feat(api): add inputs array to UserInputBlock proto

Author: stainless-app[bot]

.stats.yml

@@ -1,4 +1,4 @@
 configured_endpoints: 170
-openapi_spec_url: https://storage.googleapis.com/stainless-sdk-openapi-specs/gitpod%2Fgitpod-eeec67988ceb123d93514d894e04b0fa00dbf0c4678b95baa80868c103aaa0c1.yml
-openapi_spec_hash: 3bf178a4e70c15f12fdc23531fe81aea
+openapi_spec_url: https://storage.googleapis.com/stainless-sdk-openapi-specs/gitpod%2Fgitpod-f838b3c6096ed0143c969fcdd6acf63cb619865ef707e1213194d3b82afe10bf.yml
+openapi_spec_hash: 00a77c0d2749b833f8e29c8493b7ea98
 config_hash: d726afb2a92132197e4eae04303e8041

src/gitpod/types/user_input_block_param.py

@@ -2,30 +2,61 @@
 
 from __future__ import annotations
 
-from typing import Union
+from typing import Union, Iterable
 from datetime import datetime
-from typing_extensions import Annotated, TypedDict
+from typing_extensions import Literal, Annotated, TypedDict
 
 from .._types import Base64FileInput
 from .._utils import PropertyInfo
 from .._models import set_pydantic_config
 
-__all__ = ["UserInputBlockParam", "Image", "Text"]
+__all__ = ["UserInputBlockParam", "Image", "Input", "InputImage", "InputText", "Text"]
 
 
 class Image(TypedDict, total=False):
     """
     ImageInput allows sending images to the agent.
-     Media type is inferred from magic bytes by the backend.
+     Client must provide the MIME type; backend validates against magic bytes.
     """
 
     data: Annotated[Union[str, Base64FileInput], PropertyInfo(format="base64")]
-    """Raw image data (max 4MB). Supported formats: PNG, JPEG, WebP."""
+    """Raw image data (max 4MB). Supported formats: PNG, JPEG."""
+
+    mime_type: Annotated[Literal["image/png", "image/jpeg"], PropertyInfo(alias="mimeType")]
 
 
 set_pydantic_config(Image, {"arbitrary_types_allowed": True})
 
 
+class InputImage(TypedDict, total=False):
+    """
+    ImageInput allows sending images to the agent.
+     Client must provide the MIME type; backend validates against magic bytes.
+    """
+
+    data: Annotated[Union[str, Base64FileInput], PropertyInfo(format="base64")]
+    """Raw image data (max 4MB). Supported formats: PNG, JPEG."""
+
+    mime_type: Annotated[Literal["image/png", "image/jpeg"], PropertyInfo(alias="mimeType")]
+
+
+set_pydantic_config(InputImage, {"arbitrary_types_allowed": True})
+
+
+class InputText(TypedDict, total=False):
+    content: str
+
+
+class Input(TypedDict, total=False):
+    image: InputImage
+    """
+    ImageInput allows sending images to the agent. Client must provide the MIME
+    type; backend validates against magic bytes.
+    """
+
+    text: InputText
+
+
 class Text(TypedDict, total=False):
     content: str
 
@@ -38,8 +69,10 @@ class UserInputBlockParam(TypedDict, total=False):
 
     image: Image
     """
-    ImageInput allows sending images to the agent. Media type is inferred from magic
-    bytes by the backend.
+    ImageInput allows sending images to the agent. Client must provide the MIME
+    type; backend validates against magic bytes.
     """
 
+    inputs: Iterable[Input]
+
     text: Text

tests/api_resources/test_agents.py

@@ -361,7 +361,19 @@ def test_method_send_to_execution_with_all_params(self, client: Gitpod) -> None:
             user_input={
                 "id": "id",
                 "created_at": parse_datetime("2019-12-27T18:11:19.117Z"),
-                "image": {"data": "U3RhaW5sZXNzIHJvY2tz"},
+                "image": {
+                    "data": "U3RhaW5sZXNzIHJvY2tz",
+                    "mime_type": "image/png",
+                },
+                "inputs": [
+                    {
+                        "image": {
+                            "data": "U3RhaW5sZXNzIHJvY2tz",
+                            "mime_type": "image/png",
+                        },
+                        "text": {"content": "x"},
+                    }
+                ],
                 "text": {"content": "Generate a report based on the latest logs."},
             },
         )
@@ -874,7 +886,19 @@ async def test_method_send_to_execution_with_all_params(self, async_client: Asyn
             user_input={
                 "id": "id",
                 "created_at": parse_datetime("2019-12-27T18:11:19.117Z"),
-                "image": {"data": "U3RhaW5sZXNzIHJvY2tz"},
+                "image": {
+                    "data": "U3RhaW5sZXNzIHJvY2tz",
+                    "mime_type": "image/png",
+                },
+                "inputs": [
+                    {
+                        "image": {
+                            "data": "U3RhaW5sZXNzIHJvY2tz",
+                            "mime_type": "image/png",
+                        },
+                        "text": {"content": "x"},
+                    }
+                ],
                 "text": {"content": "Generate a report based on the latest logs."},
             },
         )