retrievalbase 2.3.0__tar.gz → 2.3.1__tar.gz

{retrievalbase-2.3.0 → retrievalbase-2.3.1}/CHANGELOG.md

@@ -1,3 +1,10 @@
+## [2.3.1](https://gitlab.com/efysent/agentic-core/retrievalbase/compare/v2.3.0...v2.3.1) (2026-05-24)
+
+
+### Bug Fixes
+
+* add diagonal relaxed dataframe concat ([cbc9d66](https://gitlab.com/efysent/agentic-core/retrievalbase/commit/cbc9d66985389cc6b28fce916acc699af6df4aa7))
+
 # [2.3.0](https://gitlab.com/efysent/agentic-core/retrievalbase/compare/v2.2.0...v2.3.0) (2026-05-24)
 
 

{retrievalbase-2.3.0 → retrievalbase-2.3.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: retrievalbase
-Version: 2.3.0
+Version: 2.3.1
 Author-email: jalal <jalalkhaldi3@gmail.com>
 Requires-Python: <3.13,>=3.11
 Requires-Dist: faiss-cpu<2.0.0,>=1.13.2

{retrievalbase-2.3.0 → retrievalbase-2.3.1}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "retrievalbase"
-version = "2.3.0"
+version = "2.3.1"
 description = ""
 authors = [
     { name = "jalal", email = "jalalkhaldi3@gmail.com" }

{retrievalbase-2.3.0 → retrievalbase-2.3.1}/src/retrievalbase/connector/minio.py

@@ -34,7 +34,7 @@ class MinioDatasetConnector(DatasetConnector[MinioDatasetConnectorSettings]):
         ]
         if not dataframes:
             raise MinioParquetObjectsNotFoundError(paths or (self.config.key,))
-        return pl.concat(dataframes)
+        return pl.concat(dataframes, how="diagonal_relaxed")
 
     def _read_parquet_object(self, bucket: str, object_name: str) -> pl.DataFrame:
         response = self.client.get_object(bucket, object_name)

{retrievalbase-2.3.0 → retrievalbase-2.3.1}/src/retrievalbase/connector/parquet.py

@@ -24,7 +24,7 @@ class ParquetDatasetConnector(DatasetConnector[ParquetDatasetConnectorSettings])
             raise ParquetFilesNotFoundError(paths or (self.config.path,))
         if self.config.lazy:
             return pl.scan_parquet([str(path) for path in parquet_paths])
-        return pl.concat([pl.read_parquet(path) for path in parquet_paths])
+        return pl.concat([pl.read_parquet(path) for path in parquet_paths], how="diagonal_relaxed")
 
     def _iter_parquet_paths(self, paths: tuple[str, ...]) -> list[Path]:
         parquet_paths: list[Path] = []

{retrievalbase-2.3.0 → retrievalbase-2.3.1}/tests/unit/test_connector/test_connectors.py

@@ -119,6 +119,24 @@ def test_parquet_connector_loads_recursive_paths(tmp_path) -> None:
     }
 
 
+def test_parquet_connector_loads_recursive_paths_with_relaxed_schema(tmp_path) -> None:
+    root = tmp_path / "root"
+    root.mkdir()
+    first = pl.DataFrame({"page_content": ["one"], "metadata": pl.Series([None], dtype=pl.Null)})
+    second = pl.DataFrame({"page_content": ["two"], "metadata": ["source"]})
+    first.write_parquet(root / "first.parquet")
+    second.write_parquet(root / "second.parquet")
+
+    connector = ParquetDatasetConnector(ParquetDatasetConnectorSettings(module_path="x", path=str(root), lazy=False))
+
+    loaded = connector.load_recursive_text()
+
+    assert loaded.polars.to_dict(as_series=False) == {
+        "page_content": ["one", "two"],
+        "metadata": [None, "source"],
+    }
+
+
 def test_parquet_connector_raises_when_recursive_path_has_no_parquet(tmp_path) -> None:
     empty = tmp_path / "empty"
     empty.mkdir()
@@ -307,6 +325,63 @@ def test_minio_connector_loads_recursive_parquet_prefixes(monkeypatch: pytest.Mo
     }
 
 
+def test_minio_connector_loads_recursive_parquet_prefixes_with_relaxed_schema(
+    monkeypatch: pytest.MonkeyPatch,
+) -> None:
+    frames = {
+        "a/first.parquet": pl.DataFrame({"page_content": ["one"], "metadata": pl.Series([None], dtype=pl.Null)}),
+        "a/second.parquet": pl.DataFrame({"page_content": ["two"], "metadata": ["source"]}),
+    }
+
+    def parquet_payload(df: pl.DataFrame) -> bytes:
+        buffer = io.BytesIO()
+        df.write_parquet(buffer)
+        return buffer.getvalue()
+
+    class FakeResponse:
+        def __init__(self, payload: bytes) -> None:
+            self.payload = payload
+
+        def read(self) -> bytes:
+            return self.payload
+
+        def close(self) -> None:
+            pass
+
+        def release_conn(self) -> None:
+            pass
+
+    class FakeMinio:
+        def __init__(self, endpoint: str, access_key: str, secret_key: str, secure: bool) -> None:
+            pass
+
+        def list_objects(self, bucket: str, prefix: str, recursive: bool) -> list[SimpleNamespace]:
+            return [SimpleNamespace(object_name=key) for key in frames if key.startswith(prefix)]
+
+        def get_object(self, bucket: str, key: str) -> FakeResponse:
+            return FakeResponse(parquet_payload(frames[key]))
+
+    monkeypatch.setattr("retrievalbase.connector.minio.Minio", FakeMinio)
+
+    connector = MinioDatasetConnector(
+        MinioDatasetConnectorSettings(
+            module_path="x",
+            endpoint="https://minio.local",
+            bucket="datasets",
+            key="a",
+            access_key=SecretStr("access"),
+            secret_key=SecretStr("secret"),
+        )
+    )
+
+    loaded = connector.load_recursive_text()
+
+    assert loaded.polars.to_dict(as_series=False) == {
+        "page_content": ["one", "two"],
+        "metadata": [None, "source"],
+    }
+
+
 def test_minio_connector_raises_when_prefix_has_no_parquet(monkeypatch: pytest.MonkeyPatch) -> None:
     class FakeMinio:
         def __init__(self, endpoint: str, access_key: str, secret_key: str, secure: bool) -> None:

{retrievalbase-2.3.0 → retrievalbase-2.3.1}/uv.lock

@@ -2584,7 +2584,7 @@ wheels = [
 
 [[package]]
 name = "retrievalbase"
-version = "2.3.0"
+version = "2.3.1"
 source = { editable = "." }
 dependencies = [
     { name = "faiss-cpu" },