abstractvision 0.1.0__py3-none-any.whl → 0.2.1__py3-none-any.whl

abstractvision/assets/vision_model_capabilities.json

@@ -0,0 +1,406 @@
+{
+  "schema_version": "1.0",
+  "tasks": {
+    "text_to_image": {
+      "description": "Generate an image from text (prompt + optional negative prompt)."
+    },
+    "image_to_image": {
+      "description": "Edit/transform an input image using a text prompt (optionally with a mask)."
+    },
+    "multi_view_image": {
+      "description": "Generate multiple consistent views/angles for a concept (optionally conditioned on a reference image)."
+    },
+    "text_to_video": {
+      "description": "Generate a video from text (prompt + optional negative prompt)."
+    },
+    "image_to_video": {
+      "description": "Generate a video conditioned on an image (optionally with text guidance)."
+    }
+  },
+  "models": {
+    "Qwen/Qwen-Image-2512": {
+      "provider": "huggingface",
+      "license": "apache-2.0",
+      "notes": "Primary text-to-image model (Diffusers: QwenImagePipeline).",
+      "tasks": {
+        "text_to_image": {
+          "inputs": ["text"],
+          "outputs": ["image"],
+          "params": {
+            "prompt": {"required": true},
+            "negative_prompt": {"required": false},
+            "width": {"required": false},
+            "height": {"required": false},
+            "steps": {"required": false},
+            "guidance_scale": {"required": false},
+            "seed": {"required": false}
+          }
+        }
+      }
+    },
+    "Qwen/Qwen-Image": {
+      "provider": "huggingface",
+      "license": "apache-2.0",
+      "notes": "Older Qwen Image release (Diffusers: QwenImagePipeline).",
+      "tasks": {
+        "text_to_image": {
+          "inputs": ["text"],
+          "outputs": ["image"],
+          "params": {
+            "prompt": {"required": true},
+            "negative_prompt": {"required": false},
+            "width": {"required": false},
+            "height": {"required": false},
+            "steps": {"required": false},
+            "guidance_scale": {"required": false},
+            "seed": {"required": false}
+          }
+        }
+      }
+    },
+    "black-forest-labs/FLUX.2-klein-4B": {
+      "provider": "huggingface",
+      "license": "apache-2.0",
+      "notes": "FLUX 2 klein (4B, open) (Diffusers: Flux2KleinPipeline). Requires Diffusers from source (diffusers@main) today; runtime stays offline-only. Supports generation + editing pipelines.",
+      "tasks": {
+        "text_to_image": {
+          "inputs": ["text"],
+          "outputs": ["image"],
+          "params": {
+            "prompt": {"required": true},
+            "negative_prompt": {"required": false},
+            "width": {"required": false},
+            "height": {"required": false},
+            "steps": {"required": false},
+            "guidance_scale": {"required": false},
+            "seed": {"required": false}
+          }
+        },
+        "image_to_image": {
+          "inputs": ["image", "text"],
+          "outputs": ["image"],
+          "params": {
+            "prompt": {"required": true},
+            "image": {"required": true},
+            "mask": {"required": false},
+            "negative_prompt": {"required": false},
+            "steps": {"required": false},
+            "guidance_scale": {"required": false},
+            "seed": {"required": false}
+          }
+        }
+      }
+    },
+    "black-forest-labs/FLUX.2-dev": {
+      "provider": "huggingface",
+      "license": "flux-dev-non-commercial-license",
+      "notes": "FLUX 2 dev (gated on HF; non-commercial) (Diffusers: Flux2Pipeline). Supports generation + editing pipelines.",
+      "tasks": {
+        "text_to_image": {
+          "inputs": ["text"],
+          "outputs": ["image"],
+          "params": {
+            "prompt": {"required": true},
+            "negative_prompt": {"required": false},
+            "width": {"required": false},
+            "height": {"required": false},
+            "steps": {"required": false},
+            "guidance_scale": {"required": false},
+            "seed": {"required": false}
+          }
+        },
+        "image_to_image": {
+          "inputs": ["image", "text"],
+          "outputs": ["image"],
+          "params": {
+            "prompt": {"required": true},
+            "image": {"required": true},
+            "mask": {"required": false},
+            "negative_prompt": {"required": false},
+            "steps": {"required": false},
+            "guidance_scale": {"required": false},
+            "seed": {"required": false}
+          }
+        }
+      }
+    },
+    "runwayml/stable-diffusion-v1-5": {
+      "provider": "huggingface",
+      "license": "creativeml-openrail-m",
+      "notes": "Stable Diffusion 1.5 (Diffusers: StableDiffusionPipeline / StableDiffusionImg2ImgPipeline / StableDiffusionInpaintPipeline).",
+      "tasks": {
+        "text_to_image": {
+          "inputs": ["text"],
+          "outputs": ["image"],
+          "params": {
+            "prompt": {"required": true},
+            "negative_prompt": {"required": false},
+            "width": {"required": false},
+            "height": {"required": false},
+            "steps": {"required": false},
+            "guidance_scale": {"required": false},
+            "seed": {"required": false}
+          }
+        },
+        "image_to_image": {
+          "inputs": ["image", "text"],
+          "outputs": ["image"],
+          "params": {
+            "prompt": {"required": true},
+            "image": {"required": true},
+            "mask": {"required": false},
+            "negative_prompt": {"required": false},
+            "steps": {"required": false},
+            "guidance_scale": {"required": false},
+            "seed": {"required": false}
+          }
+        }
+      }
+    },
+    "stabilityai/stable-diffusion-3.5-large-turbo": {
+      "provider": "huggingface",
+      "license": "stabilityai-ai-community",
+      "notes": "Stable Diffusion 3.5 Large Turbo (gated on HF; Stability AI Community License) (Diffusers: StableDiffusion3Pipeline). Optimized for low step counts.",
+      "tasks": {
+        "text_to_image": {
+          "inputs": ["text"],
+          "outputs": ["image"],
+          "params": {
+            "prompt": {"required": true},
+            "negative_prompt": {"required": false},
+            "width": {"required": false},
+            "height": {"required": false},
+            "steps": {"required": false},
+            "guidance_scale": {"required": false},
+            "seed": {"required": false}
+          }
+        }
+      }
+    },
+    "stabilityai/stable-diffusion-3.5-large": {
+      "provider": "huggingface",
+      "license": "stabilityai-ai-community",
+      "notes": "Stable Diffusion 3.5 Large (gated on HF; Stability AI Community License) (Diffusers: StableDiffusion3Pipeline).",
+      "tasks": {
+        "text_to_image": {
+          "inputs": ["text"],
+          "outputs": ["image"],
+          "params": {
+            "prompt": {"required": true},
+            "negative_prompt": {"required": false},
+            "width": {"required": false},
+            "height": {"required": false},
+            "steps": {"required": false},
+            "guidance_scale": {"required": false},
+            "seed": {"required": false}
+          }
+        }
+      }
+    },
+    "stabilityai/stable-diffusion-3.5-medium": {
+      "provider": "huggingface",
+      "license": "stabilityai-ai-community",
+      "notes": "Stable Diffusion 3.5 Medium (gated on HF; Stability AI Community License) (Diffusers: StableDiffusion3Pipeline).",
+      "tasks": {
+        "text_to_image": {
+          "inputs": ["text"],
+          "outputs": ["image"],
+          "params": {
+            "prompt": {"required": true},
+            "negative_prompt": {"required": false},
+            "width": {"required": false},
+            "height": {"required": false},
+            "steps": {"required": false},
+            "guidance_scale": {"required": false},
+            "seed": {"required": false}
+          }
+        }
+      }
+    },
+    "Tongyi-MAI/Z-Image-Turbo": {
+      "provider": "huggingface",
+      "license": "unknown",
+      "notes": "Turbo text-to-image option; backend may clamp/ignore some params.",
+      "tasks": {
+        "text_to_image": {
+          "inputs": ["text"],
+          "outputs": ["image"],
+          "params": {
+            "prompt": {"required": true},
+            "negative_prompt": {"required": false},
+            "width": {"required": false},
+            "height": {"required": false},
+            "steps": {"required": false},
+            "guidance_scale": {"required": false},
+            "seed": {"required": false}
+          }
+        }
+      }
+    },
+    "zai-org/GLM-Image": {
+      "provider": "huggingface",
+      "license": "unknown",
+      "notes": "General image model family (generation + edit in some pipelines). Requires Diffusers from source (diffusers@main) today; runtime stays offline-only.",
+      "tasks": {
+        "text_to_image": {
+          "inputs": ["text"],
+          "outputs": ["image"],
+          "params": {
+            "prompt": {"required": true},
+            "negative_prompt": {"required": false},
+            "width": {"required": false},
+            "height": {"required": false},
+            "steps": {"required": false},
+            "guidance_scale": {"required": false},
+            "seed": {"required": false}
+          }
+        },
+        "image_to_image": {
+          "inputs": ["image", "text"],
+          "outputs": ["image"],
+          "params": {
+            "prompt": {"required": true},
+            "image": {"required": true},
+            "mask": {"required": false},
+            "negative_prompt": {"required": false},
+            "steps": {"required": false},
+            "guidance_scale": {"required": false},
+            "seed": {"required": false}
+          }
+        }
+      }
+    },
+    "Qwen/Qwen-Image-Edit-2511": {
+      "provider": "huggingface",
+      "license": "unknown",
+      "notes": "Primary image edit model (image-to-image).",
+      "tasks": {
+        "image_to_image": {
+          "inputs": ["image", "text"],
+          "outputs": ["image"],
+          "params": {
+            "prompt": {"required": true},
+            "image": {"required": true},
+            "mask": {"required": false},
+            "negative_prompt": {"required": false},
+            "steps": {"required": false},
+            "guidance_scale": {"required": false},
+            "seed": {"required": false}
+          }
+        }
+      }
+    },
+    "fal/Qwen-Image-Edit-2511-Multiple-Angles-LoRA": {
+      "provider": "huggingface",
+      "license": "apache-2.0",
+      "notes": "LoRA adapter used with a compatible base (typically Qwen-Image-Edit-2511) to generate multiple angles.",
+      "tasks": {
+        "multi_view_image": {
+          "inputs": ["text", "image_optional"],
+          "outputs": ["image[]"],
+          "requires": {
+            "base_model_id": "Qwen/Qwen-Image-Edit-2511"
+          },
+          "params": {
+            "prompt": {"required": true},
+            "reference_image": {"required": false},
+            "angles": {"required": false},
+            "negative_prompt": {"required": false},
+            "steps": {"required": false},
+            "guidance_scale": {"required": false},
+            "seed": {"required": false}
+          }
+        }
+      }
+    },
+    "Wan-AI/Wan2.2-T2V-A14B": {
+      "provider": "huggingface",
+      "license": "unknown",
+      "notes": "Text-to-video foundation model.",
+      "tasks": {
+        "text_to_video": {
+          "inputs": ["text"],
+          "outputs": ["video"],
+          "params": {
+            "prompt": {"required": true},
+            "negative_prompt": {"required": false},
+            "width": {"required": false},
+            "height": {"required": false},
+            "fps": {"required": false},
+            "num_frames": {"required": false},
+            "steps": {"required": false},
+            "guidance_scale": {"required": false},
+            "seed": {"required": false}
+          }
+        }
+      }
+    },
+    "tencent/HunyuanVideo-1.5": {
+      "provider": "huggingface",
+      "license": "unknown",
+      "notes": "Text-to-video model.",
+      "tasks": {
+        "text_to_video": {
+          "inputs": ["text"],
+          "outputs": ["video"],
+          "params": {
+            "prompt": {"required": true},
+            "negative_prompt": {"required": false},
+            "width": {"required": false},
+            "height": {"required": false},
+            "fps": {"required": false},
+            "num_frames": {"required": false},
+            "steps": {"required": false},
+            "guidance_scale": {"required": false},
+            "seed": {"required": false}
+          }
+        }
+      }
+    },
+    "genmo/mochi-1-preview": {
+      "provider": "huggingface",
+      "license": "apache-2.0",
+      "notes": "State-of-the-art open text-to-video model (Diffusers).",
+      "tasks": {
+        "text_to_video": {
+          "inputs": ["text"],
+          "outputs": ["video"],
+          "params": {
+            "prompt": {"required": true},
+            "negative_prompt": {"required": false},
+            "width": {"required": false},
+            "height": {"required": false},
+            "fps": {"required": false},
+            "num_frames": {"required": false},
+            "steps": {"required": false},
+            "guidance_scale": {"required": false},
+            "seed": {"required": false}
+          }
+        }
+      }
+    },
+    "Lightricks/LTX-2": {
+      "provider": "huggingface",
+      "license": "ltx-2-community-license-agreement",
+      "notes": "Image-to-video model family; parameters have divisibility constraints (implementation detail).",
+      "tasks": {
+        "image_to_video": {
+          "inputs": ["image", "text_optional"],
+          "outputs": ["video"],
+          "params": {
+            "image": {"required": true},
+            "prompt": {"required": false},
+            "negative_prompt": {"required": false},
+            "width": {"required": false},
+            "height": {"required": false},
+            "fps": {"required": false},
+            "num_frames": {"required": false},
+            "steps": {"required": false},
+            "guidance_scale": {"required": false},
+            "seed": {"required": false}
+          }
+        }
+      }
+    }
+  }
+}

abstractvision/backends/__init__.py

@@ -0,0 +1,43 @@
+"""Backend exports.
+
+Important: this package must stay import-light.
+
+Some backends are intentionally heavy (Torch/Diffusers). Import them lazily so
+`import abstractvision` (and AbstractCore plugin discovery) does not pull GPU
+stacks unless the caller explicitly selects a local backend.
+"""
+
+from .base_backend import VisionBackend
+
+__all__ = [
+    "VisionBackend",
+    "OpenAICompatibleBackendConfig",
+    "OpenAICompatibleVisionBackend",
+    "HuggingFaceDiffusersBackendConfig",
+    "HuggingFaceDiffusersVisionBackend",
+    "StableDiffusionCppBackendConfig",
+    "StableDiffusionCppVisionBackend",
+]
+
+
+def __getattr__(name: str):
+    if name in {"OpenAICompatibleBackendConfig", "OpenAICompatibleVisionBackend"}:
+        from .openai_compatible import OpenAICompatibleBackendConfig, OpenAICompatibleVisionBackend
+
+        return OpenAICompatibleBackendConfig if name == "OpenAICompatibleBackendConfig" else OpenAICompatibleVisionBackend
+
+    if name in {"StableDiffusionCppBackendConfig", "StableDiffusionCppVisionBackend"}:
+        from .stable_diffusion_cpp import StableDiffusionCppBackendConfig, StableDiffusionCppVisionBackend
+
+        return StableDiffusionCppBackendConfig if name == "StableDiffusionCppBackendConfig" else StableDiffusionCppVisionBackend
+
+    if name in {"HuggingFaceDiffusersBackendConfig", "HuggingFaceDiffusersVisionBackend"}:
+        from .huggingface_diffusers import HuggingFaceDiffusersBackendConfig, HuggingFaceDiffusersVisionBackend
+
+        return (
+            HuggingFaceDiffusersBackendConfig
+            if name == "HuggingFaceDiffusersBackendConfig"
+            else HuggingFaceDiffusersVisionBackend
+        )
+
+    raise AttributeError(f"module {__name__!r} has no attribute {name!r}")

abstractvision/backends/base_backend.py

@@ -0,0 +1,63 @@
+from __future__ import annotations
+
+from abc import ABC, abstractmethod
+from typing import Callable, Optional
+
+from ..types import (
+    GeneratedAsset,
+    ImageEditRequest,
+    ImageGenerationRequest,
+    ImageToVideoRequest,
+    MultiAngleRequest,
+    VideoGenerationRequest,
+    VisionBackendCapabilities,
+)
+
+
+class VisionBackend(ABC):
+    """Backend interface for generative vision tasks."""
+
+    def generate_image_with_progress(
+        self,
+        request: ImageGenerationRequest,
+        progress_callback: Optional[Callable[[int, Optional[int]], None]] = None,
+    ) -> GeneratedAsset:
+        """Generate an image, optionally reporting progress (best-effort)."""
+        _ = progress_callback
+        return self.generate_image(request)
+
+    def edit_image_with_progress(
+        self,
+        request: ImageEditRequest,
+        progress_callback: Optional[Callable[[int, Optional[int]], None]] = None,
+    ) -> GeneratedAsset:
+        """Edit an image, optionally reporting progress (best-effort)."""
+        _ = progress_callback
+        return self.edit_image(request)
+
+    def get_capabilities(self) -> Optional[VisionBackendCapabilities]:
+        """Return backend-level capability constraints (optional)."""
+        return None
+
+    def preload(self) -> None:
+        """Best-effort: load model weights into memory for faster first inference."""
+        return None
+
+    def unload(self) -> None:
+        """Best-effort: release model weights from memory."""
+        return None
+
+    @abstractmethod
+    def generate_image(self, request: ImageGenerationRequest) -> GeneratedAsset: ...
+
+    @abstractmethod
+    def edit_image(self, request: ImageEditRequest) -> GeneratedAsset: ...
+
+    @abstractmethod
+    def generate_angles(self, request: MultiAngleRequest) -> list[GeneratedAsset]: ...
+
+    @abstractmethod
+    def generate_video(self, request: VideoGenerationRequest) -> GeneratedAsset: ...
+
+    @abstractmethod
+    def image_to_video(self, request: ImageToVideoRequest) -> GeneratedAsset: ...