feat(config): align LLM timeout/retry SLA with FastGPT client (60s/3)

Replaces the unequal SLA (LLM 30s/2 retries vs FastGPT client 60s/3)
that made the LangGraph backend look less stable under gray release.

- core/fastgpt_client: FASTGPT_TIMEOUT_SECONDS / FASTGPT_MAX_RETRIES
  module constants (were inline literals in lifespan)
- agent/llm: OpenAILLMService defaults 60.0s / 3 retries
- core/config: named-profile fallback defaults 60/3 (prefixed and
  JSON-dict forms)
- contract test test/backends/test_sla_alignment.py pins both sides to
  the same constants so the SLA cannot drift apart again
- retry backoff intentionally differs (FastGPT fixed 1s vs LLM
  exponential) and is documented; aligned parameters are per-call
  timeout and max retry count

Verified: 215 unit tests (3 new SLA contract tests)

.env.example

@@ -17,6 +17,8 @@ VOICE_CONFIG=config/voice.json
 # 每组 LLM_<名>_BASE_URL / _API_KEY / _MODEL 必填；
 # 可选 _TIMEOUT_SECONDS / _MAX_RETRIES / _TEMPERATURE / _EXTRA_BODY（JSON 字符串，
 # 如思考开关 KIMI 网关为 {"reasoning_effort":"none"}）。
+# 超时/重试缺省 60 秒 / 3 次，与 FastGPT 客户端 SLA 对齐
+# （契约测试 test/backends/test_sla_alignment.py）。
 # LLM_PROVIDERS=deepseek,kimi,gpt6luna
 # LLM_PROVIDER=deepseek
 # LLM_DEEPSEEK_BASE_URL=https://api.openai.com/v1

src/agent/llm.py

@@ -100,8 +100,8 @@ class OpenAILLMService:
         model: str,
         api_key: str,
         base_url: str = "",
-        timeout_seconds: float = 30.0,
+        timeout_seconds: float = 60.0,
-        max_retries: int = 2,
+        max_retries: int = 3,
         temperature: float | None = None,
         extra_body: dict | None = None,
     ):

src/core/config.py

@@ -91,8 +91,8 @@ class Config:
                 "base_url": base_url,
                 "api_key": api_key,
                 "model": model,
-                "timeout_seconds": float(spec.get("timeout_seconds", 30)),
+                "timeout_seconds": float(spec.get("timeout_seconds", 60)),
-                "max_retries": int(spec.get("max_retries", 2)),
+                "max_retries": int(spec.get("max_retries", 3)),
                 "temperature": float(temperature) if temperature is not None else None,
                 "extra_body": str(extra_body),
             }
@@ -113,8 +113,8 @@ class Config:
                 "base_url": base_url,
                 "api_key": api_key,
                 "model": model,
-                "timeout_seconds": float(os.getenv(prefix + "TIMEOUT_SECONDS", "30")),
+                "timeout_seconds": float(os.getenv(prefix + "TIMEOUT_SECONDS", "60")),
-                "max_retries": int(os.getenv(prefix + "MAX_RETRIES", "2")),
+                "max_retries": int(os.getenv(prefix + "MAX_RETRIES", "3")),
                 "temperature": float(temperature) if temperature else None,
                 "extra_body": os.getenv(prefix + "EXTRA_BODY", ""),
             }

src/core/fastgpt_client.py

@@ -16,6 +16,14 @@ _chat_backend: ChatBackend | None = None
 _langgraph_graph = None
 _app_service: AccidentAppService | None = None
 
+# 后端超时/重试 SLA（两后端对齐；契约测试防止再漂移）。
+# FastGPT 侧：客户端每次请求总超时 + 重试次数；
+# LangGraph 侧：LLM 每次调用总超时 + 重试次数（退避为指数式，
+# 与 FastGPT 固定 1s 不同——总尝试次数与单次超时可比，见
+# test/backends/test_sla_alignment.py）。
+FASTGPT_TIMEOUT_SECONDS = 60.0
+FASTGPT_MAX_RETRIES = 3
+
 
 @asynccontextmanager
 async def lifespan(app: FastAPI):
@@ -39,8 +47,8 @@ async def lifespan(app: FastAPI):
         _fastgpt_client = AsyncChatClient(
             api_key=Config.FASTGPT_API_KEY,
             base_url=Config.FASTGPT_BASE_URL,
-            timeout=60.0,
+            timeout=FASTGPT_TIMEOUT_SECONDS,
-            max_retries=3,
+            max_retries=FASTGPT_MAX_RETRIES,
             retry_delay=1.0,
             enable_logging=Config.DEBUG,
         )

test/backends/test_sla_alignment.py

@@ -0,0 +1,42 @@
+"""后端超时/重试 SLA 对齐契约测试。
+
+背景（可替代性清单）：FastGPT 客户端为 60s/3 次，LangGraph LLM 侧
+曾是 30s/2 次，客户端感知到的超时行为不等价。本测试把两端常量
+钉在一起，防止再次漂移。
+
+注：退避策略有意不同——FastGPT 固定 1s，LLM 指数退避（限流场景
+更友好）；对齐的是"单次调用总超时"与"最大重试次数"两个 SLA 参数。
+"""
+
+import inspect
+
+from src.agent.llm import OpenAILLMService
+from src.core.fastgpt_client import FASTGPT_MAX_RETRIES, FASTGPT_TIMEOUT_SECONDS
+
+
+def test_llm_default_timeout_matches_fastgpt_client():
+    signature = inspect.signature(OpenAILLMService.__init__)
+    default_timeout = signature.parameters["timeout_seconds"].default
+    assert default_timeout == FASTGPT_TIMEOUT_SECONDS == 60.0
+
+
+def test_llm_default_max_retries_matches_fastgpt_client():
+    signature = inspect.signature(OpenAILLMService.__init__)
+    default_retries = signature.parameters["max_retries"].default
+    assert default_retries == FASTGPT_MAX_RETRIES == 3
+
+
+def test_named_profile_defaults_match_fastgpt_client(monkeypatch):
+    """未显式配置 _TIMEOUT_SECONDS/_MAX_RETRIES 的 profile 取 SLA 默认值。"""
+    monkeypatch.setenv("LLM_PROVIDERS", "probe")
+    monkeypatch.setenv("LLM_PROBE_BASE_URL", "http://127.0.0.1:1/v1")
+    monkeypatch.setenv("LLM_PROBE_API_KEY", "probe-key")
+    monkeypatch.setenv("LLM_PROBE_MODEL", "probe-model")
+
+    from src.core.config import Config
+
+    # LLM_PROVIDERS 是 import 时固化的类属性，需直接替换
+    monkeypatch.setattr(Config, "LLM_PROVIDERS", "probe")
+    profile = Config.llm_profile("probe")
+    assert profile["timeout_seconds"] == FASTGPT_TIMEOUT_SECONDS
+    assert profile["max_retries"] == FASTGPT_MAX_RETRIES