fix: exclude prose colons from drainage URL boundaries

api/src/send-chain/drainage-authorization.spec.ts

@@ -30,6 +30,28 @@ const material = (id: string, url: string, channels: string[], auditStatus = 'ap
   })),
 });
 describe('drainage authorization', () => {
+  it.each(['详情：https://qa0915.example.com', '详情:https://qa0915.example.com'])(
+    'keeps prose separators outside explicit URLs: %s',
+    (content) => {
+      const text = 'https://qa0915.example.com';
+      const start = content.indexOf(text);
+      const targets = drainageTargets(content, [
+        {
+          ruleId: 'url',
+          ruleCode: 'URL',
+          ruleName: 'URL',
+          category: 'url',
+          text,
+          normalizedText: text,
+          start,
+          end: start + text.length,
+        },
+      ]);
+      expect(targets).toHaveLength(1);
+      expect(targets[0].text).toBe(text);
+      expect(materialMatches(targets[0], text)).toBe(true);
+    },
+  );
   test.each([
     'lisglo.cn',
     'sms.lisglo.cn',

api/src/send-chain/drainage-authorization.ts

@@ -81,7 +81,11 @@ export function drainageTargets(content: string, matches: DrainageDetectionMatch
     if (start < 0 || end <= 0) throw new ServiceUnavailableException('引流识别位置无效');
     // Extend the entire URL token, including suffix labels, userInfo and query.
     const token = /[a-z0-9:/?&=.%_+@#~!$*()[\]-]/i;
-    while (start > 0 && token.test(normalized.text[start - 1])) start--;
+    while (start > 0 && token.test(normalized.text[start - 1])) {
+      // A prose colon before an explicit scheme is a separator, including normalized Chinese colons.
+      if (normalized.text[start - 1] === ':' && /^https?:\/\//i.test(normalized.text.slice(start, end))) break;
+      start--;
+    }
     while (end < normalized.text.length && token.test(normalized.text[end])) end++;
     const text = normalized.text.slice(start, end).replace(/[.,;!]+$/, '');
     const value = drainageHost(text);