fix(im_llm): 代理绕过Docker内部hostname

backend/python/app/api/im_llm.py

@@ -16,6 +16,7 @@ IM Bot 真实 LLM 调用。
   * 平台知识库注入（get_platform_knowledge，技能清单实时同步）
 """
 import asyncio
+import os
 import json
 import logging
 import re as _re
@@ -48,7 +49,8 @@ def _proxy_for_url(url: str) -> str | None:
     host = (_urlparse(url).hostname or "").lower()
     if "agnes-ai.com" in host or "agnes-ai.space" in host:
         return _HK_PROXY
-    if "qihuyun.com" in host:
+    # 2026-07-30 修复 3+: 不代理 Docker 内部 hostname 和 localhost
+    if "qihuyun.com" in host or "localhost" in host or host.startswith("127.") or "qihu-aiflow" in host:
         return None
     _no_proxy = [h.strip().lower() for h in os.environ.get("NO_PROXY", "").split(",") if h.strip()]
     for _np in _no_proxy:
@@ -127,7 +129,12 @@ def _call_real_llm(db: Session, user=None, prompt: str = "", platform: str = "im
         )
 
         # ===== 1) 媒体意图路由（Issue 7）=====
-        intent = detect_media_intent(prompt)
+        # 2026-07-30 fix: 只对当前用户消息做媒体意图检测，避免历史中的图像请求污染
+        _cur_msg = prompt
+        _marker = "用户当前问题："
+        if _marker in prompt:
+            _cur_msg = prompt[prompt.rfind(_marker) + len(_marker):].strip()
+        intent = detect_media_intent(_cur_msg)
         if intent:
             ai_model = resolve_auto_model(intent, db)
             if ai_model:
@@ -196,6 +203,10 @@ def _call_real_llm(db: Session, user=None, prompt: str = "", platform: str = "im
             provider = db.query(ModelProvider).filter(ModelProvider.id == ai_model.provider_id).first()
             api_key = ai_model.api_key or (provider.api_key if provider else None)
             api_url = ai_model.endpoint
+            # 2026-07-30: LiteLLM 端点统一用 master key（与 chat_service.py 对齐）
+            if api_url and "qihu-aiflow-litellm" in api_url:
+                _litellm_key = os.environ.get("LITELLM_MASTER_KEY") or "sk-litellm-master-key-aiflow-2026"
+                api_key = _litellm_key
             if not api_url and provider and provider.api_base_url:
                 api_url = f"{provider.api_base_url.rstrip('/')}/v1/chat/completions"
             if not (api_key and api_url):
@@ -221,13 +232,15 @@ def _call_real_llm(db: Session, user=None, prompt: str = "", platform: str = "im
                 ai_content = ""
                 _http_ok = True
                 for round_num in range(max_rounds):
-                    request_body = {"model": ai_model.name, "messages": messages, "max_tokens": 2048}
+                    request_body = {"model": ai_model.name, "messages": messages, "max_tokens": 2048, "enable_thinking": False, "temperature": 0.7}
                     if function_definitions and skills:
                         request_body["tools"] = [
                             {"type": "function", "function": fd} for fd in function_definitions
                         ]
                         request_body["tool_choice"] = "auto"
                     try:
+                        import logging as _l5
+                        _l5.warning(f'[im_llm REQ] {ai_model.name}: url={api_url[:60]}, proxy={_proxy_for_url(api_url)}, api_key={api_key[:20]}..., messages_count={len(request_body.get("messages", []))}, tools_count={len(request_body.get("tools", []))}')
                         data = chat_complete(
                             api_url,
                             api_key,
@@ -235,40 +248,58 @@ def _call_real_llm(db: Session, user=None, prompt: str = "", platform: str = "im
                             proxy=_proxy_for_url(api_url),
                             timeout=ai_model.timeout or 120,
                         )
+                        _l5.warning(f'[im_llm RES] {ai_model.name}: data_keys={list(data.keys()) if data else None}, resp_type={type(data).__name__}')
                     except Exception as _exc:
                         _retry_log.append(f"{ai_model.name}: {str(_exc)[:60]}")
+                        import traceback as _tb
                         logger.warning(f"[im_llm] 模型 {ai_model.name} 异常: {_exc}")
+                        logger.warning(f"[im_llm] {ai_model.name} 异常详情: {_tb.format_exc()[:2000]}")
                         _http_ok = False
                         break
-                        message = data.get("choices", [{}])[0].get("message", {})
-                        tool_calls = message.get("tool_calls", [])
-                        if tool_calls and skills:
+                    message = data.get("choices", [{}])[0].get("message", {})
+                    tool_calls = message.get("tool_calls", [])
+                    # 2026-07-30 CRITICAL DEBUG
+                    import logging as _lcr
+                    _raw_content = message.get("content")
+                    _raw_reasoning = message.get("reasoning_content")
+                    _msg_keys = list(message.keys()) if message else []
+                    _tool_calls = tool_calls
+                    _lcr.warning(f'[CRITICAL] MODEL={ai_model.name} msg_keys={_msg_keys} content_len={len(_raw_content or "")} reasoning_len={len(_raw_reasoning or "")} has_tool_calls={bool(_tool_calls)}')
+                    ai_content = message.get("content", "") or message.get("reasoning_content") or ""
+                    if not ai_content and (message.get("reasoning_content") or ""):
+                        ai_content = f"（AI 思考：{message.get('reasoning_content')[:500]}...）"
+                    if tool_calls and skills:
+                        messages.append({
+                            "role": "assistant",
+                            "content": message.get("content") or message.get("reasoning_content") or "",
+                            "tool_calls": tool_calls,
+                        })
+                        for tc in tool_calls:
+                            func_name = tc.get("function", {}).get("name", "")
+                            try:
+                                func_args = json.loads(tc.get("function", {}).get("arguments", "{}"))
+                            except Exception:
+                                func_args = {"input": tc.get("function", {}).get("arguments", "")}
+                            tool_result = _execute_skill_tool(db, func_name, func_args, skills, prompt)
                             messages.append({
-                                "role": "assistant",
-                                "content": message.get("content") or "",
-                                "tool_calls": tool_calls,
+                                "role": "tool",
+                                "tool_call_id": tc.get("id", "call_1"),
+                                "content": tool_result,
                             })
-                            for tc in tool_calls:
-                                func_name = tc.get("function", {}).get("name", "")
-                                try:
-                                    func_args = json.loads(tc.get("function", {}).get("arguments", "{}"))
-                                except Exception:
-                                    func_args = {"input": tc.get("function", {}).get("arguments", "")}
-                                tool_result = _execute_skill_tool(db, func_name, func_args, skills, prompt)
-                                messages.append({
-                                    "role": "tool",
-                                    "tool_call_id": tc.get("id", "call_1"),
-                                    "content": tool_result,
-                                })
-                            continue
-                        ai_content = message.get("content", "")
-                        break
+                        if round_num >= max_rounds - 1:
+                            break
+                        continue
+                    break
 
                 # 检查本轮调用是否成功
+                import logging as _log2
+                _log2.warning(f'[im_llm trace] {ai_model.name}: http_ok={_http_ok}, ai_content_len={len(ai_content) if ai_content else 0}, ai_content_preview={repr(ai_content)[:200]}')
                 if _http_ok and ai_content:
                     return ai_content or "（AI 回复为空）"
                 # 本轮失败：记录日志，切换备选
-                logger.warning(f"[im_llm] 模型 {ai_model.name} 失败，切换备选...")
+                logger.warning('[im_llm] 模型 %s 失败 (http_ok=%s, content=%s), 切换备选...' % (
+                    ai_model.name, _http_ok, repr(ai_content)[:100]
+                ))
                 continue
             except Exception as exc:
                 _retry_log.append(f"{ai_model.name}: {str(exc)[:60]}")