feat: KOC 资源导入后自动爬取主页资料并调用 LLM 打标签

- 新增 structured_text_process 打标客户端（文本构造、标签解析清洗、MCP 调用）
- 写库后后台打标（tagUntaggedAccounts），单条新增 / 批量导入 / 每日补全三个入口统一
- 标签只补空，UPDATE 带 tags 为空守卫，不覆盖已有标签
- 新增 AI_TAGGING_TOOL 配置项（留空关闭），更新配置示例与文档
- 新增 16 个打标测试用例，全量 131 个测试通过

lib/account-enrichment-service.ts

@@ -1,9 +1,12 @@
 import {
+  generateAccountTagsFromMcp,
   resolveXhsPublicAccountDetails,
   resolveAccountProfileFromMcp,
   resolveProfileDetailsFromMcp,
+  type AccountTaggingEnv,
   type CollectionMcpConfig,
 } from "./mcp-collection-client";
+import { getRuntimeEnv } from "./runtime-env";
 import { hashText } from "./mvp-db";
 import type { DatabaseClient } from "./database";
 
@@ -260,6 +263,7 @@ export async function backfillAccountProfiles(
   mcpConfig: CollectionMcpConfig,
   limit = 10,
 ) {
+  const taggingEnv: AccountTaggingEnv = getRuntimeEnv();
   const rows = await db
     .prepare(
       `SELECT
@@ -334,6 +338,7 @@ export async function backfillAccountProfiles(
     const isDemoAccount = row.platform_uid?.startsWith("xhs-");
     let attemptedThisRow = false;
     let publicProfileUpdated = false;
+    let tagsUpdated = false;
     if (
       row.platform === "小红书" &&
       !isDemoAccount &&
@@ -341,7 +346,8 @@ export async function backfillAccountProfiles(
       (!row.public_account_id ||
         Number(row.followers ?? 0) === 0 ||
         !row.gender ||
-        !row.bio)
+        !row.bio ||
+        !row.tags?.trim())
     ) {
       attempted += 1;
       attemptedThisRow = true;
@@ -402,6 +408,39 @@ export async function backfillAccountProfiles(
           .run();
         publicProfileUpdated = true;
       }
+      if (
+        row.account_id &&
+        !row.tags?.trim() &&
+        (details.bio ||
+          details.providerTags.length > 0 ||
+          details.recentNoteTitles.length > 0)
+      ) {
+        const tags = await generateAccountTagsFromMcp(
+          {
+            platform: "小红书",
+            nickname: row.nickname || details.nickname || "",
+            bio: details.bio,
+            gender: details.gender,
+            ipLocation: details.ipLocation ?? "",
+            followers: details.followers,
+            providerTags: details.providerTags,
+            recentNoteTitles: details.recentNoteTitles,
+          },
+          mcpConfig,
+          taggingEnv,
+        ).catch(() => []);
+        if (tags.length > 0) {
+          await db
+            .prepare(
+              `UPDATE accounts
+               SET tags = CASE WHEN ? != '' THEN ? ELSE tags END
+               WHERE id = ?`,
+            )
+            .bind(tags.join(","), tags.join(","), row.account_id)
+            .run();
+          tagsUpdated = true;
+        }
+      }
       if (
         (details.redId || row.public_account_id) &&
         Number(details.followers ?? row.followers ?? 0) > 0
@@ -410,7 +449,7 @@ export async function backfillAccountProfiles(
         continue;
       }
       if (attempted >= Math.max(1, Math.min(25, limit))) {
-        if (publicProfileUpdated) updated += 1;
+        if (publicProfileUpdated || tagsUpdated) updated += 1;
         continue;
       }
     }
@@ -430,7 +469,7 @@ export async function backfillAccountProfiles(
       row.platform_uid?.startsWith("pending-") ||
       Boolean(noteId && row.platform_uid === noteId));
     if (!needsEnrichment || !row.publish_url) {
-      if (publicProfileUpdated) updated += 1;
+      if (publicProfileUpdated || tagsUpdated) updated += 1;
       continue;
     }
 
@@ -443,9 +482,9 @@ export async function backfillAccountProfiles(
         row.nickname || "待识别账号",
         mcpConfig,
       );
-      if (result.updated || publicProfileUpdated) updated += 1;
+      if (result.updated || publicProfileUpdated || tagsUpdated) updated += 1;
     } catch {
-      if (publicProfileUpdated) updated += 1;
+      if (publicProfileUpdated || tagsUpdated) updated += 1;
       else failed += 1;
     }
   }