any(k in msg for k in ["큐 상태", "queue", "처리 큐"]):
+            return {"action": "tools", "tool": "infra", "operation": "queue", "params": {}}
+        # 검증
+        if any(k in msg for k in ["verify", "검증"]):
+            return {"action": "tools", "tool": "infra", "operation": "verify", "params": {"check_name": "gpu-snapshot"}}
 
     # 시스템 상태 질문 — 마커만 반환 (worker에서 비동기 조회)
     if any(k in msg for k in ["추론 모델", "gemma", "젬마", "서버 상태", "시스템 상태"]) or \
@@ -254,11 +263,9 @@ async def run(job: Job) -> None:
         # --- 사전 라우팅 (키워드 기반, EXAONE 스킵) ---
         pre = _pre_route(job.message)
         classify_latency = 0
-        print(f"[TRACE] Job {job.id} pre_route result: {pre}", flush=True)
 
         if pre:
             classification = pre
-            print(f"[TRACE] Job {job.id} PRE-ROUTED: {pre.get('tool','')}.{pre.get('operation','')}", flush=True)
             logger.info("Job %s pre-routed: %s.%s", job.id, pre.get("tool", ""), pre.get("operation", pre.get("action", "")))
         else:
             # --- EXAONE 분류기 호출 ---
@@ -291,7 +298,6 @@ async def run(job: Job) -> None:
         response_text = classification.get("response", "")
         route_prompt = classification.get("prompt", "")
 
-        print(f"[TRACE] Job {job.id} final action='{action}' classification={classification}", flush=True)
         logger.info("Job %s classified as '%s'", job.id, action)
 
         # 대화 기록: 사용자 메시지

nanoclaude/tools/infra_tool.py

@@ -14,6 +14,8 @@ from infra.core.health import service_health, VALID_SERVICES
 from infra.core.system import disk_usage
 from infra.core.network import tailscale_status
 from infra.core.models import ollama_models, mlx_models
+from infra.core.docserver import scheduler_status as _scheduler_status, queue_status as _queue_status
+from infra.core.verify import run_verify as _run_verify, VERIFY_COMMANDS
 
 logger = logging.getLogger(__name__)
 
@@ -111,3 +113,18 @@ async def models(host: str = "gpu") -> dict:
     summary = f"{result.source} on {result.host}: {len(result.models)}개 모델"
     return {"ok": True, "tool": "infra", "operation": "models",
             "data": data, "summary": summary, "error": ""}
+
+
+async def scheduler() -> dict:
+    """Document Server scheduler status."""
+    return await _scheduler_status()
+
+
+async def queue() -> dict:
+    """Document Server queue status."""
+    return await _queue_status()
+
+
+async def verify(check_name: str = "gpu-snapshot") -> dict:
+    """Run predefined verify command."""
+    return await _run_verify(check_name)

nanoclaude/tools/registry.py

@@ -21,7 +21,7 @@ ALLOWED_OPS = {
     "calendar": {"today", "search", "create_draft", "create_confirmed"},
     "email": {"search", "read"},
     "document": {"search", "read"},
-    "infra": {"status", "health", "disk", "network", "models"},
+    "infra": {"status", "health", "disk", "network", "models", "scheduler", "queue", "verify"},
 }
 
 # payload hard limit
@@ -113,6 +113,12 @@ async def _exec_infra(operation: str, params: dict) -> dict:
         return await infra_tool.network()
     elif operation == "models":
         return await infra_tool.models(params.get("host", "gpu"))
+    elif operation == "scheduler":
+        return await infra_tool.scheduler()
+    elif operation == "queue":
+        return await infra_tool.queue()
+    elif operation == "verify":
+        return await infra_tool.verify(params.get("check_name", "gpu-snapshot"))
     return _error("infra", operation, "미구현")