feat: search that looks at what an article says

Both halves of hybrid retrieval were reading the same 331 titles and summaries.
The lexical half was fixed earlier; this is the semantic one. `content` is NULL
for 323 articles because the generator writes into `sections`, so the vector for
98% of the library described the heading and nothing under it.

Depth is carried by the section index, where the longest section in the corpus
is under the embedding clamp — so every sentence of every body is embedded whole
somewhere, and nothing is truncated at that level at all. The article vector is
a topical signal instead: title, summary, the full outline, and an even slice of
every section's opening, budgeted so the clamp never silently fires. Round-robin
rather than head-and-tail, because truncating the head of a twelve-section
article stops in the pathophysiology and drops treatment and management — which
is where the words somebody actually searches for live.

`article_section_index` is populated and stays populated. The rebuild was a
private helper in one router, so the three other writers that save sections —
the generation task, the pipeline script and the seeds — silently skipped it.
That is how 323 articles came to have no rows at all. The generator itself is
one line poorer for it now.

A retrieval bug found on the way: the section-to-article rollup concatenated
rather than fused, so a section matching at rank 1 landed behind every weak
whole-article match and never reached the page. And `/articles/?q=` had no
rollup at all.

3,833 vectors in 332 seconds, batched 32 to a request — a normal article save
is now one round trip rather than fourteen. Proved against the vectors restored
from backup: "surgery for infant stridor that fails to improve" found
Laryngomalacia at rank 159, below the floor and invisible; it is rank 1 now, and
the section corpus answers it at rank 1 having previously been unable to answer
it at all.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01TqXevQJhxFrM7jJg82cgZN

backend/app/services/article_writer.py

@@ -214,6 +214,10 @@ def write_article(db: Session, topic: str, category_id: int | None = None,
     article_service.record_slug(db, article)
     db.commit()
     db.refresh(article)
+    # The section index is what makes the body searchable at all. This writer
+    # produced 323 articles without it, so the library's whole depth was
+    # invisible to search while every one of them looked fine on the page.
+    article_service.reindex(db, article)
     return {"topic": topic, "status": "written", "article_id": article.id,
             "sections": len(sections), "references": len(article.references_json or []),
             "passages": len(passages)}

backend/scripts/reindex_article_search.py

@@ -36,6 +36,8 @@ import sys
 import time
 from datetime import datetime
 
+from sqlalchemy import text as sa_text
+
 from app.database import SessionLocal
 from app.models.article import Article, ArticleSectionIndex
 from app.services import article_service, embedding_service
@@ -59,6 +61,22 @@ def _needs_article_vector(article, active_model: str, started: datetime, force:
     return force and article.embedded_at < started
 
 
+def _settle(db, articles) -> None:
+    """Record that the stored vector matches the row as it now stands.
+
+    `Article.updated_at` carries `onupdate`, so writing the vector bumps it a
+    few milliseconds past the `embedded_at` the same write set. An exact
+    `embedded_at < updated_at` test therefore reports every article stale
+    forever, and a re-run pays for the whole corpus again. Raw SQL because a
+    Core update would fire `onupdate` once more and lose the race a second time.
+    """
+    if not articles:
+        return
+    db.execute(sa_text("UPDATE articles SET embedded_at = GREATEST(embedded_at, updated_at) "
+                       "WHERE id = ANY(:ids)"), {"ids": [a.id for a in articles]})
+    db.commit()
+
+
 def main() -> int:
     apply_changes = "--apply" in sys.argv
     force = "--all" in sys.argv
@@ -91,6 +109,7 @@ def main() -> int:
                 for article in chunk:
                     section_vectors += article_service.rebuild_section_index(db, article)
                 db.commit()
+                _settle(db, wanted)
             except Exception as exc:
                 # One bad article must not cost the run; it is reported and the
                 # next chunk carries on, because a sweep that has to be restarted