perf: multi-stage Docker builds, .dockerignore, singleton StarRocks connection

backend/.dockerignore

+# ═══ Git & IDE ═══
+.git
+.gitignore
+.vscode
+.idea
+*.swp
+*.swo
+
+# ═══ Python cache ═══
 __pycache__
 *.pyc
-.env
+*.pyo
+*.egg-info
+.pytest_cache
+.mypy_cache
+.ruff_cache
+
+# ═══ Virtual environments ═══
 .venv
 venv
-.git
-.gitignore
+env
+
+# ═══ Docker ═══
+Dockerfile*
+docker-compose*
 .dockerignore
-logs
-data
+
+# ═══ Dev & test files ═══
+tests/
+*.test.py
+setup_readonly_user.py
+
+# ═══ Docs & artifacts ═══
+*.md
+LICENSE
+
+# ═══ OS files ═══
+.DS_Store
+Thumbs.db
+
+# ═══ Logs ═══
+*.log
+logs/
+
+# ═══ Local env backup ═══
+.env.local
+.env.backup
+.env.example

backend/Dockerfile.dev

 # ============================================================
-# DOCKERFILE.DEV - Local Development (Hot Reload + Cache)
+# DOCKERFILE.DEV — Multi-stage Dev Build (Hot Reload)
 # ============================================================
-# Sử dụng Python 3.11 Slim để tối ưu dung lượng
+# Stage 1: Builder — install dependencies
+# Stage 2: Runtime — dev server with reload
+# ============================================================
+
+# ──── Stage 1: Builder ────
+FROM python:3.11-slim AS builder
+
+WORKDIR /build
+
+RUN apt-get update && apt-get install -y --no-install-recommends \
+    gcc libpq-dev \
+    && rm -rf /var/lib/apt/lists/*
+
+COPY requirements.txt .
+
+RUN pip install --no-cache-dir --user -r requirements.txt \
+    && pip install --no-cache-dir --user watchdog[watchmedo]
+
+
+# ──── Stage 2: Runtime ────
 FROM python:3.11-slim
 
-# Thiết lập thư mục làm việc
 WORKDIR /app
 
-# Thiết lập biến môi trường
+RUN apt-get update && apt-get install -y --no-install-recommends \
+    libpq5 curl \
+    && rm -rf /var/lib/apt/lists/* \
+    && apt-get purge -y --auto-remove
+
 ENV PYTHONUNBUFFERED=1
 ENV PYTHONDONTWRITEBYTECODE=1
 ENV ENV=development
 
-# Copy requirements.txt trước để tận dụng Docker cache
-COPY requirements.txt .
-
-# Cài đặt thư viện Python (Docker layer cache)
-RUN pip install -r requirements.txt && pip install watchdog[watchmedo]
+# Copy installed packages from builder
+COPY --from=builder /root/.local /root/.local
+ENV PATH=/root/.local/bin:$PATH
 
-# Copy toàn bộ source code vào image
+# Source code mounted via volume in docker-compose (not COPY)
 COPY . .
 
-# Expose port 5000
 EXPOSE 5000
 
-# Health check (optional)
 HEALTHCHECK --interval=10s --timeout=5s --start-period=5s --retries=2 \
-    CMD python -c "import requests; requests.get('http://localhost:5000/docs')" || exit 1
+    CMD curl -f http://localhost:5000/docs || exit 1
 
 CMD ["gunicorn", "--workers", "1", "--worker-class", "uvicorn.workers.UvicornWorker", "--bind", "0.0.0.0:5000", "--timeout", "120", "--reload", "server:app"]

backend/Dockerfile.prod

+# ============================================================
+# DOCKERFILE.PROD — Multi-stage Production Build
+# ============================================================
+# Stage 1: Build — install dependencies in isolated layer
+# Stage 2: Runtime — minimal image with only what's needed
+# ============================================================
 
-FROM python:3.11-slim
+# ──── Stage 1: Builder ────
+FROM python:3.11-slim AS builder
+
+WORKDIR /build
+
+# Install only build-time system deps (if any wheels need compiling)
+RUN apt-get update && apt-get install -y --no-install-recommends \
+    gcc libpq-dev \
+    && rm -rf /var/lib/apt/lists/*
+
+COPY requirements.txt .
 
+# Install to user site-packages (no root pollution)
+RUN pip install --no-cache-dir --user -r requirements.txt
+
+
+# ──── Stage 2: Runtime ────
+FROM python:3.11-slim
 
 WORKDIR /app
 
+# Minimal runtime deps only
+RUN apt-get update && apt-get install -y --no-install-recommends \
+    libpq5 curl \
+    && rm -rf /var/lib/apt/lists/* \
+    && apt-get purge -y --auto-remove
 
 ENV PYTHONUNBUFFERED=1
 ENV PYTHONDONTWRITEBYTECODE=1
-ENV ENV=development
+ENV ENV=production
 
-COPY requirements.txt .
-RUN pip install -r requirements.txt
+# Copy only installed packages from builder (no pip cache, no gcc)
+COPY --from=builder /root/.local /root/.local
+ENV PATH=/root/.local/bin:$PATH
 
+# Copy application source
 COPY . .
 
 EXPOSE 5000
@@ -19,4 +48,13 @@ EXPOSE 5000
 ENV WORKERS=1
 ENV TIMEOUT=60
 
-CMD gunicorn server:app --workers $WORKERS --worker-class uvicorn.workers.UvicornWorker --bind 0.0.0.0:5000 --timeout $TIMEOUT
+HEALTHCHECK --interval=30s --timeout=5s --start-period=10s --retries=3 \
+    CMD curl -f http://localhost:5000/docs || exit 1
+
+CMD gunicorn server:app \
+    --workers $WORKERS \
+    --worker-class uvicorn.workers.UvicornWorker \
+    --bind 0.0.0.0:5000 \
+    --timeout $TIMEOUT \
+    --access-logfile - \
+    --error-logfile -

backend/api/product_route.py

@@ -9,7 +9,7 @@ from typing import Optional
 from fastapi import APIRouter, Query
 from fastapi.responses import JSONResponse
 
-from common.starrocks_connection import StarRocksConnection
+from common.starrocks_connection import get_db_connection
 from agent.tools.product_mapping import resolve_color, resolve_product_name, SYNONYM_TO_DB
 
 logger = logging.getLogger(__name__)
@@ -21,7 +21,7 @@ TABLE_NAME = "shared_source.magento_product_dimension_with_text_embedding"
 @router.get("/api/products/overview", summary="Product catalog KPI overview")
 async def products_overview():
     """Return aggregate KPIs for the product catalog."""
-    db = StarRocksConnection()
+    db = get_db_connection()
     try:
         rows = await db.execute_query_async(f"""
             SELECT
@@ -104,7 +104,7 @@ async def products_list(
     offset: int = Query(0, ge=0),
 ):
     """Return paginated product listing grouped by internal_ref_code."""
-    db = StarRocksConnection()
+    db = get_db_connection()
 
     # Build WHERE clauses
     clauses = []
@@ -234,7 +234,7 @@ async def products_colors(
     code: str = Query(..., description="internal_ref_code"),
 ):
     """Return all color variants for a given internal_ref_code."""
-    db = StarRocksConnection()
+    db = get_db_connection()
     try:
         variants = await db.execute_query_async(
             f"""
@@ -266,7 +266,7 @@ async def products_colors(
 @router.get("/api/products/filters", summary="Available filter options")
 async def products_filters():
     """Return all available filter values for color, product_line, age, season."""
-    db = StarRocksConnection()
+    db = get_db_connection()
     try:
         colors = await db.execute_query_async(f"""
             SELECT master_color, COUNT(DISTINCT internal_ref_code) AS cnt