feat: engine 透传 error_code（与 error 同源取最后一次异常）

src/genesis/inference/engine.py

@@ -101,6 +101,7 @@ class InferenceEngine:
 
         start = time.monotonic()
         last_error: str | None = None
+        last_error_code: str | None = None
         for idx, name in enumerate(self._model_names(model)):
             try:
                 text, usage = self._call(
@@ -115,12 +116,13 @@ class InferenceEngine:
                 )
             except LLMError as exc:
                 last_error = str(exc)
+                last_error_code = exc.error_code   # 同源：取最后一次失败异常
 
         return ChatResult(
             text="", model=name,
             prompt_version=getattr(prompt, "version", "inline"),
             usage=TokenUsage(), duration_ms=int((time.monotonic() - start) * 1000),
-            status="failed", error=last_error,
+            status="failed", error=last_error, error_code=last_error_code,
         )
 
     def chat_structured(
@@ -144,6 +146,7 @@ class InferenceEngine:
         attempts = 0
         last_raw = ""
         last_error: str | None = None
+        last_error_code: str | None = None
 
         while attempts <= retry_count:
             attempts += 1
@@ -166,6 +169,7 @@ class InferenceEngine:
                 )
             except json.JSONDecodeError as exc:
                 last_error = f"JSON 解析失败: {exc}"
+                last_error_code = "LLM_PARSE_ERROR"
                 # 带错误信息重试
                 base_rendered = base_rendered + f"\n\n上次解析失败：{exc}。请重新输出合法 JSON。"
             except LLMError as exc:
@@ -175,7 +179,7 @@ class InferenceEngine:
                     prompt_version=getattr(prompt, "version", "inline"),
                     usage=TokenUsage(),
                     duration_ms=int((time.monotonic() - start) * 1000),
-                    status="failed", error=str(exc),
+                    status="failed", error=str(exc), error_code=exc.error_code,
                 )
 
         return StructuredResult(
@@ -184,5 +188,5 @@ class InferenceEngine:
             prompt_version=getattr(prompt, "version", "inline"),
             usage=TokenUsage(),
             duration_ms=int((time.monotonic() - start) * 1000),
-            status="parse_error", error=last_error,
+            status="parse_error", error=last_error, error_code=last_error_code,
         )