fix(matrix): add formatted_body with HTML rendering for cron deliveries

[ygd58]

Fixes #5233

Cron deliveries to Matrix now include formatted_body with HTML-rendered markdown. Also converts h1-h6 to bold for Element X compatibility.

Gracefully falls back to plain text if markdown library is unavailable.

[fxfitz]

Nice fix! This addresses the exact issue from #5233 — cron deliveries to Matrix missing formatted_body. The core approach (markdown → HTML, Element X h-tag workaround, graceful fallback) is spot-on.

A few suggestions after reviewing the diff:

1. Regex should handle header attributes

The current pattern:

html = _re.sub(r'<h[1-6]>(.*?)</h[1-6]>', r'<p><strong>\1</strong></p>', html)

Won't match headers with attributes like <h2 id="summary">. The markdown library can produce these depending on extensions. A more robust pattern:

html = _re.sub(r'<h[1-6][^>]*>(.*?)</h[1-6]>', r'<p><strong>\1</strong></p>', html)

2. Catch all exceptions, not just ImportError

If markdown is installed but throws on unexpected input (malformed text, extension conflicts, etc.), the current code would propagate the exception and break the entire message delivery. Consider:

try:
    import markdown as _md
    import re as _re
    html = _md.markdown(message, extensions=["nl2br", "fenced_code", "tables"])
    html = _re.sub(r'<h[1-6][^>]*>(.*?)</h[1-6]>', r'<p><strong>\1</strong></p>', html)
    if html != message:
        msg_content["format"] = "org.matrix.custom.html"
        msg_content["formatted_body"] = html
except ImportError:
    pass  # markdown not installed
except Exception as err:
    logger.warning("_send_matrix: markdown conversion failed: %s", err)

This way it always falls back to plain text rather than risking a send failure.

3. Consider adding logging

A pass on the ImportError makes debugging harder in production. Even a logger.debug or logger.warning would help operators understand why their messages aren't rendering as HTML.

4. Minor: single-paragraph unwrapping

For short one-line messages, markdown wraps the output in <p>...</p>. This is technically fine per the Matrix spec, but stripping the wrapper when there's only a single paragraph produces cleaner HTML:

if html.count("<p>") == 1:
    html = html.replace("<p>", "").replace("</p>", "")

Not a blocker — just a polish thing.

---

Overall this is a clean, minimal fix that solves the problem. The suggestions above are mostly about resilience and edge cases. Thanks for picking this up so quickly! 🙏

---

This review was generated by a Hermes agent running claude-opus-4-6 (Anthropic).

[teknium1]

Your formatted_body HTML rendering was incorporated into PR #5271 — _send_matrix() now converts markdown to HTML with Element X compatibility. Your authorship is credited. Thanks!