feat(F-16): chunk quality gate + section classification

- codex/quality.py: 3-signal filter (length / alpha-ratio / bib-score)
  + rule-based section classifier + run_quality_pass retroactive DB pass
- codex/ingest.py: promote quality imports to module level; apply
  filter_chunks before embedding; store section column in chunks INSERT
- codex/config.py: CHUNK_MIN_CHARS / CHUNK_MIN_ALPHA_RATIO / CHUNK_MAX_BIB_SCORE
- infra/schema.sql: ALTER TABLE chunks ADD COLUMN IF NOT EXISTS section TEXT
- .env.example: document F-16 quality thresholds
- codex/cli.py: quality run sub-command (scope by --paper-id or all papers)
- tests/quality/: 35 new tests covering all quality functions
- tests/ingest/: patch filter_chunks in source-path tests to isolate ingest

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

codex/ingest.py

@@ -13,6 +13,7 @@ from codex.config import get_settings
 from codex.db import get_conn
 from codex.embed import get_embedder
 from codex.models import Citation, Paper
+from codex.quality import classify_section, filter_chunks
 from codex.sources import openalex, semanticscholar
 
 logger = logging.getLogger(__name__)
@@ -80,7 +81,6 @@ def ingest_paper(
     # ---------------------------------------------------------------
     paper: Paper | None = openalex.fetch_paper(paper_id)
 
-
     if paper is None:
         # Detect arXiv IDs: numeric pattern like "2301.07041" or "arxiv:..." prefix
         pid_lower = paper_id.lower()
@@ -177,7 +177,8 @@ def ingest_paper(
             else:
                 logger.warning("Unbekannter Dateityp: %s — kein Text-Parsing", source_path)
 
-            chunks_text = chunk_text(text) if text else []
+            raw_chunks = chunk_text(text) if text else []
+            chunks_text = filter_chunks(raw_chunks, settings=get_settings())
 
             # Delete existing chunks for this paper
             conn.execute("DELETE FROM chunks WHERE paper_id = %s", (paper.id,))
@@ -186,13 +187,19 @@ def ingest_paper(
                 # Embed all chunks in one batch
                 chunk_embeddings = embedder.encode_dense(chunks_text)
                 chunk_rows = [
-                    (paper.id, ord_idx, content, chunk_embeddings[ord_idx].tolist())
+                    (
+                        paper.id,
+                        ord_idx,
+                        content,
+                        chunk_embeddings[ord_idx].tolist(),
+                        classify_section(content),
+                    )
                     for ord_idx, content in enumerate(chunks_text)
                 ]
                 with conn.cursor() as cur:
                     cur.executemany(
-                        "INSERT INTO chunks (paper_id, ord, content, embedding)"
-                        " VALUES (%s, %s, %s, %s)",
+                        "INSERT INTO chunks (paper_id, ord, content, embedding, section)"
+                        " VALUES (%s, %s, %s, %s, %s)",
                         chunk_rows,
                     )
                 chunks_upserted = len(chunk_rows)
@@ -207,8 +214,7 @@ def ingest_paper(
             # OpenAlex returns citing_id=openalex_id, but papers.id uses the DOI.
             # Rewrite citing_id to paper.id so the FK constraint is satisfied.
             api_citations = [
-                Citation(citing_id=paper.id, cited_id=c.cited_id, context=c.context)
-                for c in raw
+                Citation(citing_id=paper.id, cited_id=c.cited_id, context=c.context) for c in raw
             ]
         else:
             api_citations = semanticscholar.fetch_references(paper_id)