Fix JSONL loader handling of blank/invalid lines

.semversioner/next-release/patch-20260710221345787890.json

@@ -0,0 +1,4 @@
+{
+  "type": "patch",
+  "description": "Fix JSONL input loader to skip blank lines and ignore invalid JSON rows"
+}

packages/graphrag-input/graphrag_input/jsonl.py

@@ -34,5 +34,28 @@ async def read_file(self, path: str) -> list[TextDocument]:
             - output - list with a TextDocument for each row in the file.
         """
         text = await self._storage.get(path, encoding=self._encoding)
-        rows = [json.loads(line) for line in text.splitlines()]
+        rows: list[dict] = []
+        for line_number, line in enumerate(text.splitlines(), start=1):
+            if not line.strip():
+                continue
+
+            try:
+                parsed_row = json.loads(line)
+            except json.JSONDecodeError:
+                logger.warning(
+                    "Skipping malformed JSONL row in %s at line %s",
+                    path,
+                    line_number,
+                )
+                continue
+
+            if isinstance(parsed_row, dict):
+                rows.append(parsed_row)
+            else:
+                logger.warning(
+                    "Skipping non-object JSONL row in %s at line %s",
+                    path,
+                    line_number,
+                )
+
         return await self.process_data_columns(rows, path)

tests/unit/indexing/input/data/jsonl-with-invalid-and-blank-lines/input.jsonl

@@ -0,0 +1,7 @@
+{ "title": "Hello", "text": "Hi how are you today?"}
+
+not json
+{ "title": "Goodbye", "text": "I'm outta here"}
+
+["not", "an", "object"]
+{ "title": "Adios", "text": "See you later"}

tests/unit/indexing/input/test_jsonl_loader.py

@@ -40,3 +40,20 @@ async def test_jsonl_loader_one_file_with_title():
     documents = await reader.read_files()
     assert len(documents) == 3
     assert documents[0].title == "Hello"
+
+
+async def test_jsonl_loader_skips_blank_and_invalid_rows():
+    config = InputConfig(
+        type=InputType.JsonLines,
+        title_column="title",
+    )
+    storage = create_storage(
+        StorageConfig(
+            base_dir="tests/unit/indexing/input/data/jsonl-with-invalid-and-blank-lines",
+        )
+    )
+    reader = create_input_reader(config, storage)
+    documents = await reader.read_files()
+
+    assert len(documents) == 3
+    assert [document.title for document in documents] == ["Hello", "Goodbye", "Adios"]