feat(metadata): extract title from markdown

Author: Daniele Briggi

src/sqlite_rag/engine.py

@@ -215,6 +215,24 @@ def versions(self) -> dict:
             "vector_version": row["vector_version"],
         }
 
+    def extract_document_title(
+        self, text: str, fallback_first_line: bool = False
+    ) -> str | None:
+        """Extract title from markdown content."""
+        # Look for first level-1 heading
+        match = re.search(r"^# (.+)$", text, re.MULTILINE)
+        if match:
+            return match.group(1).strip()
+
+        # Fallback: first non-empty line
+        if fallback_first_line:
+            for line in text.splitlines():
+                line = line.strip()
+                if line:
+                    return line
+
+        return None
+
     def close(self):
         """Close the database connection."""
         try:

src/sqlite_rag/sqliterag.py

@@ -102,6 +102,11 @@ def add(
                     else str(file_path.absolute())
                 )
                 document = Document(content=content, uri=uri, metadata=metadata)
+                document.metadata["generated"]["title"] = (
+                    self._engine.extract_document_title(
+                        document.content, fallback_first_line=True
+                    )
+                )
 
                 exists = self._repository.document_exists_by_hash(document.hash())
                 if exists:

tests/test_engine.py

@@ -1,8 +1,11 @@
+import pytest
+
 from sqlite_rag.chunker import Chunker
 from sqlite_rag.engine import Engine
 from sqlite_rag.models.chunk import Chunk
 from sqlite_rag.models.document import Document
 from sqlite_rag.repository import Repository
+from sqlite_rag.settings import Settings
 
 
 class TestEngine:
@@ -192,3 +195,32 @@ def test_search_exact_match(self, db_conn):
         assert len(results) > 0
         assert doc1_id == results[0].document.id
         assert 0.0 == results[0].vec_distance
+
+    def test_extract_document_title(self):
+        text = """# This is the Title
+        This is the content of the document.
+        It has multiple lines.
+        """
+
+        engine = Engine(None, Settings(), None)  # type: ignore
+
+        title = engine.extract_document_title(text)
+        assert title == "This is the Title"
+
+    @pytest.mark.parametrize(
+        "fallback, expected_title",
+        [
+            (True, "This is the first line of the document without a title."),
+            (False, None),
+        ],
+    )
+    def test_extract_document_title_from_first_line(self, fallback, expected_title):
+        text = """
+        This is the first line of the document without a title.
+        It has multiple lines.
+        """
+
+        engine = Engine(None, Settings(), None)  # type: ignore
+
+        title = engine.extract_document_title(text, fallback)
+        assert title == expected_title