AI-Talent-Force/dev_caio

1"""2ShortSmith v2 - Configuration Module3 4Centralized configuration for all components including model paths,5thresholds, domain presets, and runtime settings.6"""7 8import os9from dataclasses import dataclass, field10from typing import Dict, Optional11from enum import Enum12 13 14class ContentDomain(Enum):15    """Supported content domains with different hype characteristics."""16    SPORTS = "sports"17    VLOGS = "vlogs"18    MUSIC = "music"19    PODCASTS = "podcasts"20    GAMING = "gaming"21    GENERAL = "general"22 23 24@dataclass25class DomainWeights:26    """Weight configuration for visual vs audio scoring per domain."""27    visual_weight: float28    audio_weight: float29    motion_weight: float = 0.030 31    def __post_init__(self):32        """Normalize weights to sum to 1.0."""33        total = self.visual_weight + self.audio_weight + self.motion_weight34        if total > 0:35            self.visual_weight /= total36            self.audio_weight /= total37            self.motion_weight /= total38 39 40# Domain-specific weight presets41DOMAIN_PRESETS: Dict[ContentDomain, DomainWeights] = {42    ContentDomain.SPORTS: DomainWeights(visual_weight=0.35, audio_weight=0.50, motion_weight=0.15),43    ContentDomain.VLOGS: DomainWeights(visual_weight=0.70, audio_weight=0.20, motion_weight=0.10),44    ContentDomain.MUSIC: DomainWeights(visual_weight=0.40, audio_weight=0.50, motion_weight=0.10),45    ContentDomain.PODCASTS: DomainWeights(visual_weight=0.10, audio_weight=0.85, motion_weight=0.05),46    ContentDomain.GAMING: DomainWeights(visual_weight=0.50, audio_weight=0.35, motion_weight=0.15),47    ContentDomain.GENERAL: DomainWeights(visual_weight=0.50, audio_weight=0.40, motion_weight=0.10),48}49 50 51@dataclass52class ModelConfig:53    """Configuration for AI models."""54    # Visual model (Qwen2-VL)55    visual_model_id: str = "Qwen/Qwen2-VL-2B-Instruct"56    visual_model_quantization: str = "int4"  # Options: "int4", "int8", "none"57    visual_max_frames: int = 3258 59    # Audio model60    audio_model_id: str = "facebook/wav2vec2-base-960h"61    use_advanced_audio: bool = False  # Use Wav2Vec2 instead of just Librosa62 63    # Face recognition (InsightFace)64    face_detection_model: str = "buffalo_l"  # SCRFD model65    face_similarity_threshold: float = 0.466 67    # Body recognition (OSNet)68    body_model_name: str = "osnet_x1_0"69    body_similarity_threshold: float = 0.570 71    # Motion detection (RAFT)72    motion_model: str = "raft-things"73    motion_threshold: float = 5.074 75    # Device settings76    device: str = "cuda"  # Options: "cuda", "cpu", "mps"77 78    def __post_init__(self):79        """Validate and adjust device based on availability."""80        import torch81        if self.device == "cuda" and not torch.cuda.is_available():82            self.device = "cpu"83        elif self.device == "mps" and not (hasattr(torch.backends, "mps") and torch.backends.mps.is_available()):84            self.device = "cpu"85 86 87@dataclass88class ProcessingConfig:89    """Configuration for video processing pipeline."""90    # Sampling settings91    coarse_sample_interval: float = 5.0  # Seconds between frames in first pass92    dense_sample_fps: float = 3.0  # FPS for dense sampling on candidates93    min_motion_for_dense: float = 2.0  # Threshold to trigger dense sampling94 95    # Clip settings96    min_clip_duration: float = 10.0  # Minimum clip length in seconds97    max_clip_duration: float = 20.0  # Maximum clip length in seconds98    default_clip_duration: float = 15.0  # Default clip length99    min_gap_between_clips: float = 30.0  # Minimum gap between clip starts100 101    # Output settings102    default_num_clips: int = 3103    max_num_clips: int = 10104    output_format: str = "mp4"105    output_codec: str = "libx264"106    output_audio_codec: str = "aac"107 108    # Scene detection109    scene_threshold: float = 27.0  # PySceneDetect threshold110 111    # Hype scoring112    hype_threshold: float = 0.3  # Minimum normalized score to consider113    diversity_weight: float = 0.2  # Weight for temporal diversity in ranking114 115    # Performance116    batch_size: int = 8  # Frames per batch for model inference117    max_video_duration: float = 12000.0  118 119    # Temporary files120    temp_dir: Optional[str] = None121    cleanup_temp: bool = True122 123 124@dataclass125class AppConfig:126    """Main application configuration."""127    model: ModelConfig = field(default_factory=ModelConfig)128    processing: ProcessingConfig = field(default_factory=ProcessingConfig)129 130    # Logging131    log_level: str = "INFO"132    log_file: Optional[str] = "shortsmith.log"133    log_to_console: bool = True134 135    # API settings (for future extensibility)136    api_key: Optional[str] = None137 138    # UI settings139    share_gradio: bool = False140    server_port: int = 7860141 142    @classmethod143    def from_env(cls) -> "AppConfig":144        """Create configuration from environment variables."""145        config = cls()146 147        # Override from environment148        if os.environ.get("SHORTSMITH_LOG_LEVEL"):149            config.log_level = os.environ["SHORTSMITH_LOG_LEVEL"]150 151        if os.environ.get("SHORTSMITH_DEVICE"):152            config.model.device = os.environ["SHORTSMITH_DEVICE"]153 154        if os.environ.get("SHORTSMITH_API_KEY"):155            config.api_key = os.environ["SHORTSMITH_API_KEY"]156 157        if os.environ.get("HF_TOKEN"):158            # HuggingFace token for accessing gated models159            pass160 161        return config162 163 164# Global configuration instance165_config: Optional[AppConfig] = None166 167 168def get_config() -> AppConfig:169    """Get the global configuration instance."""170    global _config171    if _config is None:172        _config = AppConfig.from_env()173    return _config174 175 176def set_config(config: AppConfig) -> None:177    """Set the global configuration instance."""178    global _config179    _config = config180 181 182# Export commonly used items183__all__ = [184    "ContentDomain",185    "DomainWeights",186    "DOMAIN_PRESETS",187    "ModelConfig",188    "ProcessingConfig",189    "AppConfig",190    "get_config",191    "set_config",192]193