protocol ¶

Classes¶

fastvideo.entrypoints.openai.protocol.VideoGenerationRequest ¶

Bases: BaseModel

OpenAI/vLLM-Omni-compatible video generation request.

Model-specific parameters belong in extra_params. FastVideo keeps the legacy input_reference and reference_url fields for clients that predate vLLM-Omni's typed reference objects.

Methods:¶

fastvideo.entrypoints.openai.protocol.VideoGenerationRequest.resolve_video_params ¶

resolve_video_params() -> VideoParams

Resolve top-level, nested, and size dimensions like vLLM-Omni.

Source code in fastvideo/entrypoints/openai/protocol.py

def resolve_video_params(self) -> VideoParams:
    """Resolve top-level, nested, and ``size`` dimensions like vLLM-Omni."""
    params = VideoParams(
        width=self.width,
        height=self.height,
        fps=self.fps,
        num_frames=self.num_frames,
    )
    if self.video_params is not None:
        params.width = params.width or self.video_params.width
        params.height = params.height or self.video_params.height
        params.fps = params.fps or self.video_params.fps
        params.num_frames = params.num_frames or self.video_params.num_frames
    if self.size is not None:
        width, height = self.size.split("x", 1)
        params.width, params.height = int(width), int(height)
    if params.fps is None:
        params.fps = DEFAULT_FPS
    if params.num_frames is None and self.seconds is not None:
        params.num_frames = int(self.seconds) * params.fps
    return params

fastvideo.entrypoints.openai.protocol.VideoParams ¶

Bases: BaseModel

Optional vLLM-Omni-compatible video parameter block.

Functions:¶

fastvideo.entrypoints.openai.protocol.generate_request_id ¶

generate_request_id() -> str

Generate a unique request ID

Source code in fastvideo/entrypoints/openai/protocol.py

def generate_request_id() -> str:
    """Generate a unique request ID"""
    return uuid.uuid4().hex