hokusai/bulkgen/providers/models.py

133 lines
3.5 KiB
Python

"""Registry of supported models and their capabilities."""
from __future__ import annotations
from dataclasses import dataclass
from enum import StrEnum
from typing import Literal
class Capability(StrEnum):
"""Capabilities a model may support."""
TEXT_GENERATION = "text generation"
VISION = "vision"
TEXT_TO_IMAGE = "text-to-image"
HIGH_RESOLUTION = "high resolution"
REFERENCE_IMAGES = "reference images"
CONTROL_IMAGES = "control images"
EDGE_DETECTION = "edge detection"
DEPTH_MAP = "depth map"
INPAINTING = "inpainting"
OUTPAINTING = "outpainting"
@dataclass(frozen=True)
class ModelInfo:
"""Describes a supported model and its capabilities."""
name: str
provider: str
type: Literal["text", "image"]
capabilities: list[Capability]
TEXT_MODELS: list[ModelInfo] = [
ModelInfo(
name="mistral-large-latest",
provider="Mistral",
type="text",
capabilities=[Capability.TEXT_GENERATION],
),
ModelInfo(
name="mistral-small-latest",
provider="Mistral",
type="text",
capabilities=[Capability.TEXT_GENERATION],
),
ModelInfo(
name="pixtral-large-latest",
provider="Mistral",
type="text",
capabilities=[Capability.TEXT_GENERATION, Capability.VISION],
),
ModelInfo(
name="pixtral-12b-latest",
provider="Mistral",
type="text",
capabilities=[Capability.TEXT_GENERATION, Capability.VISION],
),
]
IMAGE_MODELS: list[ModelInfo] = [
ModelInfo(
name="flux-dev",
provider="BlackForestLabs",
type="image",
capabilities=[Capability.TEXT_TO_IMAGE],
),
ModelInfo(
name="flux-pro",
provider="BlackForestLabs",
type="image",
capabilities=[Capability.TEXT_TO_IMAGE],
),
ModelInfo(
name="flux-pro-1.1",
provider="BlackForestLabs",
type="image",
capabilities=[Capability.TEXT_TO_IMAGE],
),
ModelInfo(
name="flux-pro-1.1-ultra",
provider="BlackForestLabs",
type="image",
capabilities=[Capability.TEXT_TO_IMAGE, Capability.HIGH_RESOLUTION],
),
ModelInfo(
name="flux-2-pro",
provider="BlackForestLabs",
type="image",
capabilities=[Capability.TEXT_TO_IMAGE, Capability.REFERENCE_IMAGES],
),
ModelInfo(
name="flux-kontext-pro",
provider="BlackForestLabs",
type="image",
capabilities=[Capability.TEXT_TO_IMAGE, Capability.REFERENCE_IMAGES],
),
ModelInfo(
name="flux-pro-1.0-canny",
provider="BlackForestLabs",
type="image",
capabilities=[
Capability.TEXT_TO_IMAGE,
Capability.CONTROL_IMAGES,
Capability.EDGE_DETECTION,
],
),
ModelInfo(
name="flux-pro-1.0-depth",
provider="BlackForestLabs",
type="image",
capabilities=[
Capability.TEXT_TO_IMAGE,
Capability.CONTROL_IMAGES,
Capability.DEPTH_MAP,
],
),
ModelInfo(
name="flux-pro-1.0-fill",
provider="BlackForestLabs",
type="image",
capabilities=[Capability.TEXT_TO_IMAGE, Capability.INPAINTING],
),
ModelInfo(
name="flux-pro-1.0-expand",
provider="BlackForestLabs",
type="image",
capabilities=[Capability.TEXT_TO_IMAGE, Capability.OUTPAINTING],
),
]
ALL_MODELS: list[ModelInfo] = TEXT_MODELS + IMAGE_MODELS