update backend

2026-09-20 01:06:13 +08:00
parent f7ae917603
commit b99dc3dca1
115 changed files with 24732 additions and 10842 deletions
--- a/scripts/diag_brand/main.go
+++ b/scripts/diag_brand/main.go
@ -17,7 +17,7 @@ func main() {

 	// 1) brand 3100 的走秀档案数（only_with_articles 子查询依赖这个）
 	var artCount int64
-	db.Raw(`SELECT COUNT(*) FROM brand_runway WHERE is_deleted=0 AND brand_id=3100`).Scan(&artCount)
+	db.Raw(`SELECT COUNT(*) FROM brand_runways WHERE is_deleted=0 AND brand_id=3100`).Scan(&artCount)
 	fmt.Printf("brand 3100 走秀档案数 = %d\n", artCount)

 	// 2) 复现后端完整搜索：only_with_articles=1 + keyword=猿人头
@ -30,9 +30,9 @@ func main() {
 	}
 	var rows []Row
 	db.Raw(`
-		SELECT id, name, name_en, name_cn FROM brand
+		SELECT id, name, name_en, name_cn FROM brands
 		WHERE is_deleted=0
-		  AND id IN (SELECT DISTINCT brand_id FROM brand_runway WHERE is_deleted=0)
+		  AND id IN (SELECT DISTINCT brand_id FROM brand_runways WHERE is_deleted=0)
 		  AND (name_en LIKE ? OR name_cn LIKE ?)`,
 		kw, kw).Scan(&rows)
 	fmt.Printf("\n[搜索 keyword=猿人头, only_with_articles=1] 命中 %d 条:\n", len(rows))
@ -43,7 +43,7 @@ func main() {
 	// 3) 对比：only_with_articles=0（不过滤档案）时的命中
 	var rows2 []Row
 	db.Raw(`
-		SELECT id, name, name_en, name_cn FROM brand
+		SELECT id, name, name_en, name_cn FROM brands
 		WHERE is_deleted=0
 		  AND (name_en LIKE ? OR name_cn LIKE ?)`,
 		kw, kw).Scan(&rows2)
--- a/scripts/diag_hot/main.go
+++ b/scripts/diag_hot/main.go
@ -26,7 +26,7 @@ func main() {
 	var rows []Row
 	db.Raw(`
 		SELECT brand_id, COALESCE(SUM(image_count),0) AS images, COUNT(*) AS shows
-		FROM brand_runway
+		FROM brand_runways
 		WHERE is_deleted = 0
 		GROUP BY brand_id
 		ORDER BY images DESC
@ -36,24 +36,24 @@ func main() {
 	fmt.Println("== FeaturedIDs(images, 30) 原始排名 ==")
 	for i, r := range rows {
 		var name string
-		db.Raw(`SELECT name_en FROM brand WHERE id = ?`, r.BrandID).Scan(&name)
+		db.Raw(`SELECT name_en FROM brands WHERE id = ?`, r.BrandID).Scan(&name)
 		fmt.Printf("%2d. brand_id=%-6d images=%-6d shows=%-4d %s\n", i+1, r.BrandID, r.Images, r.Shows, name)
 	}

 	// 2) image_count 分布：有多少品牌有图、多少归零
 	var zeroBrands, hasImgBrands int64
 	var totalRows, zeroRows int64
-	db.Raw(`SELECT COUNT(*) FROM (SELECT brand_id, SUM(image_count) s FROM brand_runway WHERE is_deleted=0 GROUP BY brand_id) t WHERE s = 0`).Scan(&zeroBrands)
-	db.Raw(`SELECT COUNT(*) FROM (SELECT brand_id, SUM(image_count) s FROM brand_runway WHERE is_deleted=0 GROUP BY brand_id) t WHERE s > 0`).Scan(&hasImgBrands)
-	db.Raw(`SELECT COUNT(*) FROM brand_runway WHERE is_deleted=0`).Scan(&totalRows)
-	db.Raw(`SELECT COUNT(*) FROM brand_runway WHERE is_deleted=0 AND image_count=0`).Scan(&zeroRows)
+	db.Raw(`SELECT COUNT(*) FROM (SELECT brand_id, SUM(image_count) s FROM brand_runways WHERE is_deleted=0 GROUP BY brand_id) t WHERE s = 0`).Scan(&zeroBrands)
+	db.Raw(`SELECT COUNT(*) FROM (SELECT brand_id, SUM(image_count) s FROM brand_runways WHERE is_deleted=0 GROUP BY brand_id) t WHERE s > 0`).Scan(&hasImgBrands)
+	db.Raw(`SELECT COUNT(*) FROM brand_runways WHERE is_deleted=0`).Scan(&totalRows)
+	db.Raw(`SELECT COUNT(*) FROM brand_runways WHERE is_deleted=0 AND image_count=0`).Scan(&zeroRows)
 	fmt.Printf("\n== image_count 分布 ==\n有图品牌=%d 零图品牌=%d | 走秀总行=%d 其中 image_count=0 的行=%d\n", hasImgBrands, zeroBrands, totalRows, zeroRows)

 	// 3) 对比口径：按走秀数(shows) DESC 前 10
 	var byShows []Row
 	db.Raw(`
 		SELECT brand_id, COALESCE(SUM(image_count),0) AS images, COUNT(*) AS shows
-		FROM brand_runway
+		FROM brand_runways
 		WHERE is_deleted = 0
 		GROUP BY brand_id
 		ORDER BY shows DESC
@ -62,7 +62,7 @@ func main() {
 	fmt.Println("\n== 若按 shows DESC 前 10 ==")
 	for i, r := range byShows {
 		var name string
-		db.Raw(`SELECT name_en FROM brand WHERE id = ?`, r.BrandID).Scan(&name)
+		db.Raw(`SELECT name_en FROM brands WHERE id = ?`, r.BrandID).Scan(&name)
 		fmt.Printf("%2d. brand_id=%-6d images=%-6d shows=%-4d %s\n", i+1, r.BrandID, r.Images, r.Shows, name)
 	}
 }
--- a/scripts/pgvector/initdb/01-extensions.sql
+++ b/scripts/pgvector/initdb/01-extensions.sql
@ -0,0 +1,3 @@
+-- 首次初始化（数据卷为空时由 postgres 官方 entrypoint 自动执行）即启用 pgvector 扩展。
+-- 用 IF NOT EXISTS 保证重复执行幂等。
+CREATE EXTENSION IF NOT EXISTS vector;
--- a/scripts/pgvector/initdb/03-migrate-phash-bigint-to-vector.sql
+++ b/scripts/pgvector/initdb/03-migrate-phash-bigint-to-vector.sql
@ -0,0 +1,49 @@
+-- 03-migrate-phash-bigint-to-vector.sql
+-- 兼容旧库：phash 早期以 bigint 存储 64-bit dHash 指纹（uint64），现模型改为 vector(64)。
+-- pgvector 不支持 bigint -> vector 直接转换，需逐位拆成 0/1 向量（位 i -> 第 i 维，
+-- 与 Go 侧 phash.ToVectorBits 完全一致，保证旧数据与新写入向量在同一语义空间可比）。
+--
+-- 用法（任选其一）：
+--   1) 后端启动时由 AutoMigrate 自动执行（已内置，幂等，无需手动）；
+--   2) 手动执行：  psql "postgres://fashion:fashion_dev_2026@localhost:5432/fashion" -f 03-migrate-phash-bigint-to-vector.sql
+--
+-- 已为 vector/text 类型的列自动跳过；可重复执行。
+
+DO $$
+DECLARE
+  t  text;
+  dt text;
+BEGIN
+  FOR t IN
+    SELECT tablename FROM pg_tables
+    WHERE schemaname = 'public'
+      AND tablename IN ('brand_runway_images','brand_runway_draft_images','street_snap_images','street_snap_draft_images')
+  LOOP
+    SELECT data_type INTO dt
+    FROM information_schema.columns
+    WHERE table_schema = 'public' AND table_name = t AND column_name = 'phash';
+
+    IF dt = 'bigint' THEN
+      -- bigint 即 64-bit dHash 指纹；pgvector 不支持 bigint->vector 直接转换，且 ALTER ... USING 的
+      -- transform 表达式不允许子查询(0A000)。改用纯表达式：bigint::bit(64) 取 64 位模式(MSB 在左)，
+      -- reverse 成 LSB 在左(与 phash.ToVectorBits 维度顺序一致：第 i 维 = bit i)，插逗号后 text::vector(64)。
+      -- NULL 自然透传。
+      EXECUTE format($e$
+        ALTER TABLE %I ALTER COLUMN phash TYPE vector(64)
+        USING (
+          ('[' ||
+           rtrim(regexp_replace(reverse(phash::bit(64)::text), '(.)', '\1,', 'g'), ',') ||
+           ']')::vector(64)
+        )
+      $e$, t);
+      RAISE NOTICE 'migrated phash (bigint -> vector) on %', t;
+    ELSIF dt IN ('character varying','text') THEN
+      EXECUTE format($e$
+        ALTER TABLE %I ALTER COLUMN phash TYPE vector(64) USING phash::vector(64)
+      $e$, t);
+      RAISE NOTICE 'migrated phash (text -> vector) on %', t;
+    ELSE
+      RAISE NOTICE 'phash on % already type %, skipped', t, dt;
+    END IF;
+  END LOOP;
+END $$;
--- a/scripts/setup_pg_wsl.sh
+++ b/scripts/setup_pg_wsl.sh
@ -0,0 +1,164 @@
+#!/usr/bin/env bash
+#
+# setup_pg_wsl.sh — WSL (Ubuntu/Debian) 下一键装 PostgreSQL + pgvector 开发环境
+#
+# 用法（在 WSL 终端里，不要用 Windows 的 CMD/PowerShell 跑）:
+#   sudo bash setup_pg_wsl.sh
+#
+# 三态幂等：
+#   1) 未安装        -> 安装 + 启动
+#   2) 已安装未运行  -> 仅启动
+#   3) 已安装且运行  -> 完全不动（不重装 / 不重启；仅幂等确保库/用户/扩展存在）
+#
+# 仅用于本地开发 / 迁移验证，生产环境请改强密码并单独评审配置。
+#
+set -euo pipefail
+
+# ============================ 配置（按需修改）============================
+PG_VERSION=16                 # PostgreSQL 大版本，pgvector HNSW 在 13+ 均可用
+DB_NAME=fashion               # 业务库名（替换成你的项目库）
+DB_USER=fashion               # 应用账号
+DB_PASS="fashion_dev_2026"    # 仅本地开发；生产务必换成强密码
+# =========================================================================
+
+log() { echo -e "\033[32m==>\033[0m $*"; }
+
+# 需要 root
+if [ "$(id -u)" -ne 0 ]; then
+  echo "请以 root 运行： sudo bash $0"; exit 1
+fi
+
+# 发行版探测（仅提示，不强制）
+if [ -r /etc/os-release ]; then
+  . /etc/os-release
+  log "检测到发行版: ${PRETTY_NAME:-unknown}"
+  case "${ID:-}" in
+    ubuntu|debian) ;;
+    *) echo "⚠️  本脚本面向 Ubuntu/Debian WSL，其他发行版请自行调整 apt 命令";;
+  esac
+fi
+
+# PG 是否在运行（用超级用户连 postgres 库试探，对活实例零副作用）
+pg_is_running() { sudo -u postgres psql -d postgres -tAc "SELECT 1" >/dev/null 2>&1; }
+
+# 1) 已装则跳过安装（不重装）
+if command -v psql >/dev/null 2>&1; then
+  log "PostgreSQL 已安装 ($(psql --version 2>/dev/null || echo unknown))，跳过安装"
+else
+  log "安装依赖并添加 PostgreSQL 官方 apt 源"
+  # 关键：先清掉旧的 pgdg 源与坏 key，否则下面第一次 apt-get update 会因旧 .asc key
+  # 报 "not signed" 返回非零，被 set -e 直接退出，导致修 key 步骤永远没机会执行。
+  rm -f /etc/apt/sources.list.d/pgdg.list
+  rm -f /usr/share/postgresql-common/pgdg/apt.postgresql.org.asc
+  apt-get update -y
+  apt-get install -y curl ca-certificates gnupg lsb-release
+  # 加 pgdg 源（官方标准：armored key -> dearmor 成 .gpg，signed-by 指向 .gpg）
+  install -d /usr/share/postgresql-common/pgdg
+  curl -fsSL https://www.postgresql.org/media/keys/ACCC4CF8.asc -o /tmp/pgdg-key.asc
+  gpg --dearmor -o /usr/share/postgresql-common/pgdg/apt.postgresql.org.gpg /tmp/pgdg-key.asc
+  chmod 644 /usr/share/postgresql-common/pgdg/apt.postgresql.org.gpg
+  echo "deb [signed-by=/usr/share/postgresql-common/pgdg/apt.postgresql.org.gpg] https://apt.postgresql.org/pub/repos/apt $(lsb_release -cs)-pgdg main" \
+    > /etc/apt/sources.list.d/pgdg.list
+  apt-get update -y
+  apt-get install -y "postgresql-$PG_VERSION" "postgresql-client-$PG_VERSION"
+fi
+
+# 2) pgvector（已可用则跳过，不重装）
+if sudo -u postgres psql -d postgres -tAc "SELECT 1 FROM pg_available_extensions WHERE name='vector'" 2>/dev/null | grep -q 1; then
+  log "pgvector 已可用，跳过"
+else
+  log "安装 pgvector 扩展"
+  if apt-get install -y "postgresql-$PG_VERSION-pgvector"; then
+    :
+  else
+    echo "apt 无 pgvector 包，改用源码编译..."
+    apt-get install -y "build-essential" "postgresql-server-dev-$PG_VERSION" git
+    tmp=$(mktemp -d)
+    git clone --depth 1 https://github.com/pgvector/pgvector.git "$tmp/pgvector"
+    make -C "$tmp/pgvector"
+    make -C "$tmp/pgvector" install
+    rm -rf "$tmp"
+  fi
+fi
+
+# 3) 仅在未运行时启动；已在运行则完全不动
+if pg_is_running; then
+  log "PostgreSQL 已在运行，跳过启动 / 重启"
+else
+  log "启动 PostgreSQL"
+  pg_ctlcluster "$PG_VERSION" main start 2>/dev/null \
+    || service postgresql start 2>/dev/null \
+    || systemctl start postgresql
+  # 等实例就绪（最多 15s）
+  for i in $(seq 1 15); do
+    if pg_is_running; then break; fi
+    sleep 1
+  done
+  if ! pg_is_running; then
+    echo "❌ PostgreSQL 启动失败，请查日志： journalctl -u postgresql 或 /var/log/postgresql/"; exit 1
+  fi
+fi
+
+# 4) 幂等确保库/用户/扩展（对运行中实例无副作用，不重启）
+log "确保用户、数据库与 pgvector 扩展存在（幂等）"
+sudo -u postgres psql -c "ALTER USER postgres WITH PASSWORD '$DB_PASS';"
+sudo -u postgres psql -c "CREATE ROLE $DB_USER LOGIN PASSWORD '$DB_PASS';" 2>/dev/null \
+  || echo "  角色 $DB_USER 已存在，跳过"
+sudo -u postgres psql -tc "SELECT 1 FROM pg_database WHERE datname='$DB_NAME'" | grep -q 1 \
+  || sudo -u postgres createdb -O "$DB_USER" "$DB_NAME"
+sudo -u postgres psql -d "$DB_NAME" -c "CREATE EXTENSION IF NOT EXISTS vector;"
+
+# 5) 本地访问配置：仅当 listen 还不是 localhost 才改；运行中走在线 reload，绝不重启
+PGCONF="/etc/postgresql/$PG_VERSION/main/postgresql.conf"
+if [ -f "$PGCONF" ]; then
+  if grep -q "^listen_addresses = 'localhost'" "$PGCONF"; then
+    : # 已配置，无需改动
+  else
+    log "调整 listen_addresses=localhost（仅本地）"
+    sed -i "s/^#*listen_addresses = .*/listen_addresses = 'localhost'/" "$PGCONF"
+    if pg_is_running; then
+      pg_ctlcluster "$PG_VERSION" main reload 2>/dev/null \
+        || sudo -u postgres psql -d postgres -c "SELECT pg_reload_conf();" >/dev/null 2>&1 \
+        || echo "  ⚠️  reload 失败，配置将在下次重启生效"
+    else
+      log "（实例未运行，配置将在下次启动时生效）"
+    fi
+  fi
+fi
+
+# 6) 验证
+log "验证安装"
+sudo -u postgres psql -d "$DB_NAME" \
+  -c "SELECT version();" \
+  -c "SELECT extname, extversion FROM pg_extension WHERE extname='vector';"
+
+cat <<EOF
+
+✅ PostgreSQL $PG_VERSION + pgvector 就绪（WSL）
+   连接串 : postgres://$DB_USER:$DB_PASS@localhost:5432/$DB_NAME
+   命令行 : psql -h localhost -U $DB_USER -d $DB_NAME
+   数据目录: /var/lib/postgresql/$PG_VERSION/main
+
+⚠️  WSL 注意事项
+   - WSL 重启后 PG 不会自动起，手动： sudo service postgresql start
+   - 想开机自启：在 Windows 侧建 .bat 执行  wsl -u root service postgresql start
+     或在 /etc/wsl.conf 加  [boot] command="service postgresql start"
+   - 本脚本密码为弱密码，仅本地开发；生产请更换并 review pg_hba.conf
+
+（可选）CLIP / DINOv2 embedding 表示例：
+   CREATE TABLE image_embedding (
+     id         BIGSERIAL PRIMARY KEY,
+     image_uid  VARCHAR(32) NOT NULL,     -- 对应你现有的 hashid uid
+     kind       SMALLINT     NOT NULL,    -- 0=runway 1=street
+     embedding  vector(512),              -- 512 维特征（CLIP ViT-B/32 或 DINOv2）
+     created_at INTEGER     NOT NULL DEFAULT 0
+   );
+   CREATE INDEX image_embedding_vec_idx
+     ON image_embedding USING hnsw (embedding vector_cosine_ops);
+   -- 去重查询：找最近邻
+   SELECT image_uid, 1 - (embedding <=> '[...]'::vector) AS sim
+   FROM image_embedding
+   WHERE kind = 0
+   ORDER BY embedding <=> '[...]'::vector
+   LIMIT 20;
+EOF