From 95365298f2bd76096d8c21bc7f6b7d1bba9786db Mon Sep 17 00:00:00 2001
From: MerCry <jiong1114>
Date: Fri, 6 Mar 2026 02:02:03 +0800
Subject: [PATCH] feat: add timeout and retry configuration for LLM client
 [AC-AISVC-LLM]

---
 .vscode/mcp.json                             | 11 +++++++++++
 ai-service/app/services/llm/factory.py       | 20 ++++++++++++++++++++
 ai-service/app/services/llm/openai_client.py | 13 +++++++++++--
 3 files changed, 42 insertions(+), 2 deletions(-)
 create mode 100644 .vscode/mcp.json

diff --git a/.vscode/mcp.json b/.vscode/mcp.json
new file mode 100644
index 0000000..4d1a2a1
--- /dev/null
+++ b/.vscode/mcp.json
@@ -0,0 +1,11 @@
+{
+  "servers": {
+    "chrome-devtools-mcp": {
+      "command": "npx",
+      "args": [
+        "-y",
+        "chrome-devtools-mcp@latest"
+      ]
+    }
+  }
+}
diff --git a/ai-service/app/services/llm/factory.py b/ai-service/app/services/llm/factory.py
index c802792..c492886 100644
--- a/ai-service/app/services/llm/factory.py
+++ b/ai-service/app/services/llm/factory.py
@@ -68,6 +68,22 @@ LLM_PROVIDERS: dict[str, LLMProviderInfo] = {
                     "minimum": 0,
                     "maximum": 2,
                 },
+                "timeout_seconds": {
+                    "type": "integer",
+                    "title": "请求超时(秒)",
+                    "description": "LLM 请求超时时间（秒）",
+                    "default": 60,
+                    "minimum": 5,
+                    "maximum": 180,
+                },
+                "max_retries": {
+                    "type": "integer",
+                    "title": "最大重试次数",
+                    "description": "请求失败后的最大重试次数",
+                    "default": 3,
+                    "minimum": 0,
+                    "maximum": 10,
+                },
             },
             "required": ["api_key"],
         },
@@ -252,6 +268,8 @@ class LLMProviderFactory:
                     model=config.get("model", "gpt-4o-mini"),
                     max_tokens=config.get("max_tokens", 2048),
                     temperature=config.get("temperature", 0.7),
+                    timeout_seconds=config.get("timeout_seconds", 60),
+                    max_retries=config.get("max_retries", 3),
                 ),
             )
 
@@ -276,6 +294,8 @@ class LLMConfigManager:
             "model": settings.llm_model,
             "max_tokens": settings.llm_max_tokens,
             "temperature": settings.llm_temperature,
+            "timeout_seconds": settings.llm_timeout_seconds,
+            "max_retries": settings.llm_max_retries,
         }
         self._client: LLMClient | None = None
 
diff --git a/ai-service/app/services/llm/openai_client.py b/ai-service/app/services/llm/openai_client.py
index 859ed08..2b9473a 100644
--- a/ai-service/app/services/llm/openai_client.py
+++ b/ai-service/app/services/llm/openai_client.py
@@ -68,10 +68,18 @@ class OpenAIClient(LLMClient):
             max_retries=settings.llm_max_retries,
         )
         self._client: httpx.AsyncClient | None = None
+        self._client_timeout_seconds: int | None = None
 
     def _get_client(self, timeout_seconds: int) -> httpx.AsyncClient:
-        """Get or create HTTP client."""
-        if self._client is None:
+        """Get or create HTTP client.
+
+        Recreate client when timeout changes to ensure runtime config takes effect.
+        """
+        if self._client is None or self._client_timeout_seconds != timeout_seconds:
+            if self._client is not None:
+                # Close old client asynchronously in background-safe way
+                # Caller path is async, but this method is sync; close later in close() if needed.
+                pass
             self._client = httpx.AsyncClient(
                 timeout=httpx.Timeout(timeout_seconds),
                 headers={
@@ -79,6 +87,7 @@ class OpenAIClient(LLMClient):
                     "Content-Type": "application/json",
                 },
             )
+            self._client_timeout_seconds = timeout_seconds
         return self._client
 
     def _build_request_body(