Auto-sync from demo at Wed Nov 5 04:56:04 UTC 2025

graphgen/models/reader/parquet_reader.py

@@ -0,0 +1,20 @@
+from typing import Any, Dict, List
+
+import pandas as pd
+
+from graphgen.bases.base_reader import BaseReader
+
+
+class ParquetReader(BaseReader):
+    """
+    Read parquet files, requiring the schema to be restored to List[Dict[str, Any]].
+    """
+
+    def read(self, file_path: str) -> List[Dict[str, Any]]:
+        df = pd.read_parquet(file_path)
+        data: List[Dict[str, Any]] = df.to_dict(orient="records")
+
+        for doc in data:
+            if doc.get("type") == "text" and self.text_column not in doc:
+                raise ValueError(f"Missing '{self.text_column}' in document: {doc}")
+        return self.filter(data)