修改为东南天坐标系

backend_service/venv/lib/python3.13/site-packages/huggingface_hub/inference/_client.py

@@ -160,6 +160,8 @@ class InferenceClient:
             follow the same pattern as `openai.OpenAI` client. Cannot be used if `token` is set. Defaults to None.
     """
 
+    provider: Optional[PROVIDER_OR_POLICY_T]
+
     @validate_hf_hub_args
     def __init__(
         self,
@@ -227,7 +229,7 @@ class InferenceClient:
                 )
 
         # Configure provider
-        self.provider = provider
+        self.provider = provider  # type: ignore[assignment]
 
         self.cookies = cookies
         self.timeout = timeout
@@ -1030,6 +1032,8 @@ class InferenceClient:
         prompt_name: Optional[str] = None,
         truncate: Optional[bool] = None,
         truncation_direction: Optional[Literal["left", "right"]] = None,
+        dimensions: Optional[int] = None,
+        encoding_format: Optional[Literal["float", "base64"]] = None,
         model: Optional[str] = None,
     ) -> "np.ndarray":
         """
@@ -1056,6 +1060,12 @@ class InferenceClient:
                 Only available on server powered by Text-Embedding-Inference.
             truncation_direction (`Literal["left", "right"]`, *optional*):
                 Which side of the input should be truncated when `truncate=True` is passed.
+            dimensions (`int`, *optional*):
+                The number of dimensions the resulting output embeddings should have.
+                Only available on OpenAI-compatible embedding endpoints.
+            encoding_format (`Literal["float", "base64"]`, *optional*):
+                The format of the output embeddings. Either "float" or "base64".
+                Only available on OpenAI-compatible embedding endpoints.
 
         Returns:
             `np.ndarray`: The embedding representing the input text as a float32 numpy array.
@@ -1086,6 +1096,8 @@ class InferenceClient:
                 "prompt_name": prompt_name,
                 "truncate": truncate,
                 "truncation_direction": truncation_direction,
+                "dimensions": dimensions,
+                "encoding_format": encoding_format,
             },
             headers=self.headers,
             model=model_id,

backend_service/venv/lib/python3.13/site-packages/huggingface_hub/inference/_generated/_async_client.py

@@ -151,6 +151,8 @@ class AsyncInferenceClient:
             follow the same pattern as `openai.OpenAI` client. Cannot be used if `token` is set. Defaults to None.
     """
 
+    provider: Optional[PROVIDER_OR_POLICY_T]
+
     @validate_hf_hub_args
     def __init__(
         self,
@@ -218,7 +220,7 @@ class AsyncInferenceClient:
                 )
 
         # Configure provider
-        self.provider = provider
+        self.provider = provider  # type: ignore[assignment]
 
         self.cookies = cookies
         self.timeout = timeout
@@ -1057,6 +1059,8 @@ class AsyncInferenceClient:
         prompt_name: Optional[str] = None,
         truncate: Optional[bool] = None,
         truncation_direction: Optional[Literal["left", "right"]] = None,
+        dimensions: Optional[int] = None,
+        encoding_format: Optional[Literal["float", "base64"]] = None,
         model: Optional[str] = None,
     ) -> "np.ndarray":
         """
@@ -1083,6 +1087,12 @@ class AsyncInferenceClient:
                 Only available on server powered by Text-Embedding-Inference.
             truncation_direction (`Literal["left", "right"]`, *optional*):
                 Which side of the input should be truncated when `truncate=True` is passed.
+            dimensions (`int`, *optional*):
+                The number of dimensions the resulting output embeddings should have.
+                Only available on OpenAI-compatible embedding endpoints.
+            encoding_format (`Literal["float", "base64"]`, *optional*):
+                The format of the output embeddings. Either "float" or "base64".
+                Only available on OpenAI-compatible embedding endpoints.
 
         Returns:
             `np.ndarray`: The embedding representing the input text as a float32 numpy array.
@@ -1114,6 +1124,8 @@ class AsyncInferenceClient:
                 "prompt_name": prompt_name,
                 "truncate": truncate,
                 "truncation_direction": truncation_direction,
+                "dimensions": dimensions,
+                "encoding_format": encoding_format,
             },
             headers=self.headers,
             model=model_id,

backend_service/venv/lib/python3.13/site-packages/huggingface_hub/inference/_generated/types/__init__.py

@@ -77,6 +77,18 @@ from .image_segmentation import (
     ImageSegmentationParameters,
     ImageSegmentationSubtask,
 )
+from .image_text_to_image import (
+    ImageTextToImageInput,
+    ImageTextToImageOutput,
+    ImageTextToImageParameters,
+    ImageTextToImageTargetSize,
+)
+from .image_text_to_video import (
+    ImageTextToVideoInput,
+    ImageTextToVideoOutput,
+    ImageTextToVideoParameters,
+    ImageTextToVideoTargetSize,
+)
 from .image_to_image import ImageToImageInput, ImageToImageOutput, ImageToImageParameters, ImageToImageTargetSize
 from .image_to_text import (
     ImageToTextEarlyStoppingEnum,

backend_service/venv/lib/python3.13/site-packages/huggingface_hub/inference/_generated/types/image_text_to_image.py

@@ -0,0 +1,67 @@
+# Inference code generated from the JSON schema spec in @huggingface/tasks.
+#
+# See:
+#   - script: https://github.com/huggingface/huggingface.js/blob/main/packages/tasks/scripts/inference-codegen.ts
+#   - specs:  https://github.com/huggingface/huggingface.js/tree/main/packages/tasks/src/tasks.
+from typing import Any, Optional
+
+from .base import BaseInferenceType, dataclass_with_extra
+
+
+@dataclass_with_extra
+class ImageTextToImageTargetSize(BaseInferenceType):
+    """The size in pixels of the output image. This parameter is only supported by some
+    providers and for specific models. It will be ignored when unsupported.
+    """
+
+    height: int
+    width: int
+
+
+@dataclass_with_extra
+class ImageTextToImageParameters(BaseInferenceType):
+    """Additional inference parameters for Image Text To Image"""
+
+    guidance_scale: Optional[float] = None
+    """For diffusion models. A higher guidance scale value encourages the model to generate
+    images closely linked to the text prompt at the expense of lower image quality.
+    """
+    negative_prompt: Optional[str] = None
+    """One prompt to guide what NOT to include in image generation."""
+    num_inference_steps: Optional[int] = None
+    """For diffusion models. The number of denoising steps. More denoising steps usually lead to
+    a higher quality image at the expense of slower inference.
+    """
+    prompt: Optional[str] = None
+    """The text prompt to guide the image generation. Either this or inputs (image) must be
+    provided.
+    """
+    seed: Optional[int] = None
+    """Seed for the random number generator."""
+    target_size: Optional[ImageTextToImageTargetSize] = None
+    """The size in pixels of the output image. This parameter is only supported by some
+    providers and for specific models. It will be ignored when unsupported.
+    """
+
+
+@dataclass_with_extra
+class ImageTextToImageInput(BaseInferenceType):
+    """Inputs for Image Text To Image inference. Either inputs (image) or prompt (in parameters)
+    must be provided, or both.
+    """
+
+    inputs: Optional[str] = None
+    """The input image data as a base64-encoded string. If no `parameters` are provided, you can
+    also provide the image data as a raw bytes payload. Either this or prompt must be
+    provided.
+    """
+    parameters: Optional[ImageTextToImageParameters] = None
+    """Additional inference parameters for Image Text To Image"""
+
+
+@dataclass_with_extra
+class ImageTextToImageOutput(BaseInferenceType):
+    """Outputs of inference for the Image Text To Image task"""
+
+    image: Any
+    """The generated image returned as raw bytes in the payload."""

backend_service/venv/lib/python3.13/site-packages/huggingface_hub/inference/_generated/types/image_text_to_video.py

@@ -0,0 +1,65 @@
+# Inference code generated from the JSON schema spec in @huggingface/tasks.
+#
+# See:
+#   - script: https://github.com/huggingface/huggingface.js/blob/main/packages/tasks/scripts/inference-codegen.ts
+#   - specs:  https://github.com/huggingface/huggingface.js/tree/main/packages/tasks/src/tasks.
+from typing import Any, Optional
+
+from .base import BaseInferenceType, dataclass_with_extra
+
+
+@dataclass_with_extra
+class ImageTextToVideoTargetSize(BaseInferenceType):
+    """The size in pixel of the output video frames."""
+
+    height: int
+    width: int
+
+
+@dataclass_with_extra
+class ImageTextToVideoParameters(BaseInferenceType):
+    """Additional inference parameters for Image Text To Video"""
+
+    guidance_scale: Optional[float] = None
+    """For diffusion models. A higher guidance scale value encourages the model to generate
+    videos closely linked to the text prompt at the expense of lower image quality.
+    """
+    negative_prompt: Optional[str] = None
+    """One prompt to guide what NOT to include in video generation."""
+    num_frames: Optional[float] = None
+    """The num_frames parameter determines how many video frames are generated."""
+    num_inference_steps: Optional[int] = None
+    """The number of denoising steps. More denoising steps usually lead to a higher quality
+    video at the expense of slower inference.
+    """
+    prompt: Optional[str] = None
+    """The text prompt to guide the video generation. Either this or inputs (image) must be
+    provided.
+    """
+    seed: Optional[int] = None
+    """Seed for the random number generator."""
+    target_size: Optional[ImageTextToVideoTargetSize] = None
+    """The size in pixel of the output video frames."""
+
+
+@dataclass_with_extra
+class ImageTextToVideoInput(BaseInferenceType):
+    """Inputs for Image Text To Video inference. Either inputs (image) or prompt (in parameters)
+    must be provided, or both.
+    """
+
+    inputs: Optional[str] = None
+    """The input image data as a base64-encoded string. If no `parameters` are provided, you can
+    also provide the image data as a raw bytes payload. Either this or prompt must be
+    provided.
+    """
+    parameters: Optional[ImageTextToVideoParameters] = None
+    """Additional inference parameters for Image Text To Video"""
+
+
+@dataclass_with_extra
+class ImageTextToVideoOutput(BaseInferenceType):
+    """Outputs of inference for the Image Text To Video task"""
+
+    video: Any
+    """The generated video returned as raw bytes in the payload."""

backend_service/venv/lib/python3.13/site-packages/huggingface_hub/inference/_providers/__init__.py

@@ -56,7 +56,7 @@ from .wavespeed import (
     WavespeedAITextToImageTask,
     WavespeedAITextToVideoTask,
 )
-from .zai_org import ZaiConversationalTask
+from .zai_org import ZaiConversationalTask, ZaiTextToImageTask
 
 
 logger = logging.get_logger(__name__)
@@ -208,6 +208,7 @@ PROVIDERS: dict[PROVIDER_T, dict[str, TaskProviderHelper]] = {
     },
     "zai-org": {
         "conversational": ZaiConversationalTask(),
+        "text-to-image": ZaiTextToImageTask(),
     },
 }
 

backend_service/venv/lib/python3.13/site-packages/huggingface_hub/inference/_providers/ovhcloud.py

@@ -0,0 +1,10 @@
+from huggingface_hub.inference._providers._common import BaseConversationalTask
+
+
+_PROVIDER = "ovhcloud"
+_BASE_URL = "https://oai.endpoints.kepler.ai.cloud.ovh.net"
+
+
+class OVHcloudConversationalTask(BaseConversationalTask):
+    def __init__(self):
+        super().__init__(provider=_PROVIDER, base_url=_BASE_URL)

backend_service/venv/lib/python3.13/site-packages/huggingface_hub/inference/_providers/zai_org.py

@@ -1,13 +1,35 @@
-from typing import Any, Dict
+import time
+from abc import ABC
+from typing import Any, Optional, Union
 
-from huggingface_hub.inference._providers._common import BaseConversationalTask
+from huggingface_hub.hf_api import InferenceProviderMapping
+from huggingface_hub.inference._common import RequestParameters, _as_dict
+from huggingface_hub.inference._providers._common import BaseConversationalTask, TaskProviderHelper, filter_none
+from huggingface_hub.utils import get_session
+
+
+_PROVIDER = "zai-org"
+_BASE_URL = "https://api.z.ai"
+_POLLING_INTERVAL = 5  # seconds
+_MAX_POLL_ATTEMPTS = 60
+
+
+class ZaiTask(TaskProviderHelper, ABC):
+    def __init__(self, task: str):
+        super().__init__(provider=_PROVIDER, base_url=_BASE_URL, task=task)
+
+    def _prepare_headers(self, headers: dict, api_key: str) -> dict[str, Any]:
+        headers = super()._prepare_headers(headers, api_key)
+        headers["Accept-Language"] = "en-US,en"
+        headers["x-source-channel"] = "hugging_face"
+        return headers
 
 
 class ZaiConversationalTask(BaseConversationalTask):
     def __init__(self):
-        super().__init__(provider="zai-org", base_url="https://api.z.ai")
+        super().__init__(provider=_PROVIDER, base_url=_BASE_URL)
 
-    def _prepare_headers(self, headers: Dict, api_key: str) -> Dict[str, Any]:
+    def _prepare_headers(self, headers: dict, api_key: str) -> dict[str, Any]:
         headers = super()._prepare_headers(headers, api_key)
         headers["Accept-Language"] = "en-US,en"
         headers["x-source-channel"] = "hugging_face"
@@ -15,3 +37,91 @@ class ZaiConversationalTask(BaseConversationalTask):
 
     def _prepare_route(self, mapped_model: str, api_key: str) -> str:
         return "/api/paas/v4/chat/completions"
+
+
+class ZaiTextToImageTask(ZaiTask):
+    """Text-to-image task for ZAI provider using async API."""
+
+    def __init__(self):
+        super().__init__("text-to-image")
+
+    def _prepare_route(self, mapped_model: str, api_key: str) -> str:
+        return "/api/paas/v4/async/images/generations"
+
+    def _prepare_payload_as_dict(
+        self, inputs: Any, parameters: dict, provider_mapping_info: InferenceProviderMapping
+    ) -> Optional[dict]:
+        width = parameters.pop("width", None)
+        height = parameters.pop("height", None)
+        size = None
+        if width is not None and height is not None:
+            size = f"{width}x{height}"
+
+        payload: dict[str, Any] = {
+            "model": provider_mapping_info.provider_id,
+            "prompt": inputs,
+        }
+        if size is not None:
+            payload["size"] = size
+
+        payload.update(filter_none(parameters))
+        return payload
+
+    def get_response(
+        self,
+        response: Union[bytes, dict],
+        request_params: Optional[RequestParameters] = None,
+    ) -> Any:
+        """Handle async response by polling for results."""
+        response_dict = _as_dict(response)
+
+        task_id = response_dict.get("id")
+        if task_id is None:
+            raise ValueError("No task_id in response from ZAI API")
+
+        task_status = response_dict.get("task_status")
+        if task_status == "FAIL":
+            raise ValueError(f"ZAI image generation failed for request {task_id}")
+
+        if task_status == "PROCESSING" and request_params is not None:
+            return self._poll_for_result(task_id, request_params)
+
+        return self._extract_image(response_dict)
+
+    def _poll_for_result(self, task_id: str, request_params: RequestParameters) -> bytes:
+        """Poll the async-result endpoint until completion."""
+        session = get_session()
+        base_url = request_params.url.rsplit("/api/paas/v4/async/images/generations", 1)[0]
+        poll_url = f"{base_url}/api/paas/v4/async-result/{task_id}"
+
+        for _ in range(_MAX_POLL_ATTEMPTS):
+            poll_response = session.get(poll_url, headers=request_params.headers)
+            poll_response.raise_for_status()
+            result = poll_response.json()
+
+            task_status = result.get("task_status")
+            if task_status == "SUCCESS":
+                return self._extract_image(result)
+            elif task_status == "FAIL":
+                raise ValueError(f"Zai text-to-image generation failed for request {task_id}")
+
+            time.sleep(_POLLING_INTERVAL)
+
+        raise ValueError(
+            f"Timed out while waiting for the result from Zai API - aborting after {_MAX_POLL_ATTEMPTS} attempts"
+        )
+
+    def _extract_image(self, result: dict) -> bytes:
+        """Extract and download the image from the result."""
+        image_result = result.get("image_result")
+        if not image_result or not isinstance(image_result, list) or len(image_result) == 0:
+            raise ValueError("No image_result in response from ZAI API")
+
+        image_url = image_result[0].get("url")
+        if not image_url:
+            raise ValueError("No image URL in response from ZAI API")
+
+        session = get_session()
+        image_response = session.get(image_url)
+        image_response.raise_for_status()
+        return image_response.content