let5see/AI-Video
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat: Add VLMConfig to schema for proper config.yaml VLM support

Browse Source
This commit is contained in:
empty
2026-01-07 09:38:45 +08:00
parent bc077475c6
commit f19804facb
3 changed files with 19 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
pixelle_video/config/__init__.py
View File

@@ -29,7 +29,7 @@ Usage:
if config_manager.validate(): if config_manager.validate():
print("Config is valid!") print("Config is valid!")
""" """
from .schema import PixelleVideoConfig, LLMConfig, ComfyUIConfig, TTSSubConfig, ImageSubConfig, VideoSubConfig from .schema import PixelleVideoConfig, LLMConfig, VLMConfig, ComfyUIConfig, TTSSubConfig, ImageSubConfig, VideoSubConfig
from .manager import ConfigManager from .manager import ConfigManager
from .loader import load_config_dict, save_config_dict from .loader import load_config_dict, save_config_dict
@@ -38,7 +38,8 @@ config_manager = ConfigManager()
__all__ = [ __all__ = [
"PixelleVideoConfig", "PixelleVideoConfig",
"LLMConfig", "LLMConfig",
"VLMConfig",
"ComfyUIConfig", "ComfyUIConfig",
"TTSSubConfig", "TTSSubConfig",
"ImageSubConfig", "ImageSubConfig",

9
pixelle_video/config/schema.py
View File

@@ -26,6 +26,14 @@ class LLMConfig(BaseModel):
model: str = Field(default="", description="LLM Model Name") model: str = Field(default="", description="LLM Model Name")
class VLMConfig(BaseModel):
"""VLM (Vision Language Model) configuration for character analysis"""
provider: str = Field(default="qwen", description="VLM provider: qwen, glm, openai")
api_key: str = Field(default="", description="VLM API Key")
base_url: str = Field(default="", description="VLM API Base URL (auto-detected if empty)")
model: str = Field(default="", description="VLM Model Name (defaults based on provider)")
class TTSLocalConfig(BaseModel): class TTSLocalConfig(BaseModel):
"""Local TTS configuration (Edge TTS)""" """Local TTS configuration (Edge TTS)"""
voice: str = Field(default="zh-CN-YunjianNeural", description="Edge TTS voice ID") voice: str = Field(default="zh-CN-YunjianNeural", description="Edge TTS voice ID")
@@ -92,6 +100,7 @@ class PixelleVideoConfig(BaseModel):
"""Pixelle-Video main configuration""" """Pixelle-Video main configuration"""
project_name: str = Field(default="Pixelle-Video", description="Project name") project_name: str = Field(default="Pixelle-Video", description="Project name")
llm: LLMConfig = Field(default_factory=LLMConfig) llm: LLMConfig = Field(default_factory=LLMConfig)
vlm: VLMConfig = Field(default_factory=VLMConfig)
comfyui: ComfyUIConfig = Field(default_factory=ComfyUIConfig) comfyui: ComfyUIConfig = Field(default_factory=ComfyUIConfig)
template: TemplateConfig = Field(default_factory=TemplateConfig) template: TemplateConfig = Field(default_factory=TemplateConfig)

13
pixelle_video/services/quality/character_analyzer.py
View File

@@ -149,13 +149,14 @@ Output ONLY the JSON object, no additional text."""
# Priority: Environment variables > config.yaml > defaults # Priority: Environment variables > config.yaml > defaults
from pixelle_video.config import config_manager from pixelle_video.config import config_manager
# Try to get VLM config from config.yaml # VLM config from config.yaml (now part of PixelleVideoConfig)
vlm_config = getattr(config_manager.config, 'vlm', None) vlm_config = config_manager.config.vlm
vlm_provider = os.getenv("VLM_PROVIDER") or (vlm_config.provider if vlm_config and hasattr(vlm_config, 'provider') else "qwen") # Environment variables override config.yaml
vlm_api_key = os.getenv("VLM_API_KEY") or os.getenv("DASHSCOPE_API_KEY") or (vlm_config.api_key if vlm_config and hasattr(vlm_config, 'api_key') else None) vlm_provider = os.getenv("VLM_PROVIDER") or vlm_config.provider or "qwen"
vlm_base_url = os.getenv("VLM_BASE_URL") or (vlm_config.base_url if vlm_config and hasattr(vlm_config, 'base_url') else None) vlm_api_key = os.getenv("VLM_API_KEY") or os.getenv("DASHSCOPE_API_KEY") or vlm_config.api_key
vlm_model = os.getenv("VLM_MODEL") or (vlm_config.model if vlm_config and hasattr(vlm_config, 'model') else None) vlm_base_url = os.getenv("VLM_BASE_URL") or vlm_config.base_url
vlm_model = os.getenv("VLM_MODEL") or vlm_config.model
# Configure based on provider # Configure based on provider
if vlm_provider == "qwen": if vlm_provider == "qwen":