From 53e9222211e9c0b1b39ed7246aa17d57bded6783 Mon Sep 17 00:00:00 2001 From: oadank Date: Sun, 17 May 2026 16:37:42 +0800 Subject: [PATCH] =?UTF-8?q?fix:=20=E5=90=8C=E6=AD=A5=20async=20=E7=89=88?= =?UTF-8?q?=E6=9C=AC=E7=B2=BE=E7=BB=86=E5=8C=96=E8=BF=87=E6=BB=A4=20+=20?= =?UTF-8?q?=E7=A7=BB=E9=99=A4=E6=95=8F=E6=84=9F=E6=97=A5=E5=BF=97?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit 1. async 版本同步 sync 版本的精细化过滤： - 移除 reasoning_effort, parallel_tool_calls 等过滤（厂商支持/忽略） - 保留必要过滤：client_metadata, coding_plan, functions 等 2. 移除敏感日志输出： - 删除 tool-call arguments 前 50 字符的日志 - 删除 api_base 脱敏不完全的日志 - 删除大量 DEBUG 日志（参数状态、tools 列表等） 安全修复：响应 Greptile AI 审查问题 - Medium: Tool-call arguments logged → 已修复 - Medium: api_base 日志未脱敏 → 已删除日志 测试验证：小米/DeepSeek 正常响应 Co-Authored-By: Claude Opus 4.7 --- .../handler.py | 138 ++++-------------- 1 file changed, 31 insertions(+), 107 deletions(-) diff --git a/litellm/responses/litellm_completion_transformation/handler.py b/litellm/responses/litellm_completion_transformation/handler.py index 7a2b20e9ac0..3cd2d8e770b 100644 --- a/litellm/responses/litellm_completion_transformation/handler.py +++ b/litellm/responses/litellm_completion_transformation/handler.py @@ -116,12 +116,7 @@ class LiteLLMCompletionTransformationHandler: fixed_args = "{}" needs_rebuild = True fixed_count += 1 - args_preview = ( - str(original_args)[:50] if original_args else "None" - ) - logger.warning( - f"[DomesticFilter] Fixed invalid JSON arguments, original: {args_preview}" - ) + # 安全：不记录 original_args 内容，可能包含敏感信息 elif isinstance(args, dict): # dict 类型需要转成 JSON string fixed_args = json.dumps(args) @@ -396,39 +391,22 @@ class LiteLLMCompletionTransformationHandler: actual_model = litellm_params_model if is_domestic_model_or_endpoint(actual_model, api_base): - # DEBUG logger（先定义，后面使用） + # DEBUG logger logger = logging.getLogger("LiteLLM.DomesticFilter") - # Codex CLI 特有参数 + # ========== 必须过滤的参数 ========== + + # Codex CLI 特有参数 - LiteLLM 内部报错 acompletion_args.pop("client_metadata", None) - # OpenAI 扩展参数（国内模型不支持） - acompletion_args.pop("reasoning_effort", None) # reasoning 模式 - acompletion_args.pop("reasoning", None) # reasoning 参数 (Codex 0.130.0+) + + # Responses API 专用参数（国内厂商 Chat API 不支持） acompletion_args.pop("coding_plan", None) # Codex coding plan - acompletion_args.pop("parallel_tool_calls", None) # 并行工具调用 - # 旧格式参数（国内模型只支持 tools/tool_choice，不支持 functions/function_call） - acompletion_args.pop("functions", None) # OpenAI 旧格式 - acompletion_args.pop("function_call", None) # OpenAI 旧格式 - # 其他国内模型可能不支持的参数 - acompletion_args.pop("stream_options", None) # stream 选项 - acompletion_args.pop("modalities", None) # 多模态参数 - acompletion_args.pop("prediction", None) # 预测参数 - acompletion_args.pop("audio", None) # 音频参数 - acompletion_args.pop("store", None) # 存储参数 - acompletion_args.pop("include", None) # 包含参数 - acompletion_args.pop("prompt_cache_key", None) # 缓存键 - # OpenAI 高级参数（国内模型可能不支持） - acompletion_args.pop("frequency_penalty", None) # 频率惩罚 - acompletion_args.pop("presence_penalty", None) # 存在惩罚 - acompletion_args.pop("logprobs", None) # logprobs - acompletion_args.pop("top_logprobs", None) # top_logprobs - acompletion_args.pop("response_format", None) # 响应格式（如 json_object） - acompletion_args.pop("seed", None) # 随机种子 - acompletion_args.pop("logit_bias", None) # logit bias - acompletion_args.pop("n", None) # 返回多个结果 - acompletion_args.pop("service_tier", None) # 服务等级 - # 注意：litellm_* 内部参数必须保留，用于代理会计/spend attribution - # 这些参数会被 litellm.completion 正确处理，不会发送给上游 API + + # 旧格式参数（国内模型只支持 tools/tool_choice） + acompletion_args.pop("functions", None) + acompletion_args.pop("function_call", None) + + # LiteLLM 内部参数（不应发送给上游） acompletion_args.pop("shared_session", None) acompletion_args.pop("model_info", None) acompletion_args.pop("secret_fields", None) @@ -436,35 +414,32 @@ class LiteLLMCompletionTransformationHandler: acompletion_args.pop("use_litellm_proxy", None) acompletion_args.pop("merge_reasoning_content_in_choices", None) acompletion_args.pop("supports_function_calling", None) + acompletion_args.pop("max_retries", None) + + # ========== 厂商忽略的参数（保留过滤，避免无意义传输）========== + acompletion_args.pop("stream_options", None) + acompletion_args.pop("modalities", None) + acompletion_args.pop("prediction", None) + acompletion_args.pop("audio", None) + acompletion_args.pop("store", None) + acompletion_args.pop("include", None) + acompletion_args.pop("prompt_cache_key", None) acompletion_args.pop("caching", None) acompletion_args.pop("extra_body", None) - # 注意：custom_llm_provider 必须保留，LiteLLM 需要它识别提供商 - acompletion_args.pop("max_retries", None) + + # ========== 不过滤的参数（厂商支持或忽略）========== + # reasoning_effort - DeepSeek/Xiaomi 支持，其他厂商忽略 + # parallel_tool_calls - 厂商支持/忽略 + # frequency_penalty, presence_penalty - 厂商支持 + # response_format, seed, logprobs 等 - 厂商支持/忽略 + # tool_choice 兼容：国内模型不支持 "required"，改成 "auto" tool_choice = acompletion_args.get("tool_choice") - logger.info(f"[DomesticFilter] tool_choice before fix: {tool_choice}") if tool_choice == "required": acompletion_args["tool_choice"] = "auto" logger.info("[DomesticFilter] tool_choice converted: required -> auto") - # DEBUG: 国内模型参数过滤验证（脱敏 api_base，只显示域名） - # 安全：api_base 可能包含凭证或内部主机名，只显示基础域名 - safe_api_base = ( - api_base.split("//")[-1].split("/")[0].split("?")[0] - if api_base - else "none" - ) - logger.info( - f"[DomesticFilter] actual_model={actual_model}, api_base_host={safe_api_base}, " - f"is_domestic={is_domestic_model_or_endpoint(actual_model, api_base)}" - ) - logger.info( - f"[DomesticFilter] acompletion_args keys after filter: " - f"{list(acompletion_args.keys())}" - ) - - # 确保所有历史消息中的 tool_calls arguments 是有效 JSON 格式 - # 国内模型要求 function.arguments 必须是严格 JSON + # 确保 tool_calls arguments 是有效 JSON messages = acompletion_args.get("messages") if messages: acompletion_args["messages"] = ( @@ -472,57 +447,6 @@ class LiteLLMCompletionTransformationHandler: messages ) ) - logger.info( - "[DomesticFilter] Validated tool_calls arguments in messages" - ) - - # 打印 tools 和 messages 用于调试 - tools_debug = acompletion_args.get("tools") - messages_debug = acompletion_args.get("messages") - logger.info( - f"[DomesticFilter] tools count: {len(tools_debug) if tools_debug else 0}" - ) - if tools_debug: - for i, tool in enumerate(tools_debug[:3]): # 只打印前3个 - tool_name = ( - tool.get("function", {}).get("name", "unknown") - if isinstance(tool, dict) - else "unknown" - ) - logger.info(f"[DomesticFilter] tool[{i}] name: {tool_name}") - logger.info( - f"[DomesticFilter] messages count: {len(messages_debug) if messages_debug else 0}" - ) - if messages_debug: - for i, msg in enumerate(messages_debug[:5]): # 只打印前5条 - role = ( - msg.get("role", "unknown") - if isinstance(msg, dict) - else "unknown" - ) - has_tools = ( - "tool_calls" in msg if isinstance(msg, dict) else False - ) - logger.info( - f"[DomesticFilter] msg[{i}] role={role}, has_tool_calls={has_tools}" - ) - - # 打印第一个有 tool_calls 的消息详情（不含 arguments，安全） - for i, msg in enumerate(messages_debug or []): - if isinstance(msg, dict) and msg.get("tool_calls"): - tc = msg.get("tool_calls", []) - if tc and isinstance(tc, list) and len(tc) > 0: - first_tc = tc[0] - func_name = ( - first_tc.get("function", {}).get("name", "unknown") - if isinstance(first_tc, dict) - else "unknown" - ) - # 安全：不打印 arguments，可能包含敏感内容 - logger.info( - f"[DomesticFilter] First tool_call msg[{i}] function={func_name}" - ) - break litellm_completion_response: Union[ ModelResponse, litellm.CustomStreamWrapper