feat: NanoClaude 프로덕션 통합 — Docker, Caddy, aiosqlite 로깅

- docker-compose에 nanoclaude 서비스 추가 (포트 8100)
- Caddy /nano/* → nanoclaude 리버스 프록시 (SSE flush)
- aiosqlite 요청/응답 로깅 (request_logs 테이블)
- .env.example, CLAUDE.md 업데이트

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

.env.example

@@ -14,3 +14,7 @@ BACKENDS_CONFIG=/app/config/backends.json
 
 # DB path (in Docker)
 DB_PATH=/app/data/gateway.db
+
+# NanoClaude
+EXAONE_MODEL=exaone3.5:7.8b-instruct-q8_0
+NANOCLAUDE_API_KEY=

CLAUDE.md

@@ -9,13 +9,14 @@ GPU 서버(RTX 4070 Ti Super)에서 운영하는 중앙 AI 라우팅 서비스.
 |--------|----------|------|------|
 | Caddy | caddy/ | Caddy 2 | 80/443 |
 | hub-api | hub-api/ | FastAPI + aiosqlite | 8000 |
-| hub-web | hub-web/ | Vite + React + shadcn/ui | 3000 (Phase 2) |
+| hub-web | hub-web/ | Vite + React + shadcn/ui | 3000 |
+| NanoClaude | nanoclaude/ | FastAPI + aiosqlite | 8100 |
 
 ## 외부 연결
 
 - GPU Ollama: host.docker.internal:11434
 - 맥미니 Ollama: 100.115.153.119:11434
-- NanoClaude: 100.115.153.119:PORT (Phase 1.5)
+- NanoClaude: localhost:8100 (비동기 job 기반 AI Gateway)
 
 ## 개발
 
@@ -37,6 +38,11 @@ OpenAI 호환: `/v1/chat/completions`, `/v1/models`, `/v1/embeddings`
 인증: `/auth/login` → Cookie 또는 Bearer 토큰
 모니터링: `/health`, `/gpu`
 
+## NanoClaude API
+
+비동기 job 기반: `POST /nano/chat` → `{ job_id }`, `GET /nano/chat/{job_id}/stream` → SSE
+취소: `POST /nano/chat/{job_id}/cancel`
+
 ## 백엔드 설정
 
 `backends.json`에서 백엔드 추가/제거. 서비스 재시작 필요.

caddy/Caddyfile

@@ -16,6 +16,12 @@
 	handle /gpu {
 		reverse_proxy hub-api:8000
 	}
+	handle /nano/* {
+		uri strip_prefix /nano
+		reverse_proxy nanoclaude:8100 {
+			flush_interval -1
+		}
+	}
 	handle {
 		reverse_proxy hub-web:80
 	}

docker-compose.yml

@@ -12,6 +12,7 @@ services:
     depends_on:
       - hub-api
       - hub-web
+      - nanoclaude
     networks:
       - gateway-net
 
@@ -46,9 +47,31 @@ services:
     networks:
       - gateway-net
 
+  nanoclaude:
+    build: ./nanoclaude
+    container_name: gpu-nanoclaude
+    restart: unless-stopped
+    environment:
+      - EXAONE_BASE_URL=http://host.docker.internal:11434
+      - EXAONE_MODEL=${EXAONE_MODEL:-exaone3.5:7.8b-instruct-q8_0}
+      - DB_PATH=/app/data/nanoclaude.db
+      - API_KEY=${NANOCLAUDE_API_KEY:-}
+    volumes:
+      - nano_data:/app/data
+    extra_hosts:
+      - "host.docker.internal:host-gateway"
+    healthcheck:
+      test: ["CMD", "curl", "-f", "http://localhost:8100/health"]
+      interval: 15s
+      timeout: 5s
+      retries: 3
+    networks:
+      - gateway-net
+
 volumes:
   caddy_data:
   hub_data:
+  nano_data:
 
 networks:
   gateway-net:

nanoclaude/config.py

@@ -12,6 +12,9 @@ class Settings(BaseSettings):
     host: str = "0.0.0.0"
     port: int = 8100
 
+    # DB
+    db_path: str = "/app/data/nanoclaude.db"
+
     # Optional API key (empty = disabled)
     api_key: str = ""
 

nanoclaude/db/database.py

@@ -0,0 +1,47 @@
+"""aiosqlite DB — 요청/응답 로깅 및 메트릭."""
+
+import aiosqlite
+
+from config import settings
+
+SCHEMA = """
+CREATE TABLE IF NOT EXISTS request_logs (
+    id INTEGER PRIMARY KEY AUTOINCREMENT,
+    job_id TEXT NOT NULL,
+    message TEXT NOT NULL,
+    status TEXT NOT NULL DEFAULT 'queued',
+    model TEXT NOT NULL,
+    response_chars INTEGER DEFAULT 0,
+    latency_ms REAL DEFAULT 0,
+    created_at REAL NOT NULL,
+    completed_at REAL
+);
+
+CREATE INDEX IF NOT EXISTS idx_logs_job ON request_logs(job_id);
+CREATE INDEX IF NOT EXISTS idx_logs_created ON request_logs(created_at);
+"""
+
+
+async def init_db():
+    async with aiosqlite.connect(settings.db_path) as db:
+        await db.execute("PRAGMA journal_mode=WAL")
+        await db.executescript(SCHEMA)
+        await db.commit()
+
+
+async def log_request(job_id: str, message: str, model: str, created_at: float):
+    async with aiosqlite.connect(settings.db_path) as db:
+        await db.execute(
+            "INSERT INTO request_logs (job_id, message, model, created_at) VALUES (?, ?, ?, ?)",
+            (job_id, message, model, created_at),
+        )
+        await db.commit()
+
+
+async def log_completion(job_id: str, status: str, response_chars: int, latency_ms: float, completed_at: float):
+    async with aiosqlite.connect(settings.db_path) as db:
+        await db.execute(
+            "UPDATE request_logs SET status=?, response_chars=?, latency_ms=?, completed_at=? WHERE job_id=?",
+            (status, response_chars, latency_ms, completed_at, job_id),
+        )
+        await db.commit()

nanoclaude/main.py

@@ -3,12 +3,14 @@
 from __future__ import annotations
 
 import logging
+from contextlib import asynccontextmanager
 
 from fastapi import FastAPI, Request
 from fastapi.middleware.cors import CORSMiddleware
 from fastapi.responses import JSONResponse
 
 from config import settings
+from db.database import init_db
 from routers import chat
 
 logging.basicConfig(
@@ -16,10 +18,17 @@ logging.basicConfig(
     format="%(asctime)s %(levelname)s %(name)s — %(message)s",
 )
 
+@asynccontextmanager
+async def lifespan(app: FastAPI):
+    await init_db()
+    yield
+
+
 app = FastAPI(
     title="NanoClaude",
     version="0.1.0",
     description="비동기 job 기반 AI Gateway — Phase 1",
+    lifespan=lifespan,
 )
 
 app.add_middleware(

nanoclaude/requirements.txt

@@ -2,3 +2,4 @@ fastapi==0.115.0
 uvicorn[standard]==0.30.0
 httpx==0.27.0
 pydantic-settings==2.5.0
+aiosqlite==0.20.0

nanoclaude/services/worker.py

@@ -4,7 +4,10 @@ from __future__ import annotations
 
 import asyncio
 import logging
+from time import time
 
+from config import settings
+from db.database import log_completion, log_request
 from models.schemas import JobStatus
 from services.exaone_adapter import stream_chat
 from services.job_manager import Job, job_manager
@@ -18,6 +21,14 @@ HEARTBEAT_INTERVAL = 4.0
 
 async def run(job: Job) -> None:
     """EXAONE 호출 → SSE 이벤트 발행."""
+    start_time = time()
+
+    # DB 로깅: 요청 기록
+    try:
+        await log_request(job.id, job.message, settings.exaone_model, job.created_at)
+    except Exception:
+        logger.warning("Failed to log request for job %s", job.id, exc_info=True)
+
     try:
         # --- ACK ---
         await state_stream.push(job.id, "ack", {"message": "요청을 확인했습니다. 분석을 시작합니다."})
@@ -50,9 +61,19 @@ async def run(job: Job) -> None:
         if not collected:
             job_manager.set_status(job.id, JobStatus.failed)
             await state_stream.push(job.id, "error", {"message": "EXAONE으로부터 응답을 받지 못했습니다."})
+            status = "failed"
         else:
             job_manager.set_status(job.id, JobStatus.completed)
             await state_stream.push(job.id, "done", {"message": "완료"})
+            status = "completed"
+
+        # DB 로깅: 완료 기록
+        latency_ms = (time() - start_time) * 1000
+        response_text = "".join(collected)
+        try:
+            await log_completion(job.id, status, len(response_text), latency_ms, time())
+        except Exception:
+            logger.warning("Failed to log completion for job %s", job.id, exc_info=True)
 
     except asyncio.CancelledError:
         job_manager.set_status(job.id, JobStatus.cancelled)