fix(pipeline): slug clashes inside one batch; the source page is never the one skipped (review of #211, pass 2)

dropSlugClashes only compared against what was on disk, so a clip named like an
entity the model extracted wrote both pages and neither could be updated again.
In-batch winners: source over entity over concept. Against the brain, only
entities/concepts are skipped; the source's slug was already chosen free or its
own by resolveSourceSlug, so an existing bad pair no longer locks the source out.
Mutation-checked.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

ingestion/internal/pipeline/pipeline.go

@@ -372,15 +372,53 @@ func resolveSourceSlug(brainDir, wing, base string) (string, error) {
 }
 
 // dropSlugClashes removes, from the pages about to be written in a wing, any
-// page whose slug already belongs to a different page elsewhere in the brain.
-// The page's own path is not a clash (a shared entity merging into itself).
-// A skipped page is a warning, not a failure: the ingest as a whole is fine,
-// and one extracted page must not be able to lock a source out of re-ingest.
+// page whose slug belongs to a different page. A slug is the graph identity, so
+// two pages must not share one; and a page that shares one cannot be updated
+// again, because every later ingest sees the other as its clash.
+//
+// Two kinds of clash, both by slug, neither failing the ingest (one extracted
+// page must not be able to lock a source out of re-ingest):
+//   - inside the batch (a clip named notes.md whose model output also has an
+//     entity "Notes"): the source wins over an entity, an entity over a concept;
+//   - against the brain: an entity or concept whose slug is another page's. The
+//     SOURCE page is exempt: it is the note's own output, and its slug was
+//     already chosen free (or its own) by resolveSourceSlug.
+//
+// A page's own path is never a clash (a shared entity merging into itself).
 func dropSlugClashes(brainDir string, pages []wiki.Page) ([]wiki.Page, []string, error) {
+	rank := func(p wiki.Page) int {
+		switch {
+		case isSourcePath(p.Path):
+			return 0
+		case strings.HasPrefix(p.Path, "wiki/entities/"):
+			return 1
+		case strings.HasPrefix(p.Path, "wiki/concepts/"):
+			return 2
+		}
+		return 3
+	}
+	slugOf := func(p wiki.Page) string { return strings.TrimSuffix(filepath.Base(p.Path), ".md") }
+
+	winner := map[string]wiki.Page{}
+	for _, p := range pages {
+		w, seen := winner[slugOf(p)]
+		if !seen || (w.Path != p.Path && rank(p) < rank(w)) {
+			winner[slugOf(p)] = p
+		}
+	}
+
 	var kept []wiki.Page
 	var warnings []string
 	for _, p := range pages {
-		slug := strings.TrimSuffix(filepath.Base(p.Path), ".md")
+		slug := slugOf(p)
+		if w := winner[slug]; w.Path != p.Path {
+			warnings = append(warnings, fmt.Sprintf("skipped %s: slug %q is also used by %s in this batch", p.Path, slug, w.Path))
+			continue
+		}
+		if rank(p) == 0 {
+			kept = append(kept, p)
+			continue
+		}
 		existing, err := wiki.FindSlugs(brainDir, slug)
 		if err != nil {
 			return nil, nil, fmt.Errorf("check slug %q: %w", slug, err)

ingestion/internal/pipeline/wing_layout_test.go

@@ -233,3 +233,62 @@ func TestResolveSourceSlug_AFurtherClashGetsANumber(t *testing.T) {
 	require.NoError(t, err)
 	assert.Equal(t, "talk-b-2", got)
 }
+
+// From the second review of #211: slug clashes INSIDE one batch. dropSlugClashes
+// only looked at what was already on disk, so a source and an entity extracted
+// together with one slug were both written, and neither could ever be updated.
+func TestRun_ASourceAndAnEntityWithOneSlugInTheSameBatchWriteOnlyTheSource(t *testing.T) {
+	dir := t.TempDir()
+	pages := []RawPage{
+		{Title: "Notes", Type: "source", Subtype: "note", Content: "## Summary\n\nfirst\n"},
+		{Title: "Notes", Type: "entity", Subtype: "tool", Content: "## Definition\n\nan entity named like its source\n"},
+	}
+	res, err := Run(context.Background(), wingCfg("homelab", pages...), dir, "body one", "notes", false)
+	require.NoError(t, err)
+	assert.True(t, exists(dir, "wiki/homelab/sources/notes.md"))
+	assert.False(t, exists(dir, "wiki/entities/notes.md"), "two pages must not share a slug")
+	assert.Contains(t, strings.Join(res.Warnings, "\n"), "notes")
+
+	pages[0].Content = "## Summary\n\nfirst\n\n## Key Claims\n\n- second\n"
+	_, err = Run(context.Background(), wingCfg("homelab", pages...), dir, "body two, changed", "notes", false)
+	require.NoError(t, err)
+	assert.Contains(t, read(t, dir, "wiki/homelab/sources/notes.md"), "second", "the source can still be updated")
+}
+
+func TestRun_AConceptAndAnEntityWithOneSlugInTheSameBatchWriteOne(t *testing.T) {
+	dir := t.TempDir()
+	pages := []RawPage{
+		{Title: "Talk", Type: "source", Subtype: "video", Content: "## Summary\n\nx\n"},
+		{Title: "Foo", Type: "concept", Content: "## Definition\n\na concept\n"},
+		{Title: "Foo", Type: "entity", Subtype: "tool", Content: "## Definition\n\nan entity\n"},
+	}
+	res, err := Run(context.Background(), wingCfg("homelab", pages...), dir, "body", "talk", false)
+	require.NoError(t, err)
+	written := 0
+	for _, rel := range []string{"wiki/concepts/foo.md", "wiki/entities/foo.md"} {
+		if exists(dir, rel) {
+			written++
+		}
+	}
+	assert.Equal(t, 1, written, "one slug, one page")
+	assert.Contains(t, strings.Join(res.Warnings, "\n"), "foo")
+}
+
+// A brain that already holds the bad pair (written before this fix) must not
+// lock the source out: the source page is the note's own output and is never
+// the one skipped.
+func TestRun_AnExistingSourceEntityPairDoesNotLockTheSourceOut(t *testing.T) {
+	dir := t.TempDir()
+	require.NoError(t, os.MkdirAll(filepath.Join(dir, "wiki/homelab/sources"), 0o755))
+	require.NoError(t, os.MkdirAll(filepath.Join(dir, "wiki/entities"), 0o755))
+	require.NoError(t, os.WriteFile(filepath.Join(dir, "wiki/homelab/sources/notes.md"),
+		[]byte("---\ntitle: 'Notes'\ntype: 'note'\nwing: homelab\nhall: sources\naliases:\n  - 'Notes'\n---\n\n## Summary\n\nold\n"), 0o644))
+	require.NoError(t, os.WriteFile(filepath.Join(dir, "wiki/entities/notes.md"),
+		[]byte("---\ntitle: 'Notes'\naliases:\n  - 'Notes'\n---\n\n## Definition\n\nstale entity\n"), 0o644))
+
+	pages := []RawPage{{Title: "Notes", Type: "source", Subtype: "note", Content: "## Summary\n\nnew text\n\n## Key Claims\n\n- fresh\n"}}
+	_, err := Run(context.Background(), wingCfg("homelab", pages...), dir, "changed body", "notes", false)
+	require.NoError(t, err)
+	assert.Contains(t, read(t, dir, "wiki/homelab/sources/notes.md"), "fresh", "the source took the update")
+	assert.Contains(t, read(t, dir, "wiki/entities/notes.md"), "stale entity", "the entity is untouched")
+}