feat: 增强工具调用代理功能，支持多工具调用和消息历史转换

主要改进:
- 新增 convert_tool_calls_to_content 函数，将消息历史中的 tool_calls 转换为 LLM 可理解的 XML 格式
- 修复 response_parser 支持同时解析多个 tool_calls
- 优化响应解析逻辑，支持 content 和 tool_calls 同时存在
- 添加完整的测试覆盖，包括多工具调用、消息转换和混合响应

技术细节:
- services.py: 实现工具调用历史到 content 的转换
- response_parser.py: 使用非贪婪匹配支持多个 tool_calls 解析
- main.py: 集成消息转换功能，确保消息历史正确传递给 LLM

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

app/main.py

@@ -8,7 +8,7 @@ from fastapi import FastAPI, HTTPException, Depends, Request
 from starlette.responses import StreamingResponse
 
 from .models import IncomingRequest, ProxyResponse
-from .services import process_chat_request, stream_llm_api, inject_tools_into_prompt, parse_llm_response_from_content, _parse_sse_data
+from .services import process_chat_request, stream_llm_api, inject_tools_into_prompt, parse_llm_response_from_content, _parse_sse_data, convert_tool_calls_to_content
 from .core.config import get_settings, Settings
 from .database import init_db, log_request, update_request_log
 
@@ -87,8 +87,13 @@ async def chat_completions(
         raise HTTPException(status_code=500, detail="LLM API Key or URL is not configured.")
 
     messages_to_llm = request_obj.messages
+
+    # Convert assistant messages with tool_calls to content format
+    messages_to_llm = convert_tool_calls_to_content(messages_to_llm)
+    logger.info(f"Converted tool calls to content format for log ID: {log_id}")
+
     if request_obj.tools:
-        messages_to_llm = inject_tools_into_prompt(request_obj.messages, request_obj.tools)
+        messages_to_llm = inject_tools_into_prompt(messages_to_llm, request_obj.tools)
 
     # Handle streaming request
     if request_obj.stream: