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/resource-write.ts

@@ -1,8 +1,10 @@
 import { ensureSchema, getRawDb } from "./mvp-db";
 import {
+  generateAccountTagsFromMcp,
   resolveCollectionMcpConfig,
   resolveProfileDetailsFromMcp,
   resolveXhsPublicAccountDetails,
+  type AccountTaggingEnv,
   type CollectionMcpBindings,
 } from "./mcp-collection-client";
 import { getRuntimeEnv } from "./runtime-env";
@@ -135,9 +137,8 @@ export async function enrichRows(
   rows: ResourceImportRow[],
   accounts: AccountRow[],
 ) {
-  const mcpConfig = resolveCollectionMcpConfig(
-    getRuntimeEnv() as unknown as CollectionMcpBindings,
-  );
+  const env = getRuntimeEnv() as unknown as CollectionMcpBindings;
+  const mcpConfig = resolveCollectionMcpConfig(env);
   const baselineRows = mergeExistingFields(rows, accounts);
   return mapConcurrent(baselineRows, 4, async (baseline) => {
     const row = baseline;
@@ -216,6 +217,89 @@ export async function enrichRows(
   });
 }
 
+type UntaggedAccountRow = {
+  id: string;
+  platform: string;
+  platform_uid: string;
+  profile_url: string;
+  nickname: string;
+};
+
+export async function tagUntaggedAccounts(
+  rows: Array<Pick<AnalyzedRow, "platform" | "platformUid">>,
+) {
+  const env = getRuntimeEnv() as unknown as CollectionMcpBindings &
+    AccountTaggingEnv;
+  const toolName = String(env.AI_TAGGING_TOOL ?? "").trim();
+  const refs = new Set(
+    rows
+      .filter((row) => row.platformUid)
+      .map((row) => `${row.platform}:${row.platformUid}`),
+  );
+  if (!toolName || refs.size === 0) return;
+
+  const db = getRawDb();
+  const uids = [...new Set([...refs].map((key) => key.slice(key.indexOf(":") + 1)))];
+  const candidates: UntaggedAccountRow[] = [];
+  for (let index = 0; index < uids.length; index += 100) {
+    const chunk = uids.slice(index, index + 100);
+    const placeholders = chunk.map(() => "?").join(", ");
+    const result = await db
+      .prepare(
+        `SELECT id, platform, platform_uid, profile_url, nickname
+         FROM accounts
+         WHERE (tags = '' OR tags IS NULL)
+           AND profile_url IS NOT NULL AND profile_url != ''
+           AND platform_uid IN (${placeholders})`,
+      )
+      .bind(...chunk)
+      .all<UntaggedAccountRow>();
+    candidates.push(...result.results);
+  }
+  const targets = candidates.filter((account) =>
+    refs.has(`${account.platform}:${account.platform_uid}`),
+  );
+  if (targets.length === 0) return;
+
+  const mcpConfig = resolveCollectionMcpConfig(env);
+  await mapConcurrent(targets, 4, async (account) => {
+    const details = await resolveProfileDetailsFromMcp(
+      account.profile_url,
+      account.platform === "抖音" ? "抖音" : "小红书",
+      mcpConfig,
+    ).catch(() => null);
+    if (
+      !details ||
+      (!details.bio &&
+        details.providerTags.length === 0 &&
+        details.recentNoteTitles.length === 0)
+    ) {
+      return;
+    }
+    const tags = await generateAccountTagsFromMcp(
+      {
+        platform: account.platform,
+        nickname: details.nickname?.trim() || account.nickname || "",
+        bio: details.bio,
+        gender: details.gender,
+        ipLocation: details.ipLocation ?? "",
+        followers: details.followers,
+        providerTags: details.providerTags,
+        recentNoteTitles: details.recentNoteTitles,
+      },
+      mcpConfig,
+      { AI_TAGGING_TOOL: toolName },
+    ).catch(() => []);
+    if (tags.length === 0) return;
+    await db
+      .prepare(
+        `UPDATE accounts SET tags = ? WHERE id = ? AND (tags = '' OR tags IS NULL)`,
+      )
+      .bind(tags.join(","), account.id)
+      .run();
+  });
+}
+
 export function analyzeRows(
   rows: ResourceImportRow[],
   accountRows: AccountRow[],