diff --git a/.env.example b/.env.example
index d15026c..e0c190c 100644
--- a/.env.example
+++ b/.env.example
@@ -1,5 +1,11 @@
-LLM_API_KEY=replace-with-your-api-key
-LLM_MODEL=replace-with-your-model-name
-# Optional: your provider endpoint
-# LLM_BASE_URL=https://your-provider.example/v1
-EMBEDDING_BATCH_SIZE=32
+LLM_API_KEY=sk-xxx
+LLM_MODEL=gpt-6
+LLM_BASE_URL=https://api.openai.com/v1
+
+# Jev / TypeSafe
+TYPESAFE_API_KEY=apikey_xxx
+TYPESAFE_DEFAULT_MODEL=jev-latest
+# API 根地址，不带 /v1/systemone
+# TYPESAFE_BASE_URL=https://api.typesafe.ai
+TYPESAFE_TIMEOUT=30
+TYPESAFE_MAX_RETRIES=2
diff --git a/app/clients/llm.py b/app/clients/llm.py
index c4bcf84..9628306 100644
--- a/app/clients/llm.py
+++ b/app/clients/llm.py
@@ -7,6 +7,8 @@ from app.config import settings
 
 @cache
 def get_llm() -> ChatOpenAI:
+    if not settings.llm_api_key.strip() or not settings.llm_model.strip():
+        raise ValueError("未配置 LLM_API_KEY 或 LLM_MODEL")
     return ChatOpenAI(
         model=settings.llm_model,
         api_key=settings.llm_api_key,
diff --git a/app/config.py b/app/config.py
index 942bbec..f623598 100644
--- a/app/config.py
+++ b/app/config.py
@@ -11,15 +11,34 @@ load_dotenv(Path(__file__).resolve().parent.parent / ".env")
 class Settings:
     # LLM 配置
     llm_api_key: str = field(
-        default_factory=lambda: os.environ["LLM_API_KEY"],
+        default_factory=lambda: os.getenv("LLM_API_KEY", ""),
         repr=False,
     )
-    llm_model: str = field(default_factory=lambda: os.environ["LLM_MODEL"])
+    llm_model: str = field(default_factory=lambda: os.getenv("LLM_MODEL", ""))
     llm_base_url: str | None = field(
         default_factory=lambda: os.getenv("LLM_BASE_URL") or None
     )
     llm_timeout: float = 60.0
     llm_max_retries: int = 2
 
+    # Jev
+    typesafe_api_key: str = field(
+        default_factory=lambda: os.getenv("TYPESAFE_API_KEY", ""), repr=False
+    )
+    typesafe_model: str = field(
+        default_factory=lambda: os.getenv("TYPESAFE_DEFAULT_MODEL") or "jev-latest"
+    )
+    typesafe_base_url: str = field(
+        default_factory=lambda: (
+            os.getenv("TYPESAFE_BASE_URL") or "https://api.typesafe.ai"
+        )
+    )
+    typesafe_timeout: float = field(
+        default_factory=lambda: float(os.getenv("TYPESAFE_TIMEOUT", "30"))
+    )
+    typesafe_max_retries: int = field(
+        default_factory=lambda: int(os.getenv("TYPESAFE_MAX_RETRIES", "2"))
+    )
+
 
 settings = Settings()
diff --git a/app/exceptions.py b/app/exceptions.py
index 1abedc3..589de4d 100644
--- a/app/exceptions.py
+++ b/app/exceptions.py
@@ -1,5 +1,22 @@
 class AppError(Exception):
-    """应用业务异常。"""
+    """应用业务异常"""
+
+
+class DecisionError(AppError):
+    """决策服务错误；仅暴露安全摘要及服务端追踪信息"""
+
+    def __init__(
+        self,
+        message: str,
+        *,
+        kind: str,
+        status: int | None = None,
+        request_id: str | None = None,
+    ):
+        self.kind = kind
+        self.status = status
+        self.request_id = request_id
+        super().__init__(message)
 
 
 class ImportProcessError(AppError):
diff --git a/app/processor/base.py b/app/processor/base.py
index bfcbeb4..33ca581 100644
--- a/app/processor/base.py
+++ b/app/processor/base.py
@@ -1,9 +1,12 @@
 import logging
 import time
 from abc import ABC, abstractmethod
-from typing import TypeVar
+from typing import TYPE_CHECKING, TypeVar
 
-from app.config import Settings, settings
+from pydantic import ValidationError
+
+if TYPE_CHECKING:
+    from app.config import Settings
 from app.exceptions import AppError, ImportProcessError
 
 T = TypeVar("T")
@@ -12,18 +15,26 @@ T = TypeVar("T")
 class BaseNode(ABC):
     name: str = "base_node"
 
-    def __init__(self, config: Settings | None = None):
-        """传入配置时使用传入值，否则使用全局配置"""
-        self.config = config if config is not None else settings
+    def __init__(self, config: "Settings | None" = None):
+        """保存显式配置；全局配置在节点实际使用时才加载。"""
+        self._config = config
         self.logger = logging.getLogger(f"import.{self.name}")
 
+    @property
+    def config(self) -> "Settings":
+        if self._config is None:
+            from app.config import settings
+
+            self._config = settings
+        return self._config
+
     def __call__(self, state: T) -> T:
         """执行节点，统一记录日志、耗时并保留业务异常类型"""
         start_time = time.perf_counter()
         self.logger.info("--- %s 开始 ---", self.name)
         try:
             result = self.process(state)
-        except AppError:
+        except (AppError, ValidationError):
             self.logger.exception("%s 执行失败", self.name)
             raise
         except Exception as e:
diff --git a/app/processor/nodes/input_node.py b/app/processor/nodes/input_node.py
new file mode 100644
index 0000000..1739935
--- /dev/null
+++ b/app/processor/nodes/input_node.py
@@ -0,0 +1,60 @@
+"""输入解析与校验，可直接运行本文件修改示例进行调试"""
+
+from app.processor.base import BaseNode
+from app.processor.state import QueryState
+from app.schemas.input_context import InputContext
+
+
+class InputNode(BaseNode):
+    """纯输入节点，不依赖模型配置；校验失败保留 ValidationError"""
+
+    name = "input_node"
+
+    def process(self, state: QueryState) -> QueryState:
+        # 深拷贝隔离原始请求，后续修改上下文不会改变提交时的快照
+        context = InputContext.model_validate(state["request"]).model_copy(deep=True)
+        updated_state: QueryState = {**state, "input_context": context}
+        return updated_state
+
+
+if __name__ == "__main__":
+    import json
+
+    # 1. 创建 state：可将 screen 替换成前端的完整大屏 JSON
+    state: QueryState = {
+        "request": {
+            "prompt": "把销售额统计向右移动 50px",
+            "screen": {
+                "canvas": {
+                    "width": 1920,
+                    "height": 1080,
+                    "backgroundColor": "#ffffff10",
+                },
+                "nodes": [
+                    {
+                        "id": "sales-chart",
+                        "type": "bar-chart",
+                        "name": "销售额统计",
+                        "layout": {"x": 26, "y": 34.8, "width": 400, "height": 260},
+                        "props": {"option": {"title": {"text": "销售额统计"}}},
+                    }
+                ],
+                "dataSource": [],
+            },
+            "context": {"selected_node_ids": ["sales-chart"]},
+            "attachments": [],
+        }
+    }
+
+    # 2. 执行节点
+    node = InputNode()
+    final_state = node.process(state)
+
+    # 3. 打印完整状态
+    printable_state = {
+        **final_state,
+        "input_context": final_state["input_context"].model_dump(
+            mode="json", exclude_unset=True
+        ),
+    }
+    print(json.dumps(printable_state, ensure_ascii=False, indent=2))
diff --git a/app/schema/__init__.py b/app/processor/nodes/requirement_node.py
similarity index 100%
rename from app/schema/__init__.py
rename to app/processor/nodes/requirement_node.py
diff --git a/app/processor/state.py b/app/processor/state.py
new file mode 100644
index 0000000..1e755fb
--- /dev/null
+++ b/app/processor/state.py
@@ -0,0 +1,9 @@
+from typing import NotRequired, TypedDict
+
+from app.schemas.base import JsonObject
+from app.schemas.input_context import InputContext
+
+
+class QueryState(TypedDict):
+    request: JsonObject
+    input_context: NotRequired[InputContext]
diff --git a/app/schemas/__init__.py b/app/schemas/__init__.py
new file mode 100644
index 0000000..5de356b
--- /dev/null
+++ b/app/schemas/__init__.py
@@ -0,0 +1,18 @@
+"""前端请求与大屏公共模型"""
+
+from .input_context import InputContext
+from .request import DesignRequest, RequestContext
+from .requirement import Requirement, RequirementChange
+from .screen import Canvas, Layout, Screen, ScreenNode
+
+__all__ = [
+    "Canvas",
+    "DesignRequest",
+    "InputContext",
+    "Layout",
+    "RequestContext",
+    "Requirement",
+    "RequirementChange",
+    "Screen",
+    "ScreenNode",
+]
diff --git a/app/schemas/base.py b/app/schemas/base.py
new file mode 100644
index 0000000..9a2c2b4
--- /dev/null
+++ b/app/schemas/base.py
@@ -0,0 +1,23 @@
+"""输入协议的公共类型"""
+
+from typing import Annotated
+
+from pydantic import AfterValidator, BaseModel, ConfigDict, JsonValue
+
+
+def require_non_blank(value: str) -> str:
+    if not value.strip():
+        raise ValueError("不能为空或仅包含空白字符")
+    return value
+
+
+NonBlankString = Annotated[str, AfterValidator(require_non_blank)]
+JsonObject = dict[str, JsonValue]
+
+
+class ExtensibleModel(BaseModel):
+    """校验已知字段，保留未知 JSON 字段；不隐式转换输入类型"""
+
+    model_config = ConfigDict(extra="allow", strict=True, allow_inf_nan=False)
+
+    __pydantic_extra__: dict[str, JsonValue]
diff --git a/app/schemas/input_context.py b/app/schemas/input_context.py
new file mode 100644
index 0000000..71a75b3
--- /dev/null
+++ b/app/schemas/input_context.py
@@ -0,0 +1,7 @@
+"""输入节点校验后的上下文"""
+
+from .request import DesignRequest
+
+
+class InputContext(DesignRequest):
+    """当前复用请求协议，后续可添加解析结果而不改变前端传参"""
diff --git a/app/schemas/request.py b/app/schemas/request.py
new file mode 100644
index 0000000..1448c2a
--- /dev/null
+++ b/app/schemas/request.py
@@ -0,0 +1,21 @@
+"""前端设计请求：原始需求、当前大屏及可选补充上下文"""
+
+from pydantic import Field
+
+from .base import ExtensibleModel, JsonObject, NonBlankString
+from .screen import Screen
+
+
+class RequestContext(ExtensibleModel):
+    selected_node_ids: list[NonBlankString] = Field(default_factory=list)
+    screen_id: NonBlankString | None = None
+    screen_revision: int | None = Field(default=None, ge=0)
+    conversation_id: NonBlankString | None = None
+
+
+class DesignRequest(ExtensibleModel):
+    prompt: NonBlankString
+    screen: Screen
+    context: RequestContext = Field(default_factory=RequestContext)
+    # 只预留 JSON 描述入口，暂不规定材料种类，也不读取文件或远程资源
+    attachments: list[JsonObject] = Field(default_factory=list)
diff --git a/pyproject.toml b/pyproject.toml
index 080a6fa..2af038a 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -3,7 +3,9 @@ name = "my-v0-server"
 version = "0.1.0"
 requires-python = ">=3.11"
 dependencies = [
+    "pydantic>=2.13.5,<3",
     "langchain-openai>=1.6.2",
     "langgraph>=1.2.12",
     "python-dotenv>=1.2.3",
+    "typesafe-sdk>=0.7.1",
 ]
diff --git a/uv.lock b/uv.lock
index 6f67c95..87f44ed 100644
--- a/uv.lock
+++ b/uv.lock
@@ -551,14 +551,18 @@ source = { virtual = "." }
 dependencies = [
     { name = "langchain-openai" },
     { name = "langgraph" },
+    { name = "pydantic" },
     { name = "python-dotenv" },
+    { name = "typesafe-sdk" },
 ]
 
 [package.metadata]
 requires-dist = [
     { name = "langchain-openai", specifier = ">=1.6.2" },
     { name = "langgraph", specifier = ">=1.2.12" },
+    { name = "pydantic", specifier = ">=2.13.5,<3" },
     { name = "python-dotenv", specifier = ">=1.2.3" },
+    { name = "typesafe-sdk", specifier = ">=0.7.1" },
 ]
 
 [[package]]
@@ -1114,6 +1118,22 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/19/97/56608b2249fe206a67cd573bc93cd9896e1efb9e98bce9c163bcdc704b88/truststore-0.10.4-py3-none-any.whl", hash = "sha256:adaeaecf1cbb5f4de3b1959b42d41f6fab57b2b1666adb59e89cb0b53361d981", size = 18660, upload-time = "2025-08-12T18:49:01.46Z" },
 ]
 
+[[package]]
+name = "typesafe-sdk"
+version = "0.7.1"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "httpx2" },
+    { name = "pydantic" },
+    { name = "pydantic-core" },
+    { name = "tenacity" },
+    { name = "typing-extensions" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/dc/d4/14d9a00c759c126c814f7cfc0b2a760768214cda0708a9af49577b321ee3/typesafe_sdk-0.7.1.tar.gz", hash = "sha256:495ce9d81733b4f95edcb156e95042a7ec4bfd59972a1150ffa8ce1e11f85a54", size = 23062, upload-time = "2026-09-21T15:57:53.579Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/75/72/7e57648d27e45260b04762dd2393a81604b516a282fc5efe94cdf90ab971/typesafe_sdk-0.7.1-py3-none-any.whl", hash = "sha256:9d04eee13b5f64bbbe6f8e96cf40e3f39ca4efc934a1f4f39490cb9a43e9d2ad", size = 36366, upload-time = "2026-09-21T15:57:52.314Z" },
+]
+
 [[package]]
 name = "typing-extensions"
 version = "4.16.0"