Add mypy type checks

Author: cbornet

.github/workflows/lint.yml

@@ -45,7 +45,7 @@ jobs:
         run: uv run ruff format --check
       - name: Run ruff check
         run: uv run ruff check
-      # - name: Run mypy
-      #   run: uv run mypy .
+      - name: Run mypy
+        run: uv run mypy .
       - name: Minimize uv cache
         run: uv cache prune --ci

langchain/langchain_vectorize/retrievers.py

@@ -4,20 +4,18 @@
 
 from typing import TYPE_CHECKING, Any, Literal, Optional
 
-import vectorize_client
 from langchain_core.documents import Document
 from langchain_core.retrievers import BaseRetriever
 from typing_extensions import override
-from vectorize_client import (
-    ApiClient,
-    Configuration,
-    PipelinesApi,
-    RetrieveDocumentsRequest,
-)
+from vectorize_client.api.pipelines_api import PipelinesApi
+from vectorize_client.api_client import ApiClient
+from vectorize_client.configuration import Configuration
+from vectorize_client.models.retrieve_documents_request import RetrieveDocumentsRequest
 
 if TYPE_CHECKING:
     from langchain_core.callbacks import CallbackManagerForRetrieverRun
     from langchain_core.runnables import RunnableConfig
+    from vectorize_client.models.document import Document as VectorizeDocument
 
 _METADATA_FIELDS = {
     "relevancy",
@@ -122,7 +120,7 @@ def format_docs(docs):
     metadata_filters: list[dict[str, Any]] = []
     """The metadata filters to apply when retrieving the documents."""
 
-    _pipelines: PipelinesApi | None = None
+    _pipelines: PipelinesApi = _NOT_SET  # type: ignore[assignment]
 
     @override
     def model_post_init(self, /, context: Any) -> None:
@@ -146,7 +144,7 @@ def model_post_init(self, /, context: Any) -> None:
         self._pipelines = PipelinesApi(api)
 
     @staticmethod
-    def _convert_document(document: vectorize_client.models.Document) -> Document:
+    def _convert_document(document: VectorizeDocument) -> Document:
         metadata = {field: getattr(document, field) for field in _METADATA_FIELDS}
         return Document(id=document.id, page_content=document.text, metadata=metadata)
 
@@ -162,14 +160,29 @@ def _get_relevant_documents(
         rerank: bool | None = None,
         metadata_filters: list[dict[str, Any]] | None = None,
     ) -> list[Document]:
-        request = RetrieveDocumentsRequest(
+        request = RetrieveDocumentsRequest(  # type: ignore[call-arg]
             question=query,
             num_results=num_results or self.num_results,
             rerank=rerank or self.rerank,
             metadata_filters=metadata_filters or self.metadata_filters,
         )
+        organization_ = organization or self.organization
+        if not organization_:
+            msg = (
+                "Organization must be set either at initialization "
+                "or in the invoke method."
+            )
+            raise ValueError(msg)
+        pipeline_id_ = pipeline_id or self.pipeline_id
+        if not pipeline_id_:
+            msg = (
+                "Pipeline ID must be set either at initialization "
+                "or in the invoke method."
+            )
+            raise ValueError(msg)
+
         response = self._pipelines.retrieve_documents(
-            organization or self.organization, pipeline_id or self.pipeline_id, request
+            organization_, pipeline_id_, request
         )
         return [self._convert_document(doc) for doc in response.documents]
 
@@ -181,9 +194,10 @@ def invoke(
         *,
         organization: str = "",
         pipeline_id: str = "",
-        num_results: int = _NOT_SET,
-        rerank: bool = _NOT_SET,
-        metadata_filters: list[dict[str, Any]] = _NOT_SET,
+        num_results: int = _NOT_SET,  # type: ignore[assignment]
+        rerank: bool = _NOT_SET,  # type: ignore[assignment]
+        metadata_filters: list[dict[str, Any]] = _NOT_SET,  # type: ignore[assignment]
+        **_kwargs: Any,
     ) -> list[Document]:
         """Invoke the retriever to get relevant documents.
 
@@ -218,16 +232,15 @@ def invoke(
                 query = "what year was breath of the wild released?"
                 docs = retriever.invoke(query, num_results=2)
         """
-        kwargs = {}
         if organization:
-            kwargs["organization"] = organization
+            _kwargs["organization"] = organization
         if pipeline_id:
-            kwargs["pipeline_id"] = pipeline_id
+            _kwargs["pipeline_id"] = pipeline_id
         if num_results is not _NOT_SET:
-            kwargs["num_results"] = num_results
+            _kwargs["num_results"] = num_results
         if rerank is not _NOT_SET:
-            kwargs["rerank"] = rerank
+            _kwargs["rerank"] = rerank
         if metadata_filters is not _NOT_SET:
-            kwargs["metadata_filters"] = metadata_filters
+            _kwargs["metadata_filters"] = metadata_filters
 
-        return super().invoke(input, config, **kwargs)
+        return super().invoke(input, config, **_kwargs)

langchain/pyproject.toml

@@ -31,7 +31,7 @@ Issues = "https://github.com/vectorize-io/integrations-python/issues"
 
 [dependency-groups]
 dev = [
-    "mypy>=1.13.0",
+    "mypy>=1.17.1,<1.18",
     "pytest>=8.3.3",
     "ruff>=0.9.0,<0.10",
 ]
@@ -59,6 +59,8 @@ flake8-annotations.mypy-init-return = true
 
 [tool.mypy]
 strict = true
+strict_bytes = true
+enable_error_code = "deprecated"
 warn_unreachable = true
 pretty = true
 show_error_codes = true

langchain/tests/test_retrievers.py

@@ -8,8 +8,29 @@
 
 import pytest
 import urllib3
-import vectorize_client as v
-from vectorize_client import ApiClient
+from vectorize_client.api.connectors_api import ConnectorsApi
+from vectorize_client.api.pipelines_api import PipelinesApi
+from vectorize_client.api.uploads_api import UploadsApi
+from vectorize_client.api_client import ApiClient
+from vectorize_client.configuration import Configuration
+from vectorize_client.models.ai_platform_config_schema import AIPlatformConfigSchema
+from vectorize_client.models.ai_platform_schema import AIPlatformSchema
+from vectorize_client.models.ai_platform_type import AIPlatformType
+from vectorize_client.models.create_source_connector import CreateSourceConnector
+from vectorize_client.models.destination_connector_schema import (
+    DestinationConnectorSchema,
+)
+from vectorize_client.models.destination_connector_type import DestinationConnectorType
+from vectorize_client.models.pipeline_configuration_schema import (
+    PipelineConfigurationSchema,
+)
+from vectorize_client.models.schedule_schema import ScheduleSchema
+from vectorize_client.models.schedule_schema_type import ScheduleSchemaType
+from vectorize_client.models.source_connector_schema import SourceConnectorSchema
+from vectorize_client.models.source_connector_type import SourceConnectorType
+from vectorize_client.models.start_file_upload_to_connector_request import (
+    StartFileUploadToConnectorRequest,
+)
 
 from langchain_vectorize.retrievers import VectorizeRetriever
 
@@ -38,7 +59,7 @@ def environment() -> Literal["prod", "dev", "local", "staging"]:
     if env not in ["prod", "dev", "local", "staging"]:
         msg = "Invalid VECTORIZE_ENV environment variable."
         raise ValueError(msg)
-    return env
+    return env  # type: ignore[return-value]
 
 
 @pytest.fixture(scope="session")
@@ -56,35 +77,31 @@ def api_client(api_token: str, environment: str) -> Iterator[ApiClient]:
     else:
         host = "https://api-staging.vectorize.io/v1"
 
-    with v.ApiClient(
-        v.Configuration(host=host, access_token=api_token, debug=True),
+    with ApiClient(
+        Configuration(host=host, access_token=api_token, debug=True),
         header_name,
         header_value,
     ) as api:
         yield api
 
 
 @pytest.fixture(scope="session")
-def pipeline_id(api_client: v.ApiClient, org_id: str) -> Iterator[str]:
-    pipelines = v.PipelinesApi(api_client)
+def pipeline_id(api_client: ApiClient, org_id: str) -> Iterator[str]:
+    pipelines = PipelinesApi(api_client)
 
-    connectors_api = v.ConnectorsApi(api_client)
+    connectors_api = ConnectorsApi(api_client)
     response = connectors_api.create_source_connector(
         org_id,
-        [
-            v.CreateSourceConnector(
-                name="from api", type=v.SourceConnectorType.FILE_UPLOAD
-            )
-        ],
+        [CreateSourceConnector(name="from api", type=SourceConnectorType.FILE_UPLOAD)],
     )
     source_connector_id = response.connectors[0].id
     logging.info("Created source connector %s", source_connector_id)
 
-    uploads_api = v.UploadsApi(api_client)
+    uploads_api = UploadsApi(api_client)
     upload_response = uploads_api.start_file_upload_to_connector(
         org_id,
         source_connector_id,
-        v.StartFileUploadToConnectorRequest(
+        StartFileUploadToConnectorRequest(  # type: ignore[call-arg]
             name="research.pdf",
             content_type="application/pdf",
             metadata=json.dumps({"created-from-api": True}),
@@ -125,26 +142,26 @@ def pipeline_id(api_client: v.ApiClient, org_id: str) -> Iterator[str]:
 
     pipeline_response = pipelines.create_pipeline(
         org_id,
-        v.PipelineConfigurationSchema(
+        PipelineConfigurationSchema(  # type: ignore[call-arg]
             source_connectors=[
-                v.SourceConnectorSchema(
+                SourceConnectorSchema(
                     id=source_connector_id,
-                    type=v.SourceConnectorType.FILE_UPLOAD,
+                    type=SourceConnectorType.FILE_UPLOAD,
                     config={},
                 )
             ],
-            destination_connector=v.DestinationConnectorSchema(
+            destination_connector=DestinationConnectorSchema(
                 id=builtin_vector_db,
-                type=v.DestinationConnectorType.VECTORIZE,
+                type=DestinationConnectorType.VECTORIZE,
                 config={},
             ),
-            ai_platform=v.AIPlatformSchema(
+            ai_platform=AIPlatformSchema(
                 id=builtin_ai_platform,
-                type=v.AIPlatformType.VECTORIZE,
-                config=v.AIPlatformConfigSchema(),
+                type=AIPlatformType.VECTORIZE,
+                config=AIPlatformConfigSchema(),
             ),
             pipeline_name="Test pipeline",
-            schedule=v.ScheduleSchema(type=v.ScheduleSchemaType.MANUAL),
+            schedule=ScheduleSchema(type=ScheduleSchemaType.MANUAL),
         ),
     )
     pipeline_id = pipeline_response.data.id