Add support for image input in model resources and enhance related configurations

- Introduce a new `support_image_input` field in model resources, allowing models to indicate support for image input.
- Update the backend models, schemas, and database seed scripts to accommodate the new field.
- Enhance the AssistantConfig and related routes to handle image input capabilities, ensuring proper validation and error handling.
- Modify the frontend components to include toggles for enabling visual understanding and filtering models based on image input support.
- Implement necessary adjustments in the voice preview and pipeline to integrate video stream handling alongside audio functionalities.

backend/services/pipecat/service_factory.py

@@ -72,10 +72,15 @@ def create_llm(cfg: AssistantConfig):
     """DeepSeek 等,走 OpenAI 兼容的 /v1/chat/completions。"""
     if cfg.llm_interface_type not in {"openai-llm", "dashscope-llm"}:
         raise ValueError(f"不支持的 LLM 接口类型: {cfg.llm_interface_type}")
+    extra_body = cfg.llm_values.get("extraBody")
+    extra = {"extra_body": extra_body} if isinstance(extra_body, dict) else {}
     return OpenAILLMService(
         api_key=cfg.llm_api_key or config.LLM_API_KEY,
         base_url=cfg.llm_base_url or config.LLM_BASE_URL,
-        settings=OpenAILLMService.Settings(model=cfg.model or config.LLM_MODEL),
+        settings=OpenAILLMService.Settings(
+            model=cfg.model or config.LLM_MODEL,
+            extra=extra,
+        ),
     )