fix(ingest): normalize Markdown front-matter dates before storage

src/dynavec/ingest.py

@@ -16,6 +16,7 @@
 
 from __future__ import annotations
 
+import datetime
 import hashlib
 from collections.abc import Iterable, Iterator
 from dataclasses import dataclass, field
@@ -30,6 +31,24 @@
 Metadata = dict[str, Any]
 
 
+def _normalize_front_matter(value: Any) -> Any:
+    """Recursively normalize parsed YAML front-matter for storage.
+
+    ``yaml.safe_load()`` converts unquoted dates (``date: 2026-09-24``) into
+    ``datetime.date`` / ``datetime.datetime`` objects, which DynamoDB's
+    ``TypeSerializer`` rejects. Convert those to ISO 8601 strings, recurse
+    through mappings and sequences, and leave strings, ints, floats, and
+    booleans untouched.
+    """
+    if isinstance(value, (datetime.datetime, datetime.date)):
+        return value.isoformat()
+    if isinstance(value, dict):
+        return {k: _normalize_front_matter(v) for k, v in value.items()}
+    if isinstance(value, (list, tuple)):
+        return [_normalize_front_matter(v) for v in value]
+    return value
+
+
 @dataclass
 class Record:
     """One source document before chunking."""
@@ -360,7 +379,9 @@ def _front_matter(text: str, path: Path) -> tuple[str, Metadata]:
             metadata = {}
         if not isinstance(metadata, dict) or any(not isinstance(k, str) for k in metadata):
             raise ValueError(f"Front matter in {path} must be a mapping with string keys")
-        return "".join(lines[end + 1 :]), metadata
+        # yaml.safe_load() turns unquoted dates into datetime.date/datetime
+        # objects, which storage backends (e.g. DynamoDB) cannot serialize.
+        return "".join(lines[end + 1 :]), _normalize_front_matter(metadata)
 
     def __iter__(self) -> Iterator[Record]:
         for path in sorted(self.root.glob(self.glob)):

tests/test_markdown_source.py

@@ -115,6 +115,79 @@ def test_root_must_be_an_existing_directory(tmp_path):
 def test_empty_directory_yields_no_records(tmp_path):
     assert list(MarkdownSource(tmp_path)) == []
 
+def test_front_matter_dates_normalize_to_iso_strings(tmp_path):
+    (tmp_path / "notes.md").write_text(
+        "---\ntitle: Research Notes\ndate: 2026-09-24\n---\nBody\n",
+        encoding="utf-8",
+    )
+
+    record = next(iter(MarkdownSource(tmp_path)))
+
+    assert record.metadata["date"] == "2026-09-24"
+    assert isinstance(record.metadata["date"], str)
+
+
+def test_front_matter_timestamps_normalize_to_iso_strings(tmp_path):
+    (tmp_path / "notes.md").write_text(
+        "---\ntitle: Notes\ncreated: 2026-09-24 10:00:00\n---\nBody\n",
+        encoding="utf-8",
+    )
+
+    record = next(iter(MarkdownSource(tmp_path)))
+
+    assert record.metadata["created"] == "2026-09-24T10:00:00"
+    assert isinstance(record.metadata["created"], str)
+
+
+def test_front_matter_nested_dates_normalize(tmp_path):
+    (tmp_path / "notes.md").write_text(
+        "---\ntitle: Notes\n"
+        "schedule:\n  start: 2026-09-24\n  milestones: [2026-09-25, 2026-09-26]\n---\nBody\n",
+        encoding="utf-8",
+    )
+
+    record = next(iter(MarkdownSource(tmp_path)))
+
+    assert record.metadata["schedule"] == {
+        "start": "2026-09-24",
+        "milestones": ["2026-09-25", "2026-09-26"],
+    }
+
+
+def test_front_matter_quoted_dates_and_scalars_untouched(tmp_path):
+    (tmp_path / "notes.md").write_text(
+        "---\ntitle: Notes\ndate: '2026-09-24'\nyear: 2026\npublished: true\n"
+        "score: 4.5\n---\nBody\n",
+        encoding="utf-8",
+    )
+
+    record = next(iter(MarkdownSource(tmp_path)))
+
+    assert record.metadata["date"] == "2026-09-24"
+    assert record.metadata["year"] == 2026
+    assert record.metadata["published"] is True
+    assert record.metadata["score"] == 4.5
+
+
+def test_front_matter_dates_survive_dynamodb_serialization(tmp_path):
+    """End-to-end: ingest a dated Markdown file and prove the stored item
+    passes DynamoDB's TypeSerializer (issue #246)."""
+    from boto3.dynamodb.types import TypeSerializer
+
+    from dynavec.stores.dynamodb import _to_dynamo
+
+    (tmp_path / "notes.md").write_text(
+        "---\ntitle: Research Notes\ndate: 2026-09-24\n---\nBody\n",
+        encoding="utf-8",
+    )
+    record = next(iter(MarkdownSource(tmp_path)))
+
+    serializer = TypeSerializer()
+    serialized = serializer.serialize(_to_dynamo(record.metadata))
+
+    assert serialized["M"]["date"] == {"S": "2026-09-24"}
+
+
 
 def test_markdown_source_flows_through_ingest(tmp_path):
     (tmp_path / "guide.md").write_text("---\ntopic: aws\n---\nabcdefghij", encoding="utf-8")