feat(api): DeepSeek V4 Flash 默认、HTTP 错讯与多供应商分层

- 主链路默认 deepseek-v4-flash，DEEPSEEK_THINKING_ENABLED 对齐旧非思考 chat
- 评测台评审装配迁入 adapters/llm（deepseek_eval_judge、zhipu_eval_judge）与 eval_judge_spec
- 拆分 llm_http_openai_chat_errors 与 llm_errors（DeepSeek/智谱品牌与文档链），llm_call 支持 http_error_vendor
- EvalJudgeService 按 spec.provider 传入 allm_json_call；评测台前端文案改为 V4 Flash
- 更新 .env 示例与 staging/production 的 DEEPSEEK_MODEL；补充 openai/供应商错讯测试

Made-with: Cursor

api/app/adapters/llm/zhipu_eval_judge.py

@@ -0,0 +1,37 @@
+"""智谱 GLM 评测台评审：OpenAI 兼容端点 + ChatOpenAI 装配。"""
+
+from __future__ import annotations
+
+from langchain_openai import ChatOpenAI
+
+from app.adapters.llm.openai_base_url import normalize_openai_compatible_base_url
+from app.core.config import settings
+from app.core.eval_judge_spec import EvalJudgeLlmSpec
+
+
+def build_zhipu_eval_judge_spec(
+    judge_model: str | None,
+) -> EvalJudgeLlmSpec | None:
+    """密钥缺失时返回 None。"""
+    api_key = (settings.eval_judge_api_key or settings.zhipu_api_key or "").strip()
+    if not api_key:
+        return None
+    want = (judge_model or "").strip()
+    base = normalize_openai_compatible_base_url(
+        settings.eval_judge_base_url,
+        fallback="https://open.bigmodel.cn/api/paas/v4",
+    )
+    model = want or (settings.eval_judge_model or "glm-5")
+    ctx = int(settings.eval_judge_context_window_tokens)
+    llm_kw: dict = {
+        "api_key": api_key,
+        "base_url": base,
+        "model": model,
+        "temperature": settings.eval_judge_temperature,
+    }
+    return EvalJudgeLlmSpec(
+        llm=ChatOpenAI(**llm_kw),
+        provider="zhipu",
+        resolved_model=model,
+        context_window_tokens=ctx,
+    )