fix(cli): copy chunk metadata before id assignment to break dict aliasing

packages/leann-core/src/leann/cli.py

@@ -2081,15 +2081,15 @@ def _assign_chunk_ids(chunks: list[dict]) -> None:
         for path_key, path_chunks in by_path.items():
             for idx, c in enumerate(path_chunks):
                 sid = hashlib.sha256(f"{path_key}:{idx}".encode()).hexdigest()[:16]
-                c.setdefault("metadata", {})["id"] = sid
+                c["metadata"] = {**c.get("metadata", {}), "id": sid}
                 c["id"] = sid
 
     @staticmethod
     def _assign_unique_chunk_ids(chunks: list[dict]) -> None:
         """Assign unique IDs for incremental (avoids collision when path lookup misses some old ids)."""
         for c in chunks:
             sid = uuid.uuid4().hex[:16]
-            c.setdefault("metadata", {})["id"] = sid
+            c["metadata"] = {**c.get("metadata", {}), "id": sid}
             c["id"] = sid
 
     def _chunks_for_paths(self, all_texts: list[dict], paths: set[str]) -> list[dict]:

tests/test_incremental_build.py

@@ -371,3 +371,37 @@ def test_ivf_multiple_incremental_no_duplicates(tmp_path):
     assert len(stale_ids) == 0, (
         f"passages.jsonl has {len(stale_ids)} stale entries not in offset_map: {stale_ids[:5]}"
     )
+
+
+def test_assign_unique_chunk_ids_with_shared_metadata_dict():
+    from leann.cli import LeannCLI
+
+    # Arrange
+    shared = {"file_path": "a.md"}
+    chunks = [{"text": t, "metadata": shared} for t in ("one", "two", "three")]
+
+    # Act
+    LeannCLI._assign_unique_chunk_ids(chunks)
+
+    # Assert
+    metadata_ids = [c["metadata"]["id"] for c in chunks]
+    assert len(set(metadata_ids)) == len(chunks)
+    for c in chunks:
+        assert c["metadata"]["id"] == c["id"]
+
+
+def test_assign_chunk_ids_with_shared_metadata_dict():
+    from leann.cli import LeannCLI
+
+    # Arrange
+    shared = {"file_path": "a.md"}
+    chunks = [{"text": t, "metadata": shared} for t in ("one", "two", "three")]
+
+    # Act
+    LeannCLI._assign_chunk_ids(chunks)
+
+    # Assert
+    metadata_ids = [c["metadata"]["id"] for c in chunks]
+    assert len(set(metadata_ids)) == len(chunks)
+    for c in chunks:
+        assert c["metadata"]["id"] == c["id"]