polish: no double titles, wrapped map labels, export names its format (v0.3.3)

[mabry1985]

Three cosmetic items surfaced by the showcase QA run:

• View: page titles rendered twice — the reader's own heading plus the content's leading H1. The view now strips a leading H1 that matches the page title (a deliberately different H1 is kept).
• Map: node labels chopped at 22 chars ("Tool Execution Failure…"). Labels now word-wrap to two lines, ellipsizing only past that; the full title rides the node as a native hover tooltip.
• wiki_export: the tool reply now names which format it wrote ("markdown" or "archive"), so the tutor's confirmation is unambiguous.

102 tests green (2 new for the wrap behavior).

🤖 Generated with Claude Code

[protoreview]

QA panel review — WARN

code-review-structural · head 80510572888e · formal

Low-risk PR (title dedup + wiki_export format key): all four findings confirmed, no blockers or majors, nothing refuted — verification changed only the framing, re-reading at branch tip because the head SHA was force-pushed/orphaned (all quotes byte-for-byte). Fix first: view.py:179 — when a body is only a title-matching H1, stripDupTitle returns "" and the || fallback renders the stub placeholder (or a blank body for "# Title\n\n"); the heading still shows as the H2 above, so the invariant to fix is the body-content drop, not the placeholder. Panel agreed on substance; the verifier's nuance is that "silently empty" slightly overstates since the heading still renders. Coverage: no gaps — all three touched files carry findings and the structural pass completed; the real holes are findings 2 and 3 (the format/archive branch and stripDupTitle are both untested).

Findings

	Severity	Location	Finding	Verified
🟡	minor	view.py:179	A page whose markdown body consists solely of a leading H1 matching its title now renders the "stub — nothing filed yet" placeholder instead of the heading: …	confirmed
🟡	minor	view.py:132	stripDupTitle — the PR's headline fix for doubled page titles — ships with no test; test_view.py, the suite that asserts against the PAGE string, was not exten…	confirmed
🟡	minor	tools.py:244	The new format reply key added to wiki_export — the PR's stated purpose — is never asserted by any test, and the archive branch where format="archive" was …	confirmed
⚪	nit	knowledge_map.py:36	_wrap_label's docstring says "only the last line ellipsizes", but the return comprehension ellipsizes any line exceeding width — a non-final line holding a sin…	confirmed

findings JSON (machine-readable)

[
  {
    "file": "view.py",
    "line": 179,
    "severity": "minor",
    "category": "removed-behavior",
    "claim": "A page whose markdown body consists solely of a leading H1 matching its title now renders the \"*stub \u2014 nothing filed yet*\" placeholder instead of the heading: stripDupTitle returns \"\" (falsy) so the `||` fallback fires, and real content is silently displayed as empty.",
    "evidence": "+        <div class=\"md\"><h2>${esc(page.title)}</h2>${md(stripDupTitle(page.content_md, page.title) || \"*stub \u2014 nothing filed yet*\")}</div>\n+    return lines.slice(i + 1).join(\"\\n\");",
    "verdict": "confirmed",
    "note": "Re-read at branch tip (head SHA orphaned; retried without ref): both evidence lines verbatim in file. Traced: content \"# Title\" or \"# Title\\n\" -> split, slice(i+1).join -> \"\" (falsy) -> stub fires. Precision: \"# Title\\n\\n\" returns \"\\n\" (truthy) -> renders blank instead of stub; body content is dropped either way. Heading still renders as the H2 above, so 'silently displayed as empty' slightly overstates, but the behavior change and stub-fire are exactly as claimed."
  },
  {
    "file": "view.py",
    "line": 132,
    "severity": "minor",
    "category": "tests",
    "claim": "stripDupTitle \u2014 the PR's headline fix for doubled page titles \u2014 ships with no test; test_view.py, the suite that asserts against the PAGE string, was not extended to pin the dedup behavior or its use in openPage.",
    "evidence": "+  function stripDupTitle(src, title){\n+        <div class=\"md\"><h2>${esc(page.title)}</h2>${md(stripDupTitle(page.content_md, page.title) || \"*stub \u2014 nothing filed yet*\")}</div>",
    "verdict": "confirmed",
    "note": "test_view.py read in full (7 tests, all against PAGE string): no test references stripDupTitle or the dedup call site; no other test file touches the view JS. Both evidence lines verbatim in diff and file."
  },
  {
    "file": "tools.py",
    "line": 244,
    "severity": "minor",
    "category": "tests",
    "claim": "The new `format` reply key added to wiki_export \u2014 the PR's stated purpose \u2014 is never asserted by any test, and the archive branch where `format=\"archive\"` was added has no test coverage at all (test_wiki_export_tool still only checks ok/files).",
    "evidence": "-                return _ok(**res)\n+                return _ok(format=\"archive\", **res)\n-            return _ok(dir=target, files=n)\n+            return _ok(format=\"markdown\", dir=target, files=n)",
    "verdict": "confirmed",
    "note": "All four diff lines verbatim. test_tools.py's test_wiki_export_tool asserts only out['ok'] and out['files'] == 1 (default markdown branch); no test in the repo invokes the wiki_export tool with format=\"archive\". test_archive.py covers store.export_archive directly, but the tool's archive branch and the format reply key are unpinned \u2014 claim accurate."
  },
  {
    "file": "knowledge_map.py",
    "line": 36,
    "severity": "nit",
    "category": "conventions",
    "claim": "_wrap_label's docstring says \"only the last line ellipsizes\", but the return comprehension ellipsizes any line exceeding width \u2014 a non-final line holding a single word wider than LABEL_W is mid-word truncated too, so the comment misstates the behavior.",
    "evidence": "+    \"\"\"Word-wrap a node label to at most max_lines; only the last line ellipsizes.\"\"\"\n+    return [ln if len(ln) <= width else ln[: width - 1] + \"\u2026\" for ln in lines if ln]",
    "verdict": "confirmed",
    "note": "Both evidence lines verbatim in file. Traced char-by-char: title 'supercalifragilisticexpialidocious x' (word len 34 > 26) -> first word set via 'not cur' branch, next word pushes it into lines as a NON-final line -> comprehension mid-word truncates it (ln[:25]+'\u2026'). A non-final line does ellipsize, so the docstring misstates. Nit severity correct."
  }
]