vllm_omni.engine.messages ¶

request_ids: list[str]

type: Literal['abort'] = 'abort'

companion_id: str

companion_prompt_text: object | None

parent_id: str

prompt: EngineCoreRequest

role: str

sampling_params_list: list[OmniSamplingParams]

type: Literal["add_companion_request"] = (
    "add_companion_request"
)

args: tuple[object, ...]

kwargs: dict[str, object]

method: str

rpc_id: str

stage_ids: list[int] | None

timeout: float | None = None

type: Literal['collective_rpc'] = 'collective_rpc'

method: str

results: list[object]

rpc_id: str

stage_ids: list[int]

type: Literal["collective_rpc_result"] = (
    "collective_rpc_result"
)

error: str

fatal: bool = False

request_id: str | None = None

stage_id: int | None = None

type: Literal['error'] = 'error'

engine_outputs: OmniRequestOutput

finished: bool

metrics: StageRequestStats | None = None

replica_id: int | None = None

request_id: str

stage_id: int

stage_submit_ts: float | None = None

type: Literal['output'] = 'output'

replica_id: int

stage_id: int

type: Literal["register_remote_replica"] = (
    "register_remote_replica"
)

type: Literal['shutdown'] = 'shutdown'

metrics: StageRequestStats

replica_id: int | None = None

request_id: str

stage_id: int

stage_submit_ts: float | None = None

type: Literal['stage_metrics'] = 'stage_metrics'

enqueue_ts: float

final_output_stage_ids: list[int] | None = None

final_stage_id: int

original_prompt: EngineCoreRequest | PromptType

output_prompt_text: object | None

preprocess_ms: float

prompt: EngineCoreRequest | PromptType

request_id: str

request_timestamp: float

sampling_params_list: list[OmniSamplingParams]

type: Literal['add_request', 'streaming_update']

input_addr: str

stage_id: int

type: Literal["unregister_remote_replica"] = (
    "unregister_remote_replica"
)