arize-phoenix 4.10.2rc2__py3-none-any.whl → 4.12.0__py3-none-any.whl

phoenix/server/api/routers/v1/datasets.py

@@ -6,7 +6,6 @@ import logging
 import zlib
 from asyncio import QueueFull
 from collections import Counter
-from datetime import datetime
 from enum import Enum
 from functools import partial
 from typing import (
@@ -14,7 +13,6 @@ from typing import (
     Awaitable,
     Callable,
     Coroutine,
-    Dict,
     FrozenSet,
     Iterator,
     List,
@@ -28,17 +26,14 @@ from typing import (
 
 import pandas as pd
 import pyarrow as pa
-from fastapi import APIRouter, BackgroundTasks, HTTPException, Path, Query
-from fastapi.responses import PlainTextResponse, StreamingResponse
-from pydantic import BaseModel
 from sqlalchemy import and_, delete, func, select
 from sqlalchemy.ext.asyncio import AsyncSession
+from starlette.background import BackgroundTasks
 from starlette.concurrency import run_in_threadpool
 from starlette.datastructures import FormData, UploadFile
 from starlette.requests import Request
-from starlette.responses import Response
+from starlette.responses import JSONResponse, Response
 from starlette.status import (
-    HTTP_200_OK,
     HTTP_204_NO_CONTENT,
     HTTP_404_NOT_FOUND,
     HTTP_409_CONFLICT,
@@ -56,59 +51,79 @@ from phoenix.db.insertion.dataset import (
     ExampleContent,
     add_dataset_examples,
 )
-from phoenix.server.api.types.Dataset import Dataset as DatasetNodeType
+from phoenix.server.api.types.Dataset import Dataset
 from phoenix.server.api.types.DatasetExample import DatasetExample
-from phoenix.server.api.types.DatasetVersion import DatasetVersion as DatasetVersionNodeType
+from phoenix.server.api.types.DatasetVersion import DatasetVersion
 from phoenix.server.api.types.node import from_global_id_with_expected_type
 from phoenix.server.api.utils import delete_projects, delete_traces
 
-from .dataset_examples import router as dataset_examples_router
-from .utils import (
-    PaginatedResponseBody,
-    ResponseBody,
-    add_errors_to_responses,
-    add_text_csv_content_to_responses,
-)
-
 logger = logging.getLogger(__name__)
 
-DATASET_NODE_NAME = DatasetNodeType.__name__
-DATASET_VERSION_NODE_NAME = DatasetVersionNodeType.__name__
-
-
-router = APIRouter(tags=["datasets"])
-
-
-class Dataset(BaseModel):
-    id: str
-    name: str
-    description: Optional[str]
-    metadata: Dict[str, Any]
-    created_at: datetime
-    updated_at: datetime
-
-
-class ListDatasetsResponseBody(PaginatedResponseBody[Dataset]):
-    pass
-
-
-@router.get(
-    "/datasets",
-    operation_id="listDatasets",
-    summary="List datasets",
-    responses=add_errors_to_responses([HTTP_422_UNPROCESSABLE_ENTITY]),
-)
-async def list_datasets(
-    request: Request,
-    cursor: Optional[str] = Query(
-        default=None,
-        description="Cursor for pagination",
-    ),
-    name: Optional[str] = Query(default=None, description="An optional dataset name to filter by"),
-    limit: int = Query(
-        default=10, description="The max number of datasets to return at a time.", gt=0
-    ),
-) -> ListDatasetsResponseBody:
+NODE_NAME = "Dataset"
+
+
+async def list_datasets(request: Request) -> Response:
+    """
+    summary: List datasets with cursor-based pagination
+    operationId: listDatasets
+    tags:
+      - datasets
+    parameters:
+      - in: query
+        name: cursor
+        required: false
+        schema:
+          type: string
+        description: Cursor for pagination
+      - in: query
+        name: limit
+        required: false
+        schema:
+          type: integer
+          default: 10
+      - in: query
+        name: name
+        required: false
+        schema:
+          type: string
+        description: match by dataset name
+    responses:
+      200:
+        description: A paginated list of datasets
+        content:
+          application/json:
+            schema:
+              type: object
+              properties:
+                next_cursor:
+                  type: string
+                data:
+                  type: array
+                  items:
+                    type: object
+                    properties:
+                      id:
+                        type: string
+                      name:
+                        type: string
+                      description:
+                        type: string
+                      metadata:
+                        type: object
+                      created_at:
+                        type: string
+                        format: date-time
+                      updated_at:
+                        type: string
+                        format: date-time
+      403:
+        description: Forbidden
+      404:
+        description: No datasets found
+    """
+    name = request.query_params.get("name")
+    cursor = request.query_params.get("cursor")
+    limit = int(request.query_params.get("limit", 10))
     async with request.app.state.db() as session:
         query = select(models.Dataset).order_by(models.Dataset.id.desc())
 
@@ -117,8 +132,8 @@ async def list_datasets(
                 cursor_id = GlobalID.from_id(cursor).node_id
                 query = query.filter(models.Dataset.id <= int(cursor_id))
             except ValueError:
-                raise HTTPException(
-                    detail=f"Invalid cursor format: {cursor}",
+                return Response(
+                    content=f"Invalid cursor format: {cursor}",
                     status_code=HTTP_422_UNPROCESSABLE_ENTITY,
                 )
         if name:
@@ -129,56 +144,67 @@ async def list_datasets(
         datasets = result.scalars().all()
 
         if not datasets:
-            return ListDatasetsResponseBody(next_cursor=None, data=[])
+            return JSONResponse(content={"next_cursor": None, "data": []}, status_code=200)
 
         next_cursor = None
         if len(datasets) == limit + 1:
-            next_cursor = str(GlobalID(DATASET_NODE_NAME, str(datasets[-1].id)))
+            next_cursor = str(GlobalID(NODE_NAME, str(datasets[-1].id)))
             datasets = datasets[:-1]
 
         data = []
         for dataset in datasets:
             data.append(
-                Dataset(
-                    id=str(GlobalID(DATASET_NODE_NAME, str(dataset.id))),
-                    name=dataset.name,
-                    description=dataset.description,
-                    metadata=dataset.metadata_,
-                    created_at=dataset.created_at,
-                    updated_at=dataset.updated_at,
-                )
+                {
+                    "id": str(GlobalID(NODE_NAME, str(dataset.id))),
+                    "name": dataset.name,
+                    "description": dataset.description,
+                    "metadata": dataset.metadata_,
+                    "created_at": dataset.created_at.isoformat(),
+                    "updated_at": dataset.updated_at.isoformat(),
+                }
             )
 
-        return ListDatasetsResponseBody(next_cursor=next_cursor, data=data)
-
-
-@router.delete(
-    "/datasets/{id}",
-    operation_id="deleteDatasetById",
-    summary="Delete dataset by ID",
-    status_code=HTTP_204_NO_CONTENT,
-    responses=add_errors_to_responses(
-        [
-            {"status_code": HTTP_404_NOT_FOUND, "description": "Dataset not found"},
-            {"status_code": HTTP_422_UNPROCESSABLE_ENTITY, "description": "Invalid dataset ID"},
-        ]
-    ),
-)
-async def delete_dataset(
-    request: Request, id: str = Path(description="The ID of the dataset to delete.")
-) -> None:
-    if id:
+        return JSONResponse(content={"next_cursor": next_cursor, "data": data})
+
+
+async def delete_dataset_by_id(request: Request) -> Response:
+    """
+    summary: Delete dataset by ID
+    operationId: deleteDatasetById
+    tags:
+      - datasets
+    parameters:
+      - in: path
+        name: id
+        required: true
+        schema:
+          type: string
+    responses:
+      204:
+        description: Success
+      403:
+        description: Forbidden
+      404:
+        description: Dataset not found
+      422:
+        description: Dataset ID is invalid
+    """
+    if id_ := request.path_params.get("id"):
         try:
             dataset_id = from_global_id_with_expected_type(
-                GlobalID.from_id(id),
-                DATASET_NODE_NAME,
+                GlobalID.from_id(id_),
+                Dataset.__name__,
             )
         except ValueError:
-            raise HTTPException(
-                detail=f"Invalid Dataset ID: {id}", status_code=HTTP_422_UNPROCESSABLE_ENTITY
+            return Response(
+                content=f"Invalid Dataset ID: {id_}",
+                status_code=HTTP_422_UNPROCESSABLE_ENTITY,
             )
     else:
-        raise HTTPException(detail="Missing Dataset ID", status_code=HTTP_422_UNPROCESSABLE_ENTITY)
+        return Response(
+            content="Missing Dataset ID",
+            status_code=HTTP_422_UNPROCESSABLE_ENTITY,
+        )
     project_names_stmt = get_project_names_for_datasets(dataset_id)
     eval_trace_ids_stmt = get_eval_trace_ids_for_datasets(dataset_id)
     stmt = (
@@ -188,34 +214,59 @@ async def delete_dataset(
         project_names = await session.scalars(project_names_stmt)
         eval_trace_ids = await session.scalars(eval_trace_ids_stmt)
         if (await session.scalar(stmt)) is None:
-            raise HTTPException(detail="Dataset does not exist", status_code=HTTP_404_NOT_FOUND)
+            return Response(content="Dataset does not exist", status_code=HTTP_404_NOT_FOUND)
     tasks = BackgroundTasks()
     tasks.add_task(delete_projects, request.app.state.db, *project_names)
     tasks.add_task(delete_traces, request.app.state.db, *eval_trace_ids)
-
-
-class DatasetWithExampleCount(Dataset):
-    example_count: int
-
-
-class GetDatasetResponseBody(ResponseBody[DatasetWithExampleCount]):
-    pass
-
-
-@router.get(
-    "/datasets/{id}",
-    operation_id="getDataset",
-    summary="Get dataset by ID",
-    responses=add_errors_to_responses([HTTP_404_NOT_FOUND]),
-)
-async def get_dataset(
-    request: Request, id: str = Path(description="The ID of the dataset")
-) -> GetDatasetResponseBody:
-    dataset_id = GlobalID.from_id(id)
-
-    if (type_name := dataset_id.type_name) != DATASET_NODE_NAME:
-        raise HTTPException(
-            detail=f"ID {dataset_id} refers to a f{type_name}", status_code=HTTP_404_NOT_FOUND
+    return Response(status_code=HTTP_204_NO_CONTENT, background=tasks)
+
+
+async def get_dataset_by_id(request: Request) -> Response:
+    """
+    summary: Get dataset by ID
+    operationId: getDatasetById
+    tags:
+      - datasets
+    parameters:
+      - in: path
+        name: id
+        required: true
+        schema:
+          type: string
+    responses:
+      200:
+        description: Success
+        content:
+          application/json:
+            schema:
+              type: object
+              properties:
+                id:
+                  type: string
+                name:
+                  type: string
+                description:
+                  type: string
+                metadata:
+                  type: object
+                created_at:
+                  type: string
+                  format: date-time
+                updated_at:
+                  type: string
+                  format: date-time
+                example_count:
+                  type: integer
+      403:
+        description: Forbidden
+      404:
+        description: Dataset not found
+    """
+    dataset_id = GlobalID.from_id(request.path_params["id"])
+
+    if (type_name := dataset_id.type_name) != NODE_NAME:
+        return Response(
+            content=f"ID {dataset_id} refers to a f{type_name}", status_code=HTTP_404_NOT_FOUND
         )
     async with request.app.state.db() as session:
         result = await session.execute(
@@ -227,64 +278,97 @@ async def get_dataset(
         dataset = dataset_query[0] if dataset_query else None
         example_count = dataset_query[1] if dataset_query else 0
         if dataset is None:
-            raise HTTPException(
-                detail=f"Dataset with ID {dataset_id} not found", status_code=HTTP_404_NOT_FOUND
+            return Response(
+                content=f"Dataset with ID {dataset_id} not found", status_code=HTTP_404_NOT_FOUND
             )
 
-        dataset = DatasetWithExampleCount(
-            id=str(dataset_id),
-            name=dataset.name,
-            description=dataset.description,
-            metadata=dataset.metadata_,
-            created_at=dataset.created_at,
-            updated_at=dataset.updated_at,
-            example_count=example_count,
-        )
-        return GetDatasetResponseBody(data=dataset)
-
-
-class DatasetVersion(BaseModel):
-    version_id: str
-    description: str
-    metadata: Dict[str, Any]
-    created_at: datetime
-
-
-class ListDatasetVersionsResponseBody(PaginatedResponseBody[DatasetVersion]):
-    pass
-
-
-@router.get(
-    "/datasets/{id}/versions",
-    operation_id="listDatasetVersionsByDatasetId",
-    summary="List dataset versions",
-    responses=add_errors_to_responses([HTTP_422_UNPROCESSABLE_ENTITY]),
-)
-async def list_dataset_versions(
-    request: Request,
-    id: str = Path(description="The ID of the dataset"),
-    cursor: Optional[str] = Query(
-        default=None,
-        description="Cursor for pagination",
-    ),
-    limit: int = Query(
-        default=10, description="The max number of dataset versions to return at a time", gt=0
-    ),
-) -> ListDatasetVersionsResponseBody:
-    if id:
+        output_dict = {
+            "id": str(dataset_id),
+            "name": dataset.name,
+            "description": dataset.description,
+            "metadata": dataset.metadata_,
+            "created_at": dataset.created_at.isoformat(),
+            "updated_at": dataset.updated_at.isoformat(),
+            "example_count": example_count,
+        }
+        return JSONResponse(content={"data": output_dict})
+
+
+async def get_dataset_versions(request: Request) -> Response:
+    """
+    summary: Get dataset versions (sorted from latest to oldest)
+    operationId: getDatasetVersionsByDatasetId
+    tags:
+      - datasets
+    parameters:
+      - in: path
+        name: id
+        required: true
+        description: Dataset ID
+        schema:
+          type: string
+      - in: query
+        name: cursor
+        description: Cursor for pagination.
+        schema:
+          type: string
+      - in: query
+        name: limit
+        description: Maximum number versions to return.
+        schema:
+          type: integer
+          default: 10
+    responses:
+      200:
+        description: Success
+        content:
+          application/json:
+            schema:
+              type: object
+              properties:
+                next_cursor:
+                  type: string
+                data:
+                  type: array
+                  items:
+                    type: object
+                    properties:
+                      version_id:
+                        type: string
+                      description:
+                        type: string
+                      metadata:
+                        type: object
+                      created_at:
+                        type: string
+                        format: date-time
+      403:
+        description: Forbidden
+      422:
+        description: Dataset ID, cursor or limit is invalid.
+    """
+    if id_ := request.path_params.get("id"):
         try:
             dataset_id = from_global_id_with_expected_type(
-                GlobalID.from_id(id),
-                DATASET_NODE_NAME,
+                GlobalID.from_id(id_),
+                Dataset.__name__,
             )
         except ValueError:
-            raise HTTPException(
-                detail=f"Invalid Dataset ID: {id}",
+            return Response(
+                content=f"Invalid Dataset ID: {id_}",
                 status_code=HTTP_422_UNPROCESSABLE_ENTITY,
             )
     else:
-        raise HTTPException(
-            detail="Missing Dataset ID",
+        return Response(
+            content="Missing Dataset ID",
+            status_code=HTTP_422_UNPROCESSABLE_ENTITY,
+        )
+    try:
+        limit = int(request.query_params.get("limit", 10))
+        assert limit > 0
+    except (ValueError, AssertionError):
+        return Response(
+            content="Invalid limit parameter",
             status_code=HTTP_422_UNPROCESSABLE_ENTITY,
         )
     stmt = (
@@ -293,14 +377,15 @@ async def list_dataset_versions(
         .order_by(models.DatasetVersion.id.desc())
         .limit(limit + 1)
     )
-    if cursor:
+    if cursor := request.query_params.get("cursor"):
         try:
             dataset_version_id = from_global_id_with_expected_type(
-                GlobalID.from_id(cursor), DATASET_VERSION_NODE_NAME
+                GlobalID.from_id(cursor),
+                DatasetVersion.__name__,
             )
         except ValueError:
-            raise HTTPException(
-                detail=f"Invalid cursor: {cursor}",
+            return Response(
+                content=f"Invalid cursor: {cursor}",
                 status_code=HTTP_422_UNPROCESSABLE_ENTITY,
             )
         max_dataset_version_id = (
@@ -311,99 +396,102 @@ async def list_dataset_versions(
         stmt = stmt.filter(models.DatasetVersion.id <= max_dataset_version_id)
     async with request.app.state.db() as session:
         data = [
-            DatasetVersion(
-                version_id=str(GlobalID(DATASET_VERSION_NODE_NAME, str(version.id))),
-                description=version.description,
-                metadata=version.metadata_,
-                created_at=version.created_at,
-            )
-            async for version in await session.stream_scalars(stmt)
-        ]
-    next_cursor = data.pop().version_id if len(data) == limit + 1 else None
-    return ListDatasetVersionsResponseBody(data=data, next_cursor=next_cursor)
-
-
-class UploadDatasetData(BaseModel):
-    dataset_id: str
-
-
-class UploadDatasetResponseBody(ResponseBody[UploadDatasetData]):
-    pass
-
-
-@router.post(
-    "/datasets/upload",
-    operation_id="uploadDataset",
-    summary="Upload dataset from JSON, CSV, or PyArrow",
-    responses=add_errors_to_responses(
-        [
             {
-                "status_code": HTTP_409_CONFLICT,
-                "description": "Dataset of the same name already exists",
-            },
-            {"status_code": HTTP_422_UNPROCESSABLE_ENTITY, "description": "Invalid request body"},
-        ]
-    ),
-    # FastAPI cannot generate the request body portion of the OpenAPI schema for
-    # routes that accept multiple request content types, so we have to provide
-    # this part of the schema manually. For context, see
-    # https://github.com/tiangolo/fastapi/discussions/7786 and
-    # https://github.com/tiangolo/fastapi/issues/990
-    openapi_extra={
-        "requestBody": {
-            "content": {
-                "application/json": {
-                    "schema": {
-                        "type": "object",
-                        "required": ["name", "inputs"],
-                        "properties": {
-                            "action": {"type": "string", "enum": ["create", "append"]},
-                            "name": {"type": "string"},
-                            "description": {"type": "string"},
-                            "inputs": {"type": "array", "items": {"type": "object"}},
-                            "outputs": {"type": "array", "items": {"type": "object"}},
-                            "metadata": {"type": "array", "items": {"type": "object"}},
-                        },
-                    }
-                },
-                "multipart/form-data": {
-                    "schema": {
-                        "type": "object",
-                        "required": ["name", "input_keys[]", "output_keys[]", "file"],
-                        "properties": {
-                            "action": {"type": "string", "enum": ["create", "append"]},
-                            "name": {"type": "string"},
-                            "description": {"type": "string"},
-                            "input_keys[]": {
-                                "type": "array",
-                                "items": {"type": "string"},
-                                "uniqueItems": True,
-                            },
-                            "output_keys[]": {
-                                "type": "array",
-                                "items": {"type": "string"},
-                                "uniqueItems": True,
-                            },
-                            "metadata_keys[]": {
-                                "type": "array",
-                                "items": {"type": "string"},
-                                "uniqueItems": True,
-                            },
-                            "file": {"type": "string", "format": "binary"},
-                        },
-                    }
-                },
+                "version_id": str(GlobalID(DatasetVersion.__name__, str(version.id))),
+                "description": version.description,
+                "metadata": version.metadata_,
+                "created_at": version.created_at.isoformat(),
             }
-        },
-    },
-)
-async def upload_dataset(
-    request: Request,
-    sync: bool = Query(
-        default=False,
-        description="If true, fulfill request synchronously and return JSON containing dataset_id.",
-    ),
-) -> Optional[UploadDatasetResponseBody]:
+            async for version in await session.stream_scalars(stmt)
+        ]
+    next_cursor = data.pop()["version_id"] if len(data) == limit + 1 else None
+    return JSONResponse(content={"next_cursor": next_cursor, "data": data})
+
+
+async def post_datasets_upload(request: Request) -> Response:
+    """
+    summary: Upload dataset as either JSON or file (CSV or PyArrow)
+    operationId: uploadDataset
+    tags:
+      - datasets
+    parameters:
+      - in: query
+        name: sync
+        description: If true, fulfill request synchronously and return JSON containing dataset_id
+        schema:
+          type: boolean
+    requestBody:
+      content:
+        application/json:
+          schema:
+            type: object
+            required:
+              - name
+              - inputs
+            properties:
+              action:
+                type: string
+                enum: [create, append]
+              name:
+                type: string
+              description:
+                type: string
+              inputs:
+                type: array
+                items:
+                  type: object
+              outputs:
+                type: array
+                items:
+                  type: object
+              metadata:
+                type: array
+                items:
+                  type: object
+        multipart/form-data:
+          schema:
+            type: object
+            required:
+              - name
+              - input_keys[]
+              - output_keys[]
+              - file
+            properties:
+              action:
+                type: string
+                enum: [create, append]
+              name:
+                type: string
+              description:
+                type: string
+              input_keys[]:
+                type: array
+                items:
+                  type: string
+                uniqueItems: true
+              output_keys[]:
+                type: array
+                items:
+                  type: string
+                uniqueItems: true
+              metadata_keys[]:
+                type: array
+                items:
+                  type: string
+                uniqueItems: true
+              file:
+                type: string
+                format: binary
+    responses:
+      200:
+        description: Success
+      403:
+        description: Forbidden
+      409:
+        description: Dataset of the same name already exists
+      422:
+        description: Request body is invalid
+    """
     request_content_type = request.headers["content-type"]
     examples: Union[Examples, Awaitable[Examples]]
     if request_content_type.startswith("application/json"):
@@ -412,15 +500,15 @@ async def upload_dataset(
                 _process_json, await request.json()
             )
         except ValueError as e:
-            raise HTTPException(
-                detail=str(e),
+            return Response(
+                content=str(e),
                 status_code=HTTP_422_UNPROCESSABLE_ENTITY,
             )
         if action is DatasetAction.CREATE:
             async with request.app.state.db() as session:
                 if await _check_table_exists(session, name):
-                    raise HTTPException(
-                        detail=f"Dataset with the same name already exists: {name=}",
+                    return Response(
+                        content=f"Dataset with the same name already exists: {name=}",
                         status_code=HTTP_409_CONFLICT,
                     )
     elif request_content_type.startswith("multipart/form-data"):
@@ -436,15 +524,15 @@ async def upload_dataset(
                     file,
                 ) = await _parse_form_data(form)
             except ValueError as e:
-                raise HTTPException(
-                    detail=str(e),
+                return Response(
+                    content=str(e),
                     status_code=HTTP_422_UNPROCESSABLE_ENTITY,
                 )
             if action is DatasetAction.CREATE:
                 async with request.app.state.db() as session:
                     if await _check_table_exists(session, name):
-                        raise HTTPException(
-                            detail=f"Dataset with the same name already exists: {name=}",
+                        return Response(
+                            content=f"Dataset with the same name already exists: {name=}",
                             status_code=HTTP_409_CONFLICT,
                         )
             content = await file.read()
@@ -460,13 +548,13 @@ async def upload_dataset(
             else:
                 assert_never(file_content_type)
         except ValueError as e:
-            raise HTTPException(
-                detail=str(e),
+            return Response(
+                content=str(e),
                 status_code=HTTP_422_UNPROCESSABLE_ENTITY,
             )
     else:
-        raise HTTPException(
-            detail="Invalid request Content-Type",
+        return Response(
+            content=str("Invalid request Content-Type"),
             status_code=HTTP_422_UNPROCESSABLE_ENTITY,
         )
     operation = cast(
@@ -479,17 +567,19 @@ async def upload_dataset(
             description=description,
         ),
     )
-    if sync:
+    if request.query_params.get("sync") == "true":
         async with request.app.state.db() as session:
             dataset_id = (await operation(session)).dataset_id
-        return UploadDatasetResponseBody(data=UploadDatasetData(dataset_id=str(dataset_id)))
+        return JSONResponse(
+            content={"data": {"dataset_id": str(GlobalID(Dataset.__name__, str(dataset_id)))}}
+        )
     try:
         request.state.enqueue_operation(operation)
     except QueueFull:
         if isinstance(examples, Coroutine):
             examples.close()
-        raise HTTPException(detail="Too many requests.", status_code=HTTP_429_TOO_MANY_REQUESTS)
-    return None
+        return Response(status_code=HTTP_429_TOO_MANY_REQUESTS)
+    return Response()
 
 
 class FileContentType(Enum):
@@ -667,125 +757,151 @@ async def _parse_form_data(
     )
 
 
-# including the dataset examples router here ensures the dataset example routes
-# are included in a natural order in the openapi schema and the swagger ui
-#
-# todo: move the dataset examples routes here and remove the dataset_examples
-# sub-module
-router.include_router(dataset_examples_router)
-
-
-@router.get(
-    "/datasets/{id}/csv",
-    operation_id="getDatasetCsv",
-    summary="Download dataset examples as CSV file",
-    response_class=StreamingResponse,
-    status_code=HTTP_200_OK,
-    responses={
-        **add_errors_to_responses([HTTP_422_UNPROCESSABLE_ENTITY]),
-        **add_text_csv_content_to_responses(HTTP_200_OK),
-    },
-)
-async def get_dataset_csv(
-    request: Request,
-    response: Response,
-    id: str = Path(description="The ID of the dataset"),
-    version_id: Optional[str] = Query(
-        default=None,
-        description=(
-            "The ID of the dataset version " "(if omitted, returns data from the latest version)"
-        ),
-    ),
-) -> Response:
+async def get_dataset_csv(request: Request) -> Response:
+    """
+    summary: Download dataset examples as CSV text file
+    operationId: getDatasetCsv
+    tags:
+      - datasets
+    parameters:
+      - in: path
+        name: id
+        required: true
+        schema:
+          type: string
+        description: Dataset ID
+      - in: query
+        name: version_id
+        schema:
+          type: string
+        description: Dataset version ID. If omitted, returns the latest version.
+    responses:
+      200:
+        description: Success
+        content:
+          text/csv:
+            schema:
+              type: string
+              contentMediaType: text/csv
+              contentEncoding: gzip
+      403:
+        description: Forbidden
+      404:
+        description: Dataset does not exist.
+      422:
+        description: Dataset ID or version ID is invalid.
+    """
     try:
-        async with request.app.state.db() as session:
-            dataset_name, examples = await _get_db_examples(
-                session=session, id=id, version_id=version_id
-            )
+        dataset_name, examples = await _get_db_examples(request)
     except ValueError as e:
-        raise HTTPException(detail=str(e), status_code=HTTP_422_UNPROCESSABLE_ENTITY)
+        return Response(content=str(e), status_code=HTTP_422_UNPROCESSABLE_ENTITY)
     content = await run_in_threadpool(_get_content_csv, examples)
     return Response(
         content=content,
         headers={
             "content-disposition": f'attachment; filename="{dataset_name}.csv"',
             "content-type": "text/csv",
+            "content-encoding": "gzip",
         },
     )
 
 
-@router.get(
-    "/datasets/{id}/jsonl/openai_ft",
-    operation_id="getDatasetJSONLOpenAIFineTuning",
-    summary="Download dataset examples as OpenAI fine-tuning JSONL file",
-    response_class=PlainTextResponse,
-    responses=add_errors_to_responses(
-        [
-            {
-                "status_code": HTTP_422_UNPROCESSABLE_ENTITY,
-                "description": "Invalid dataset or version ID",
-            }
-        ]
-    ),
-)
-async def get_dataset_jsonl_openai_ft(
-    request: Request,
-    response: Response,
-    id: str = Path(description="The ID of the dataset"),
-    version_id: Optional[str] = Query(
-        default=None,
-        description=(
-            "The ID of the dataset version " "(if omitted, returns data from the latest version)"
-        ),
-    ),
-) -> bytes:
+async def get_dataset_jsonl_openai_ft(request: Request) -> Response:
+    """
+    summary: Download dataset examples as OpenAI Fine-Tuning JSONL file
+    operationId: getDatasetJSONLOpenAIFineTuning
+    tags:
+      - datasets
+    parameters:
+      - in: path
+        name: id
+        required: true
+        schema:
+          type: string
+        description: Dataset ID
+      - in: query
+        name: version_id
+        schema:
+          type: string
+        description: Dataset version ID. If omitted, returns the latest version.
+    responses:
+      200:
+        description: Success
+        content:
+          text/plain:
+            schema:
+              type: string
+              contentMediaType: text/plain
+              contentEncoding: gzip
+      403:
+        description: Forbidden
+      404:
+        description: Dataset does not exist.
+      422:
+        description: Dataset ID or version ID is invalid.
+    """
     try:
-        async with request.app.state.db() as session:
-            dataset_name, examples = await _get_db_examples(
-                session=session, id=id, version_id=version_id
-            )
+        dataset_name, examples = await _get_db_examples(request)
     except ValueError as e:
-        raise HTTPException(detail=str(e), status_code=HTTP_422_UNPROCESSABLE_ENTITY)
+        return Response(content=str(e), status_code=HTTP_422_UNPROCESSABLE_ENTITY)
     content = await run_in_threadpool(_get_content_jsonl_openai_ft, examples)
-    response.headers["content-disposition"] = f'attachment; filename="{dataset_name}.jsonl"'
-    return content
+    return Response(
+        content=content,
+        headers={
+            "content-disposition": f'attachment; filename="{dataset_name}.jsonl"',
+            "content-type": "text/plain",
+            "content-encoding": "gzip",
+        },
+    )
 
 
-@router.get(
-    "/datasets/{id}/jsonl/openai_evals",
-    operation_id="getDatasetJSONLOpenAIEvals",
-    summary="Download dataset examples as OpenAI evals JSONL file",
-    response_class=PlainTextResponse,
-    responses=add_errors_to_responses(
-        [
-            {
-                "status_code": HTTP_422_UNPROCESSABLE_ENTITY,
-                "description": "Invalid dataset or version ID",
-            }
-        ]
-    ),
-)
-async def get_dataset_jsonl_openai_evals(
-    request: Request,
-    response: Response,
-    id: str = Path(description="The ID of the dataset"),
-    version_id: Optional[str] = Query(
-        default=None,
-        description=(
-            "The ID of the dataset version " "(if omitted, returns data from the latest version)"
-        ),
-    ),
-) -> bytes:
+async def get_dataset_jsonl_openai_evals(request: Request) -> Response:
+    """
+    summary: Download dataset examples as OpenAI Evals JSONL file
+    operationId: getDatasetJSONLOpenAIEvals
+    tags:
+      - datasets
+    parameters:
+      - in: path
+        name: id
+        required: true
+        schema:
+          type: string
+        description: Dataset ID
+      - in: query
+        name: version_id
+        schema:
+          type: string
+        description: Dataset version ID. If omitted, returns the latest version.
+    responses:
+      200:
+        description: Success
+        content:
+          text/plain:
+            schema:
+              type: string
+              contentMediaType: text/plain
+              contentEncoding: gzip
+      403:
+        description: Forbidden
+      404:
+        description: Dataset does not exist.
+      422:
+        description: Dataset ID or version ID is invalid.
+    """
     try:
-        async with request.app.state.db() as session:
-            dataset_name, examples = await _get_db_examples(
-                session=session, id=id, version_id=version_id
-            )
+        dataset_name, examples = await _get_db_examples(request)
     except ValueError as e:
-        raise HTTPException(detail=str(e), status_code=HTTP_422_UNPROCESSABLE_ENTITY)
+        return Response(content=str(e), status_code=HTTP_422_UNPROCESSABLE_ENTITY)
     content = await run_in_threadpool(_get_content_jsonl_openai_evals, examples)
-    response.headers["content-disposition"] = f'attachment; filename="{dataset_name}.jsonl"'
-    return content
+    return Response(
+        content=content,
+        headers={
+            "content-disposition": f'attachment; filename="{dataset_name}.jsonl"',
+            "content-type": "text/plain",
+            "content-encoding": "gzip",
+        },
+    )
 
 
 def _get_content_csv(examples: List[models.DatasetExampleRevision]) -> bytes:
@@ -801,7 +917,7 @@ def _get_content_csv(examples: List[models.DatasetExampleRevision]) -> bytes:
         }
         for ex in examples
     ]
-    return str(pd.DataFrame.from_records(records).to_csv(index=False)).encode()
+    return gzip.compress(pd.DataFrame.from_records(records).to_csv(index=False).encode())
 
 
 def _get_content_jsonl_openai_ft(examples: List[models.DatasetExampleRevision]) -> bytes:
@@ -822,7 +938,7 @@ def _get_content_jsonl_openai_ft(examples: List[models.DatasetExampleRevision])
             ).encode()
         )
     records.seek(0)
-    return records.read()
+    return gzip.compress(records.read())
 
 
 def _get_content_jsonl_openai_evals(examples: List[models.DatasetExampleRevision]) -> bytes:
@@ -849,17 +965,18 @@ def _get_content_jsonl_openai_evals(examples: List[models.DatasetExampleRevision
             ).encode()
         )
     records.seek(0)
-    return records.read()
+    return gzip.compress(records.read())
 
 
-async def _get_db_examples(
-    *, session: Any, id: str, version_id: Optional[str]
-) -> Tuple[str, List[models.DatasetExampleRevision]]:
-    dataset_id = from_global_id_with_expected_type(GlobalID.from_id(id), DATASET_NODE_NAME)
+async def _get_db_examples(request: Request) -> Tuple[str, List[models.DatasetExampleRevision]]:
+    if not (id_ := request.path_params.get("id")):
+        raise ValueError("Missing Dataset ID")
+    dataset_id = from_global_id_with_expected_type(GlobalID.from_id(id_), Dataset.__name__)
     dataset_version_id: Optional[int] = None
-    if version_id:
+    if version_id := request.query_params.get("version_id"):
         dataset_version_id = from_global_id_with_expected_type(
-            GlobalID.from_id(version_id), DATASET_VERSION_NODE_NAME
+            GlobalID.from_id(version_id),
+            DatasetVersion.__name__,
         )
     latest_version = (
         select(
@@ -892,12 +1009,13 @@ async def _get_db_examples(
         .where(models.DatasetExampleRevision.revision_kind != "DELETE")
         .order_by(models.DatasetExampleRevision.dataset_example_id)
     )
-    dataset_name: Optional[str] = await session.scalar(
-        select(models.Dataset.name).where(models.Dataset.id == dataset_id)
-    )
-    if not dataset_name:
-        raise ValueError("Dataset does not exist.")
-    examples = [r async for r in await session.stream_scalars(stmt)]
+    async with request.app.state.db() as session:
+        dataset_name: Optional[str] = await session.scalar(
+            select(models.Dataset.name).where(models.Dataset.id == dataset_id)
+        )
+        if not dataset_name:
+            raise ValueError("Dataset does not exist.")
+        examples = [r async for r in await session.stream_scalars(stmt)]
     return dataset_name, examples