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>

infra/core/docserver.py

@@ -0,0 +1,63 @@
+"""Document Server diagnostic tools — scheduler and queue status."""
+
+from __future__ import annotations
+
+import json
+import re
+from datetime import datetime, timezone
+
+from ..config import HOSTS
+from ..schemas import BaseResult
+from .ssh import run_command, run_local, SSHError, _is_local_host
+
+
+def _now() -> str:
+    return datetime.now(timezone.utc).isoformat()
+
+
+async def scheduler_status() -> dict:
+    """Get APScheduler job status from Document Server logs."""
+    cfg = HOSTS["gpu"]
+    cmd = "docker logs hyungi_document_server-fastapi-1 --tail 100 2>&1 | grep -iE 'scheduler|apscheduler|job|trigger|cron|interval' | tail -20"
+
+    try:
+        if _is_local_host(cfg):
+            stdout, _ = await run_local(cmd)
+        else:
+            stdout, _ = await run_command(cfg, cmd)
+    except SSHError as e:
+        return {"ok": False, "checked_at": _now(), "error_type": e.error_type,
+                "error": str(e), "data": [], "summary": ""}
+
+    lines = [l.strip() for l in stdout.strip().splitlines() if l.strip()]
+    return {
+        "ok": True,
+        "checked_at": _now(),
+        "data": lines,
+        "summary": f"최근 스케줄러 로그 {len(lines)}줄",
+        "error": "",
+    }
+
+
+async def queue_status() -> dict:
+    """Get document processing queue status from Document Server logs."""
+    cfg = HOSTS["gpu"]
+    cmd = "docker logs hyungi_document_server-fastapi-1 --tail 200 2>&1 | grep -iE 'queue_consumer|pending|processing|completed|stale|batch' | tail -20"
+
+    try:
+        if _is_local_host(cfg):
+            stdout, _ = await run_local(cmd)
+        else:
+            stdout, _ = await run_command(cfg, cmd)
+    except SSHError as e:
+        return {"ok": False, "checked_at": _now(), "error_type": e.error_type,
+                "error": str(e), "data": [], "summary": ""}
+
+    lines = [l.strip() for l in stdout.strip().splitlines() if l.strip()]
+    return {
+        "ok": True,
+        "checked_at": _now(),
+        "data": lines,
+        "summary": f"최근 큐 로그 {len(lines)}줄",
+        "error": "",
+    }

infra/core/verify.py

@@ -0,0 +1,73 @@
+"""Run predefined verify commands from infra_inventory.md."""
+
+from __future__ import annotations
+
+from datetime import datetime, timezone
+
+from ..config import HOSTS
+from .ssh import run_command, run_local, SSHError, _is_local_host
+
+
+def _now() -> str:
+    return datetime.now(timezone.utc).isoformat()
+
+
+# Predefined verify commands (from infra_inventory.md)
+VERIFY_COMMANDS = {
+    "gpu-snapshot": {
+        "host": "gpu",
+        "cmd": "docker compose ls 2>/dev/null; echo '---'; ollama list 2>/dev/null; echo '---'; pgrep -af 'Plex Media' | head -1 2>/dev/null",
+        "desc": "GPU 서버 전체 스냅샷 (docker, ollama, plex)",
+    },
+    "macmini-snapshot": {
+        "host": "macmini",
+        "cmd": "bash -lc 'launchctl list | grep com.user; curl -s localhost:8800/v1/models 2>/dev/null | python3 -m json.tool 2>/dev/null; ollama list 2>/dev/null'",
+        "desc": "Mac mini 전체 스냅샷 (launchd, MLX, ollama)",
+    },
+    "docserver-health": {
+        "host": "gpu",
+        "cmd": "curl -sf http://localhost:8000/health",
+        "desc": "Document Server 헬스체크",
+    },
+    "config-model-match": {
+        "host": "gpu",
+        "cmd": "grep -E 'model:' ~/Documents/code/hyungi_Document_Server/config.yaml 2>/dev/null; echo '---'; ollama list 2>/dev/null",
+        "desc": "config.yaml 모델과 실제 설치 모델 비교",
+    },
+}
+
+
+async def run_verify(check_name: str) -> dict:
+    """Run a predefined verify command."""
+    if check_name not in VERIFY_COMMANDS:
+        available = ", ".join(VERIFY_COMMANDS.keys())
+        return {
+            "ok": False, "checked_at": _now(),
+            "error_type": "parse_error",
+            "error": f"알 수 없는 체크: '{check_name}'. 사용 가능: {available}",
+            "data": [], "summary": "", "raw": "",
+        }
+
+    spec = VERIFY_COMMANDS[check_name]
+    cfg = HOSTS[spec["host"]]
+
+    try:
+        if _is_local_host(cfg):
+            stdout, _ = await run_local(spec["cmd"], timeout=15)
+        else:
+            stdout, _ = await run_command(cfg, spec["cmd"], timeout=15)
+    except SSHError as e:
+        return {
+            "ok": False, "checked_at": _now(),
+            "error_type": e.error_type, "error": str(e),
+            "data": [], "summary": spec["desc"], "raw": "",
+        }
+
+    return {
+        "ok": True,
+        "checked_at": _now(),
+        "data": stdout.strip().splitlines(),
+        "summary": spec["desc"],
+        "error": "",
+        "raw": stdout.strip(),
+    }

infra/mcp_server.py

@@ -11,6 +11,7 @@ All actual logic lives in src/core/.
 
 from __future__ import annotations
 
+import json
 from mcp.server.fastmcp import FastMCP
 
 from .core.docker import docker_status, docker_logs
@@ -18,6 +19,8 @@ from .core.health import service_health, VALID_SERVICES
 from .core.system import disk_usage
 from .core.network import tailscale_status
 from .core.models import ollama_models, mlx_models
+from .core.docserver import scheduler_status as _scheduler_status, queue_status as _queue_status
+from .core.verify import run_verify as _run_verify, VERIFY_COMMANDS
 
 mcp = FastMCP(
     "infra",
@@ -99,6 +102,31 @@ async def check_mlx_models() -> str:
     return result.model_dump_json(indent=2)
 
 
+@mcp.tool()
+async def check_scheduler_status() -> str:
+    """Document Server APScheduler 잡 상태. 최근 스케줄러 로그에서 추출."""
+    result = await _scheduler_status()
+    return json.dumps(result, ensure_ascii=False, indent=2)
+
+
+@mcp.tool()
+async def check_queue_status() -> str:
+    """Document Server 문서 처리 큐 현황. 최근 큐 로그에서 추출."""
+    result = await _queue_status()
+    return json.dumps(result, ensure_ascii=False, indent=2)
+
+
+@mcp.tool()
+async def check_verify(check_name: str) -> str:
+    """인프라 검증 명령 실행 (infra_inventory.md 기반).
+
+    Args:
+        check_name: 체크 이름 (gpu-snapshot | macmini-snapshot | docserver-health | config-model-match)
+    """
+    result = await _run_verify(check_name)
+    return json.dumps(result, ensure_ascii=False, indent=2)
+
+
 def main():
     mcp.run(transport="stdio")