fix(llm): ignore the section heading when locating a chunk's PDF page

src/llm/pages.py

@@ -9,6 +9,7 @@ came from and finds the page containing the chunk's opening words.
 from __future__ import annotations
 
 import logging
+import re
 from dataclasses import replace
 from pathlib import Path
 
@@ -18,12 +19,20 @@ log = logging.getLogger(__name__)
 
 _PROBE_CHARS = 60
 _MIN_PROBE = 12
+_HEADING = re.compile(r"\A#{1,6}[^\n]*\n+")
 
 
 def _norm(text: str) -> str:
     return " ".join(text.split())
 
 
+def _probe_text(text: str) -> str:
+    """Chunk text minus a leading markdown heading — the first chunk of a
+    ``## attachment.pdf`` section starts with that heading, which is not
+    in the PDF."""
+    return _HEADING.sub("", text, count=1)
+
+
 def pdf_pages(path: Path) -> list[str]:
     """Whitespace-normalized text per page; ``[]`` when unreadable."""
     try:
@@ -68,7 +77,7 @@ def enrich_pdf_pages(doc: Doc, chunks: list[Chunk]) -> list[Chunk]:
         if path.lower().endswith(".pdf") and not md.get("page"):
             if path not in cache:
                 cache[path] = pdf_pages(Path(path))
-            n = locate(cache[path], c.text)
+            n = locate(cache[path], _probe_text(c.text))
             md["page"] = str(n) if n else ""
         out.append(replace(c, metadata=md))
     return out

tests/llm/test_pages.py

@@ -66,6 +66,17 @@ class TestEnrich:
         assert "attachment" not in out[1].metadata
         assert out[0].id == "c" and out[0].text == chunks[0].text
 
+    def test_leading_section_heading_ignored_when_probing(self, pdf):
+        doc = Doc(
+            key="K", text="", metadata={}, files=(("attachment_1.pdf", str(pdf)),)
+        )
+        c = self._chunk(
+            "## attachment_1.pdf\n\nPage one talks about telehealth originating sites.",
+            "attachment_1.pdf",
+        )
+        (out,) = enrich_pdf_pages(doc, [c])
+        assert out.metadata["page"] == "1"
+
     def test_unlocated_chunk_keeps_attachment_without_page(self, pdf):
         doc = Doc(
             key="K", text="", metadata={}, files=(("attachment_1.pdf", str(pdf)),)