faster_whisper_server/server_models.py

from __future__ import annotations

from typing import TYPE_CHECKING, Literal

from pydantic import BaseModel, ConfigDict, Field

from faster_whisper_server.core import Segment, Transcription, Word, segments_to_text

if TYPE_CHECKING:
    from faster_whisper.transcribe import TranscriptionInfo


# https://platform.openai.com/docs/api-reference/audio/json-object
class TranscriptionJsonResponse(BaseModel):
    text: str

    @classmethod
    def from_segments(cls, segments: list[Segment]) -> TranscriptionJsonResponse:
        return cls(text=segments_to_text(segments))

    @classmethod
    def from_transcription(cls, transcription: Transcription) -> TranscriptionJsonResponse:
        return cls(text=transcription.text)


# https://platform.openai.com/docs/api-reference/audio/verbose-json-object
class TranscriptionVerboseJsonResponse(BaseModel):
    task: str = "transcribe"
    language: str
    duration: float
    text: str
    words: list[Word]
    segments: list[Segment]

    @classmethod
    def from_segment(cls, segment: Segment, transcription_info: TranscriptionInfo) -> TranscriptionVerboseJsonResponse:
        return cls(
            language=transcription_info.language,
            duration=segment.end - segment.start,
            text=segment.text,
            words=(segment.words if isinstance(segment.words, list) else []),
            segments=[segment],
        )

    @classmethod
    def from_segments(
        cls, segments: list[Segment], transcription_info: TranscriptionInfo
    ) -> TranscriptionVerboseJsonResponse:
        return cls(
            language=transcription_info.language,
            duration=transcription_info.duration,
            text=segments_to_text(segments),
            segments=segments,
            words=Word.from_segments(segments),
        )

    @classmethod
    def from_transcription(cls, transcription: Transcription) -> TranscriptionVerboseJsonResponse:
        return cls(
            language="english",  # FIX: hardcoded
            duration=transcription.duration,
            text=transcription.text,
            words=transcription.words,
            segments=[],  # FIX: hardcoded
        )


class ModelListResponse(BaseModel):
    data: list[ModelObject]
    object: Literal["list"] = "list"


class ModelObject(BaseModel):
    id: str
    """The model identifier, which can be referenced in the API endpoints."""
    created: int
    """The Unix timestamp (in seconds) when the model was created."""
    object_: Literal["model"] = Field(serialization_alias="object")
    """The object type, which is always "model"."""
    owned_by: str
    """The organization that owns the model."""
    language: list[str] = Field(default_factory=list)
    """List of ISO 639-3 supported by the model. It's possible that the list will be empty. This field is not a part of the OpenAI API spec and is added for convenience."""  # noqa: E501

    model_config = ConfigDict(
        populate_by_name=True,
        json_schema_extra={
            "examples": [
                {
                    "id": "Systran/faster-whisper-large-v3",
                    "created": 1700732060,
                    "object": "model",
                    "owned_by": "Systran",
                },
                {
                    "id": "Systran/faster-distil-whisper-large-v3",
                    "created": 1711378296,
                    "object": "model",
                    "owned_by": "Systran",
                },
                {
                    "id": "bofenghuang/whisper-large-v2-cv11-french-ct2",
                    "created": 1687968011,
                    "object": "model",
                    "owned_by": "bofenghuang",
                },
            ]
        },
    )