fix(bridge): convert <br> to newline for Telegram HTML mode (#23)

The Telegram Bot API's HTML parse mode does not support the <br> tag
and rejects it with "400 Bad Request: can't parse entities: Unsupported
start tag 'br'". Previously, markdownToHtml kept <br> in its whitelist,
so any CommonMark hard break (trailing two spaces or backslash) produced
by markdown-it would be forwarded to Telegram and crash the bridge with
an uncaught GrammyError.

Fix by replacing <br> with a literal newline before the whitelist strip,
and removing br from the whitelist comment/regex.

Repro: send any message whose markdown contains a hard line break, e.g.
"line one  \nline two" — the bridge process exits with an unhandled
GrammyError in MessageRenderer.flushCallback.

Adds two regression tests covering both CommonMark hard break forms.

Co-authored-by: shu zf <shuzf@aicoda.tech>

Author: Xiaofuziod

bridge/src/__tests__/markdown.test.ts

@@ -21,6 +21,24 @@ describe('Telegram rendering', () => {
     // Telegram doesn't support <h1>, should be plain text or bold
     expect(result).not.toContain('<h1>');
   });
+
+  it('converts <br> from markdown hard break to newline', () => {
+    // CommonMark hard break: two trailing spaces before newline -> <br>
+    // Telegram HTML parse mode does NOT support <br> and rejects it with
+    // 400 Bad Request: can't parse entities: Unsupported start tag "br".
+    const result = markdownToTelegram('line one  \nline two');
+    expect(result).not.toContain('<br');
+    expect(result).toContain('line one');
+    expect(result).toContain('line two');
+  });
+
+  it('converts backslash hard break to newline', () => {
+    // CommonMark hard break: trailing backslash -> <br>
+    const result = markdownToTelegram('line one\\\nline two');
+    expect(result).not.toContain('<br');
+    expect(result).toContain('line one');
+    expect(result).toContain('line two');
+  });
 });
 
 describe('Discord chunking', () => {

bridge/src/markdown/ir.ts

@@ -156,8 +156,12 @@ export function markdownToHtml(text: string): string {
   // Strip <hr> tags
   html = html.replace(/<hr\s*\/?>/g, '---\n');
 
-  // Strip any remaining unsupported HTML tags (keep: b, i, s, u, code, pre, a, br)
-  html = html.replace(/<\/?(?!b>|\/b>|i>|\/i>|s>|\/s>|u>|\/u>|code>|\/code>|pre>|\/pre>|a[\s>]|\/a>|br\s*\/?>)[a-z][a-z0-9]*[^>]*>/gi, '');
+  // <br> is NOT supported by Telegram HTML parse mode — convert to newline
+  // (Telegram API rejects <br> with 400: Unsupported start tag "br")
+  html = html.replace(/<br\s*\/?>/gi, '\n');
+
+  // Strip any remaining unsupported HTML tags (keep: b, i, s, u, code, pre, a)
+  html = html.replace(/<\/?(?!b>|\/b>|i>|\/i>|s>|\/s>|u>|\/u>|code>|\/code>|pre>|\/pre>|a[\s>]|\/a>)[a-z][a-z0-9]*[^>]*>/gi, '');
 
   return html.trim();
 }