Add workflow support and enhance runtime configuration in models and services

- Introduce RuntimeModelResource and RuntimeKnowledgeBase classes to manage workflow resources.
- Update AssistantConfig to include workflow_model_resources and workflow_knowledge_bases for better integration.
- Refactor validation and processing logic in routes and services to accommodate workflow types.
- Implement dynamic variable support for workflow assistants and enhance graph normalization.
- Add ToolExecutor for reusable tool execution across different assistant types.
- Update various services to ensure compatibility with new workflow features and improve error handling.

backend/services/config_resolver.py

@@ -12,7 +12,13 @@ from db.models import (
     ModelResource,
     Tool,
 )
-from models import AssistantConfig, RuntimeTool
+from models import (
+    AssistantConfig,
+    RuntimeKnowledgeBase,
+    RuntimeModelResource,
+    RuntimeTool,
+)
+from services.node_specs import graph_references, normalize_graph
 from sqlalchemy import select
 from sqlalchemy.ext.asyncio import AsyncSession
 
@@ -83,7 +89,7 @@ def _secret(resource: ModelResource | None, key: str, default: str = "") -> str:
 
 
 async def _tools_for(session: AsyncSession, assistant: Assistant) -> list[RuntimeTool]:
-    if assistant.type != "prompt":
+    if assistant.type not in {"prompt", "workflow"}:
         return []
     tools = (
         await session.execute(
@@ -134,6 +140,43 @@ async def resolve_runtime_config(
         else None
     )
 
+    graph = normalize_graph(assistant.graph or {}) if assistant.type == "workflow" else {}
+    refs = graph_references(graph) if graph else {
+        "model_resources": set(),
+        "knowledge_bases": set(),
+    }
+    workflow_resources: dict[str, RuntimeModelResource] = {}
+    if refs["model_resources"]:
+        resources = (
+            await session.execute(
+                select(ModelResource).where(ModelResource.id.in_(refs["model_resources"]))
+            )
+        ).scalars().all()
+        workflow_resources = {
+            resource.id: RuntimeModelResource(
+                id=resource.id,
+                name=resource.name,
+                capability=resource.capability,
+                interface_type=resource.interface_type,
+                values=resource.values or {},
+                secrets=resource.secrets or {},
+            )
+            for resource in resources
+            if resource.enabled
+        }
+    workflow_knowledge: dict[str, RuntimeKnowledgeBase] = {}
+    if refs["knowledge_bases"]:
+        knowledge_rows = (
+            await session.execute(
+                select(KnowledgeBase).where(KnowledgeBase.id.in_(refs["knowledge_bases"]))
+            )
+        ).scalars().all()
+        workflow_knowledge = {
+            kb.id: RuntimeKnowledgeBase(id=kb.id, name=kb.name, description=kb.description)
+            for kb in knowledge_rows
+            if kb.status == "active"
+        }
+
     return AssistantConfig(
         name=assistant.name,
         type=assistant.type,
@@ -150,7 +193,9 @@ async def resolve_runtime_config(
         knowledge_base_description=knowledge_base.description if knowledge_base else "",
         knowledge_retrieval_config=assistant.knowledge_retrieval_config or {},
         # workflow 图:仅 workflow 类型非空,引擎据此启用图驱动对话
-        graph=(assistant.graph or {}) if assistant.type == "workflow" else {},
+        graph=graph,
+        workflow_model_resources=workflow_resources,
+        workflow_knowledge_bases=workflow_knowledge,
         # 外部托管类型连接信息(DB 存真 key,直接注入)
         dify_api_url=str(_value(agent_resource, "apiUrl", assistant.api_url)),
         dify_api_key=_secret(agent_resource, "apiKey", assistant.api_key),