diff --git a/wwwroot/api/cockpit_chat.dspy b/wwwroot/api/cockpit_chat.dspy
index 637ebeb..69f6033 100644
--- a/wwwroot/api/cockpit_chat.dspy
+++ b/wwwroot/api/cockpit_chat.dspy
@@ -145,6 +145,7 @@ async def _call_llm(model_info, messages, temperature):
     api_base = model_info.api_base.rstrip('/')
     api_key = model_info.api_key or ''
     model_id = model_info.model_id
+    debug(f'_call_llm: base={api_base} model={model_id} key_len={len(api_key)} key_prefix={api_key[:10]}')
 
     headers = {
         "Authorization": f"Bearer {api_key}",
@@ -158,13 +159,16 @@ async def _call_llm(model_info, messages, temperature):
 
     timeout = aiohttp.ClientTimeout(total=120)
     async with aiohttp.ClientSession(timeout=timeout) as session:
-        async with session.post(f"{api_base}/chat/completions", headers=headers, json=payload) as resp:
+        url = f"{api_base}/chat/completions"
+        debug(f'_call_llm: POST {url}')
+        async with session.post(url, headers=headers, json=payload) as resp:
             if resp.status != 200:
                 text = await resp.text()
+                debug(f'_call_llm: FAIL status={resp.status} body={text[:200]}')
                 raise ValueError(f"LLM API error {resp.status}: {text[:300]}")
             data = await resp.json()
             content = data["choices"][0]["message"]["content"]
-            # Truncate overly long responses
+            debug(f'_call_llm: OK reply_len={len(content)}')
             if len(content) > 8000:
                 content = content[:8000] + "\n\n...(内容过长已截断)"
             return content