feat(workflow): add per-agent vision configuration

backend/services/node_specs.py

@@ -49,7 +49,7 @@ NODE_SPECS: list[dict[str, Any]] = [
         "name": "agent",
         "displayName": "Agent",
         "category": "conversation_node",
-        "description": "阶段智能体：绑定上下文、工具、知识库及 ASR/TTS 资源。",
+        "description": "阶段智能体：绑定上下文、工具、知识库、视觉及语音资源。",
         "icon": "Bot",
         "accent": "sky",
         "addable": True,
@@ -145,6 +145,8 @@ def _normalize_agent_data(data: dict[str, Any]) -> None:
                 data.get("llmResourceId"),
                 data.get("asrResourceId"),
                 data.get("ttsResourceId"),
+                data.get("visionEnabled"),
+                data.get("visionModelResourceId"),
                 data.get("knowledgeBaseId"),
                 data.get("toolIds"),
             )
@@ -157,6 +159,8 @@ def _normalize_settings(settings: dict[str, Any], *, global_prompt: str = "") ->
     settings.setdefault("defaultLlmResourceId", "")
     settings.setdefault("defaultAsrResourceId", "")
     settings.setdefault("defaultTtsResourceId", "")
+    settings.setdefault("visionEnabled", False)
+    settings.setdefault("visionModelResourceId", "")
     settings.setdefault("toolIds", [])
     settings.setdefault("knowledgeBaseId", "")
     settings.setdefault("knowledgeMode", "automatic")
@@ -450,6 +454,11 @@ def graph_references(graph: dict[str, Any]) -> dict[str, set[str]]:
             settings.get("defaultLlmResourceId"),
             settings.get("defaultAsrResourceId"),
             settings.get("defaultTtsResourceId"),
+            (
+                settings.get("visionModelResourceId")
+                if settings.get("visionEnabled")
+                else None
+            ),
         )
         if value
     }
@@ -469,6 +478,11 @@ def graph_references(graph: dict[str, Any]) -> dict[str, set[str]]:
                 data.get("llmResourceId"),
                 data.get("asrResourceId"),
                 data.get("ttsResourceId"),
+                (
+                    data.get("visionModelResourceId")
+                    if data.get("visionEnabled")
+                    else None
+                ),
             ):
                 if resource_id:
                     resources.add(str(resource_id))