fix(chat): OpenClaw 改用 NIM llama-3.1-8b 對話 + NemoClaw timeout 120s + 老闆稱謂

1. _call_openclaw: 改用 NIM meta/llama-3.1-8b-instruct
   舊的 analyze/incident 是告警 API，回覆是告警格式，不適合對話
2. _call_nemotron: 移除 Ollama fallback，回到純 NIM
3. NEMOTRON_TIMEOUT_SECONDS: 55 → 120 (ConfigMap 已更新)
4. 修正「統帥」→「老闆」

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

apps/api/src/services/chat_manager.py

@@ -78,52 +78,35 @@ class ChatManager:
 
     async def _call_openclaw(self, system_prompt: str, user_message: str) -> str | None:
         """
-        呼叫 OpenClaw 對話 — 走 Ollama qwen2.5:7b-instruct (192.168.0.188:11434)
+        呼叫 OpenClaw 對話 — 走 NVIDIA NIM meta/llama-3.1-8b-instruct
 
         2026-04-03 ogt: OpenClaw 8088 的 analyze/incident 是告警分析 API，
-        不適合做自然語言對話（回覆會是告警格式）。
-        改用 Ollama 本地模型做 chat，速度快、無格式污染。
+        回覆是告警格式，不適合自然語言對話。
+        改用 NIM llama-3.1-8b 做 chat，與 NemoClaw 同樣走免費 NIM cloud。
         """
-        import httpx
-        from src.core.config import get_settings
-        settings = get_settings()
-
-        ollama_url = getattr(settings, 'OLLAMA_URL', 'http://192.168.0.188:11434')
-        openclaw_timeout = float(getattr(settings, 'OPENCLAW_TIMEOUT', 40.0))
+        from src.services.nvidia_provider import get_nvidia_provider
+        nvidia = get_nvidia_provider()
         try:
-            async with httpx.AsyncClient(timeout=openclaw_timeout) as client:
-                resp = await client.post(
-                    f"{ollama_url}/api/chat",
-                    json={
-                        "model": "qwen2.5:7b-instruct",
-                        "stream": False,
-                        "messages": [
-                            {"role": "system", "content": system_prompt},
-                            {"role": "user", "content": user_message},
-                        ],
-                        "options": {"num_predict": 300},
-                    },
-                )
-                resp.raise_for_status()
-                data = resp.json()
-                return data.get("message", {}).get("content", "").strip() or None
+            full_prompt = f"{system_prompt}\n\n用戶訊息: {user_message}"
+            response, success, _, _ = await nvidia.chat(
+                prompt=full_prompt,
+                model="meta/llama-3.1-8b-instruct",
+                max_tokens=300,
+            )
+            if success and response and "not configured" not in response and "Circuit Breaker" not in response:
+                return response.strip()
+            return None
         except Exception as e:
             logger.warning("openclaw_chat_failed", error=str(e))
             return None
 
     async def _call_nemotron(self, system_prompt: str, user_message: str) -> str | None:
         """
-        呼叫 NemoClaw — NIM 優先，超時則 fallback 到 Ollama llama3.2:3b
+        呼叫 NVIDIA NIM nemotron-mini-4b (NemoClaw)
 
-        2026-04-03 ogt: NIM 免費 tier 延遲 11-45s 且常超時，
-        加 Ollama fallback 確保 NemoClaw 一定有回應。
+        NIM 免費 tier 延遲 11-45s，此方法可能需要 30-120s 才回應
         """
-        import httpx
-        from src.core.config import get_settings as _get_settings
         from src.services.nvidia_provider import get_nvidia_provider
-        settings = _get_settings()
-
-        # 優先嘗試 NIM (timeout 20s，快速失敗)
         nvidia = get_nvidia_provider()
         try:
             full_prompt = f"{system_prompt}\n\n用戶訊息: {user_message}"
@@ -134,35 +117,10 @@ class ChatManager:
             )
             if success and response and "not configured" not in response and "Circuit Breaker" not in response:
                 return response.strip()
+            return None
         except Exception as e:
-            logger.warning("nemotron_nim_failed_fallback_ollama", error=str(e))
-
-        # Fallback: Ollama llama3.2:3b (本地，速度快)
-        ollama_url = getattr(settings, 'OLLAMA_URL', 'http://192.168.0.188:11434')
-        try:
-            async with httpx.AsyncClient(timeout=30.0) as client:
-                resp = await client.post(
-                    f"{ollama_url}/api/chat",
-                    json={
-                        "model": "llama3.2:3b",
-                        "stream": False,
-                        "messages": [
-                            {"role": "system", "content": system_prompt},
-                            {"role": "user", "content": user_message},
-                        ],
-                        "options": {"num_predict": 250},
-                    },
-                )
-                resp.raise_for_status()
-                data = resp.json()
-                result = data.get("message", {}).get("content", "").strip()
-                if result:
-                    logger.info("nemotron_ollama_fallback_used")
-                    return result
-        except Exception as e:
-            logger.warning("nemotron_ollama_fallback_failed", error=str(e))
-
-        return None
+            logger.warning("nemotron_chat_failed", error=str(e))
+            return None
 
     async def generate_response(
         self,
@@ -199,7 +157,7 @@ class ChatManager:
         nemo_task = asyncio.create_task(
             self._call_nemotron(
                 f"{NEMOCLAW_PERSONA}\n{context}",
-                f"統帥問了: {text}\n\n請從 NemoClaw 角度補充或評論。",
+                f"老闆問了: {text}\n\n請從 NemoClaw 角度補充或評論。",
             )
         )