fix(comments): don't attach combined.md to bib — siblings cover the same text

combined.md is the consolidated dump of every attachment's extracted
text plus the inline comment body. With per-attachment sibling MDs
also attached, the same text ends up in Zotero twice — once aggregated,
once split — bloating the library and slowing down loads.

Drop combined.md from the attach pass. The file still lives on disk
under .state/comments/ so `stack comments stats` can index it via
DuckDB; only the bib-attachment side stops carrying it.

Companion to a one-shot cleanup that removed 23,610 combined.md
attachments + storage dirs from both bib and Zotero.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

src/cli/comments.py

@@ -81,6 +81,13 @@ def _build_bib_helpers(use_bib: bool, attach: bool):
             return  # comment not in bib
         item_key = info[0]
         for md_path in sorted(comment_dir.glob("*.md")):
+            # Don't attach the aggregated combined.md — it duplicates the
+            # text already in the per-attachment siblings, doubles the
+            # row/file count in Zotero, and slows the library down.
+            # combined.md still lives on disk for `stack comments stats`
+            # to index it.
+            if md_path.name == "combined.md":
+                continue
             key = (item_key, md_path.name)
             if key in existing:
                 continue

tests/cli/test_comments.py

@@ -59,9 +59,11 @@ def test_stats_after_extract(tmp_path: Path):
     assert "1" in result.output  # at least one comment counted
 
 
-def test_extract_attaches_combined_md_and_siblings_to_bib(tmp_path: Path, monkeypatch):
+def test_extract_attaches_only_sibling_mds_to_bib(tmp_path: Path, monkeypatch):
-    """End-to-end: extract writes combined.md plus per-attachment sibling MDs,
+    """End-to-end: extract writes combined.md (for `stats` indexing) plus
-    and attaches all of them to the bib item. Idempotent on re-run."""
+    per-attachment sibling MDs, but only the siblings get attached to the
+    bib item — combined.md is intentionally skipped to avoid duplicating
+    the same text twice in Zotero. Idempotent on re-run."""
     from bib import connect
     from bib.item import Source
 
@@ -81,7 +83,7 @@ def test_extract_attaches_combined_md_and_siblings_to_bib(tmp_path: Path, monkey
 
     result = runner.invoke(app, ["extract", "--root", str(tmp_path), "--workers", "1"])
     assert result.exit_code == 0, result.output
-    assert (cdir / "combined.md").is_file()
+    assert (cdir / "combined.md").is_file()  # still on disk for stats
     assert (cdir / "attachment_1.pdf.md").is_file()
 
     def md_attachments():
@@ -98,8 +100,8 @@ def test_extract_attaches_combined_md_and_siblings_to_bib(tmp_path: Path, monkey
         store.close()
         return rows
 
-    assert md_attachments() == ["attachment_1.pdf.md", "combined.md"]
+    assert md_attachments() == ["attachment_1.pdf.md"]
 
     # Re-running must not duplicate attachments
     runner.invoke(app, ["extract", "--root", str(tmp_path), "--workers", "1"])
-    assert md_attachments() == ["attachment_1.pdf.md", "combined.md"]
+    assert md_attachments() == ["attachment_1.pdf.md"]