[Frontend] Create OmniOpenAIServeImage Class and move image api to it

vllm_omni/benchmarks/metrics/metrics.py

@@ -223,8 +223,8 @@ def calculate_metrics(
                 good_completed += 1
 
     if completed == 0:
-        warnings.formatwarning = (
-            lambda msg, category, filename, lineno, line=None: f"{filename}:{lineno}: {category.__name__}: {msg}\n"
+        warnings.formatwarning = lambda msg, category, filename, lineno, line=None: (
+            f"{filename}:{lineno}: {category.__name__}: {msg}\n"
         )
         warnings.warn(
             "All requests failed. This is likely due to a misconfiguration on the benchmark arguments.",

vllm_omni/diffusion/envs.py

@@ -20,7 +20,7 @@
     # used in distributed environment to determine the master address
     "MASTER_ADDR": lambda: os.getenv("MASTER_ADDR", ""),
     # used in distributed environment to manually set the communication port
-    "MASTER_PORT": lambda: (int(os.getenv("MASTER_PORT", "0")) if "MASTER_PORT" in os.environ else None),
+    "MASTER_PORT": lambda: int(os.getenv("MASTER_PORT", "0")) if "MASTER_PORT" in os.environ else None,
     # path to cudatoolkit home directory, under which should be bin, include,
     # and lib directories.
     "CUDA_HOME": lambda: os.environ.get("CUDA_HOME", None),

vllm_omni/entrypoints/openai/diffusion_models.py

@@ -0,0 +1,37 @@
+from vllm.entrypoints.openai.engine.protocol import (
+    ModelCard,
+    ModelList,
+    ModelPermission,
+)
+from vllm.entrypoints.openai.models.protocol import BaseModelPath
+
+
+class DiffusionServingModels:
+    """Minimal OpenAIServingModels implementation for diffusion-only servers.
+
+    vLLM's /v1/models route expects `app.state.openai_serving_models` to expose
+    `show_available_models()`. In pure diffusion mode we don't initialize the
+    full OpenAIServingModels (it depends on LLM-specific processors), so we
+    provide a lightweight fallback.
+    """
+
+    def __init__(self, base_model_paths: list[BaseModelPath]) -> None:
+        self.base_model_paths = base_model_paths
+
+    async def show_available_models(self) -> ModelList:
+        return ModelList(
+            data=[
+                ModelCard(
+                    id=base_model.name,
+                    root=base_model.model_path,
+                    permission=[ModelPermission()],
+                )
+                for base_model in self.base_model_paths
+            ]
+        )
+
+    @property
+    def model_name(self) -> str:
+        if not self.base_model_paths:
+            raise ValueError("No base models are configured; cannot determine model_name.")
+        return self.base_model_paths[0].name

vllm_omni/entrypoints/openai/protocol/__init__.py

@@ -4,6 +4,8 @@
 from vllm_omni.entrypoints.openai.protocol.chat_completion import OmniChatCompletionStreamResponse
 from vllm_omni.entrypoints.openai.protocol.images import (
     ImageData,
+    ImageEditRequest,
+    ImageEditResponse,
     ImageGenerationRequest,
     ImageGenerationResponse,
     ResponseFormat,
@@ -19,6 +21,8 @@
     "ImageData",
     "ImageGenerationRequest",
     "ImageGenerationResponse",
+    "ImageEditRequest",
+    "ImageEditResponse",
     "ResponseFormat",
     "VideoData",
     "VideoGenerationRequest",

vllm_omni/entrypoints/openai/protocol/images.py

@@ -10,7 +10,8 @@
 from enum import Enum
 from typing import Any
 
-from pydantic import BaseModel, Field, field_validator
+from fastapi import UploadFile
+from pydantic import BaseModel, ConfigDict, Field, field_validator
 
 
 class ResponseFormat(str, Enum):
@@ -127,5 +128,92 @@ class ImageGenerationResponse(BaseModel):
 
     created: int = Field(..., description="Unix timestamp of when the generation completed")
     data: list[ImageData] = Field(..., description="Array of generated images")
-    output_format: str = Field(None, description="The output format of the image generation")
-    size: str = Field(None, description="The size of the image generated")
+    output_format: str | None = None
+    size: str | None = None
+
+
+class ImageEditResponse(BaseModel):
+    """
+    OpenAI DALL-E compatible image edit response.
+
+    Returns generated images with metadata.
+    """
+
+    created: int = Field(..., description="Unix timestamp of when the generation completed")
+    data: list[ImageData] = Field(..., description="Array of generated images")
+    output_format: str = Field(..., description="The output format of the image generation")
+    size: str = Field(..., description="The size of the image generated")
+
+
+class ImageEditRequest(BaseModel):
+    image: list[UploadFile] | None = Field(default=None, description="Image file to edit")
+    url: list[str] | None = Field(default=None, description="URL of the image to edit")
+    prompt: str = Field(..., description="Text description of the desired image edit")
+    model: str | None = Field(
+        default=None,
+        description="Model to use (optional, uses server's configured model if omitted)",
+    )
+    n: int = Field(default=1, ge=1, le=10, description="Number of images to generate")
+    size: str | None = Field(
+        default=None,
+        description="Image dimensions in WIDTHxHEIGHT format (e.g., '1024x1024', uses model defaults if omitted)",
+    )
+    output_format: str = Field(
+        default="png", description="The output format of the image generation (e.g., 'png', 'jpeg')"
+    )
+    background: str | None = Field(
+        default=None,
+        description=(
+            "Background color for transparent areas in the edited image (e.g., '#FFFFFF' for white). "
+            "If not specified, transparent areas will remain transparent in formats that support it (e.g., PNG)."
+        ),
+    )
+    response_format: ResponseFormat = Field(default=ResponseFormat.B64_JSON, description="Format of the returned image")
+    user: str | None = Field(default=None, description="User identifier for tracking")
+
+    # vllm-omni extensions for diffusion control
+    negative_prompt: str | None = Field(default=None, description="Text describing what to avoid in the image")
+    num_inference_steps: int | None = Field(
+        default=None,
+        ge=1,
+        le=200,
+        description="Number of diffusion sampling steps (uses model defaults if not specified)",
+    )
+    guidance_scale: float | None = Field(
+        default=None,
+        ge=0.0,
+        le=20.0,
+        description="Classifier-free guidance scale (uses model defaults if not specified)",
+    )
+    true_cfg_scale: float | None = Field(
+        default=None,
+        ge=0.0,
+        le=20.0,
+        description="True CFG scale (model-specific parameter, may be ignored if not supported)",
+    )
+    seed: int | None = Field(default=None, description="Random seed for reproducibility")
+    generator_device: str | None = Field(
+        default=None,
+        description="Device for the seeded torch.Generator (e.g. 'cpu', 'cuda'). Defaults to the runner's device.",
+    )
+    output_compression: int = Field(
+        default=100,
+        ge=0,
+        le=100,
+        description=(
+            "Output compression/quality level for edited images (0-100). "
+            "Higher values preserve higher quality (typically larger files), lower values apply stronger compression. "
+            "Defaults to 100."
+        ),
+    )
+    lora: dict[str, Any] | None = Field(
+        default=None,
+        description=(
+            "Optional LoRA adapter for this request. Expected shape: "
+            "{name/path/scale/int_id}. Field names are flexible "
+            "(e.g. name|lora_name|adapter, path|lora_path|local_path, "
+            "scale|lora_scale, int_id|lora_int_id)."
+        ),
+    )
+
+    model_config = ConfigDict(arbitrary_types_allowed=True)