Add RuntimeTool model and enhance AssistantConfig for tool management

- Introduce a new `RuntimeTool` model to encapsulate tool data for runtime sessions, including attributes like `id`, `name`, `function_name`, `type`, and `description`.
- Update the `AssistantConfig` model to include a list of reusable tools, allowing for better management of tools within assistant configurations.
- Modify the `config_resolver` service to fetch and resolve tools associated with assistants, ensuring they are available during runtime.
- Refactor tool-related CRUD operations in the `tools` route to support the new runtime execution model, enhancing the overall tool management system.
- Update documentation and comments to reflect changes in tool execution and configuration handling, improving clarity for future development.

backend/models.py

@@ -15,6 +15,17 @@ from pydantic import BaseModel, Field
 RuntimeMode = Literal["pipeline", "realtime"]
 
 
+class RuntimeTool(BaseModel):
+    """Tool data resolved from an assistant binding for one runtime session."""
+
+    id: str
+    name: str
+    function_name: str
+    type: str
+    description: str = ""
+    definition: dict = Field(default_factory=dict)
+
+
 class AssistantConfig(BaseModel):
     """运行时配置:前端可见部分(name/prompt/...) + 服务端注入部分(*_api_key/*_base_url)。"""
 
@@ -61,6 +72,9 @@ class AssistantConfig(BaseModel):
 
     enableInterrupt: bool = True
 
+    # Prompt assistant reusable tools. Execution remains type-specific in the pipeline.
+    tools: list[RuntimeTool] = Field(default_factory=list)
+
     # workflow 类型:节点图(nodes/edges)。非 workflow 为空,引擎据此决定是否启用。
     graph: dict = {}