diff --git a/src/config/config.py b/src/config/config.py index e3ecedb9..11d59935 100644 --- a/src/config/config.py +++ b/src/config/config.py @@ -1,7 +1,7 @@ import os import re from dataclasses import dataclass, field -from typing import Dict, List, Optional, Any +from typing import Dict, List, Optional from dateutil import tz import tomli diff --git a/src/plugins/models/utils_model.py b/src/plugins/models/utils_model.py index 2447bfac..085f948a 100644 --- a/src/plugins/models/utils_model.py +++ b/src/plugins/models/utils_model.py @@ -23,7 +23,6 @@ import base64 from PIL import Image import io import os -from dotenv import load_dotenv # 导入 dotenv 用于加载 .env 文件 (如果需要直接加载) from rich.traceback import install @@ -113,12 +112,12 @@ async def _safely_record(request_content: Dict[str, Any], payload: Dict[str, Any f"{image_base64[:10]}...{image_base64[-10:]}" ) elif is_gemini_payload and "contents" in safe_payload and len(safe_payload["contents"]) > 0: - if isinstance(safe_payload["contents"][0], dict) and "parts" in safe_payload["contents"][0]: - parts = safe_payload["contents"][0]["parts"] - for i, part in enumerate(parts): - if isinstance(part, dict) and "inlineData" in part: - safe_payload["contents"][0]["parts"][i]["inlineData"]["data"] = f"{image_base64[:10]}...{image_base64[-10:]}" - break + if isinstance(safe_payload["contents"][0], dict) and "parts" in safe_payload["contents"][0]: + parts = safe_payload["contents"][0]["parts"] + for i, part in enumerate(parts): + if isinstance(part, dict) and "inlineData" in part: + safe_payload["contents"][0]["parts"][i]["inlineData"]["data"] = f"{image_base64[:10]}...{image_base64[-10:]}" + break return safe_payload @@ -165,14 +164,14 @@ class LLMRequest: parsed_keys = [loaded_keys] else: raise ValueError(f"Parsed API key for {self.model_key_name} is not a valid list or string.") - except (json.JSONDecodeError, TypeError): + except (json.JSONDecodeError, TypeError) as e: if isinstance(raw_api_key_config, list): parsed_keys = [str(key) for key in raw_api_key_config if key] is_list_config = True elif isinstance(raw_api_key_config, str) and raw_api_key_config: parsed_keys = [raw_api_key_config] else: - raise ValueError(f"Invalid or empty API key config for {self.model_key_name}: {raw_api_key_config}") + raise ValueError(f"Invalid or empty API key config for {self.model_key_name}: {raw_api_key_config}") from e if not parsed_keys: raise ValueError(f"No valid API keys found for {self.model_key_name}.") @@ -238,9 +237,9 @@ class LLMRequest: logger.error(f"解析 PROXY_MODELS ('{proxy_models_str}') 出错: {e}. 代理将不会对特定模型生效。") self.proxy_models_set = set() except ValueError: - logger.error(f"无效的代理端口号: {proxy_port}。代理将不被启用。") - self.proxy_url = None - self.proxy_models_set = set() + logger.error(f"无效的代理端口号: {proxy_port}。代理将不被启用。") + self.proxy_url = None + self.proxy_models_set = set() except Exception as e: logger.error(f"加载代理配置时发生错误: {e}") self.proxy_url = None @@ -352,7 +351,7 @@ class LLMRequest: } policy = {**default_retry, **(retry_policy or {})} - actual_endpoint = endpoint + _actual_endpoint = endpoint if self.is_gemini: action = endpoint.lstrip('/') api_url = f"{self.base_url.rstrip('/')}/{self.model_name}{action}" @@ -486,10 +485,10 @@ class LLMRequest: keys_failed_429.add(current_key) logger.info(f" (因 429 已失败 {len(keys_failed_429)}/{key_switch_limit_429} 个不同 Key)") if available_keys_pool and len(keys_failed_429) < key_switch_limit_429: - logger.info(f" 尝试因 429 切换到下一个可用 Key...") + logger.info(" 尝试因 429 切换到下一个可用 Key...") raise _SwitchKeyException() else: - logger.warning(f" 无更多 Key 可因 429 切换或已达上限。") + logger.warning(" 无更多 Key 可因 429 切换或已达上限。") else: logger.warning(f" Key ...{current_key[-4:]} 再次遇到 429，按标准重试流程。") @@ -499,12 +498,13 @@ class LLMRequest: keys_abandoned_runtime.add(current_key) LLMRequest._abandoned_keys_runtime.add(current_key) logger.critical(f" !! Key ...{current_key[-4:]} 已添加到运行时废弃列表。请考虑将其移至配置中的 'abandon_{self.model_key_name}' !!") - if current_key in available_keys_pool: available_keys_pool.remove(current_key) + if current_key in available_keys_pool: + available_keys_pool.remove(current_key) if available_keys_pool and len(keys_abandoned_runtime) < key_switch_limit_403: - logger.info(f" 尝试因 403 切换到下一个可用 Key...") + logger.info(" 尝试因 403 切换到下一个可用 Key...") raise _SwitchKeyException() else: - logger.error(f" 无更多 Key 可因 403 切换或已达上限。将中止请求。") + logger.error(" 无更多 Key 可因 403 切换或已达上限。将中止请求。") await response.read() raise PermissionDeniedException(f"Key ...{current_key[-4:]} 权限被拒，且无其他可用 Key 切换。", key_identifier=current_key) else: @@ -784,7 +784,7 @@ class LLMRequest: """处理非 HTTP 错误，支持使用合并后的参数重建 payload""" policy = request_content["policy"] payload = request_content["payload"] - wait_time = policy["base_wait"] * (2**retry_count) + _wait_time = policy["base_wait"] * (2**retry_count) keep_request = False if retry_count < policy["max_retries"] - 1: keep_request = True @@ -944,10 +944,10 @@ class LLMRequest: **params_copy, } if "max_tokens" not in payload and "max_completion_tokens" not in payload: - if "max_tokens" not in params_copy and "max_completion_tokens" not in params_copy: - payload["max_tokens"] = global_config.model_max_output_length + if "max_tokens" not in params_copy and "max_completion_tokens" not in params_copy: + payload["max_tokens"] = global_config.model_max_output_length if "max_completion_tokens" in payload: - payload["max_tokens"] = payload.pop("max_completion_tokens") + payload["max_tokens"] = payload.pop("max_completion_tokens") return payload @@ -992,13 +992,13 @@ class LLMRequest: finish_reason = candidate.get("finishReason") if finish_reason == "SAFETY": - logger.warning(f"模型 {self.model_name}: Gemini 响应因安全设置被阻止。") - content = "响应内容因安全原因被过滤。" + logger.warning(f"模型 {self.model_name}: Gemini 响应因安全设置被阻止。") + content = "响应内容因安全原因被过滤。" elif finish_reason == "RECITATION": - logger.warning(f"模型 {self.model_name}: Gemini 响应因引用限制被阻止。") - content = "响应内容因引用限制被过滤。" + logger.warning(f"模型 {self.model_name}: Gemini 响应因引用限制被阻止。") + content = "响应内容因引用限制被过滤。" elif finish_reason == "OTHER": - logger.warning(f"模型 {self.model_name}: Gemini 响应因未知原因停止。") + logger.warning(f"模型 {self.model_name}: Gemini 响应因未知原因停止。") # finishReason == "TOOL_CODE" or "FUNCTION_CALL" 是正常情况 usage = result.get("usageMetadata", {}) @@ -1043,16 +1043,16 @@ class LLMRequest: # --- 记录 Token 使用情况 --- # (代码不变) if prompt_tokens > 0 or completion_tokens > 0 or total_tokens > 0: - self._record_usage( - prompt_tokens=prompt_tokens, - completion_tokens=completion_tokens, - total_tokens=total_tokens, - user_id=user_id, - request_type=request_type, - endpoint=endpoint, - ) + self._record_usage( + prompt_tokens=prompt_tokens, + completion_tokens=completion_tokens, + total_tokens=total_tokens, + user_id=user_id, + request_type=request_type, + endpoint=endpoint, + ) else: - logger.warning(f"模型 {self.model_name}: 未能从响应中提取有效的 token 使用信息。") + logger.warning(f"模型 {self.model_name}: 未能从响应中提取有效的 token 使用信息。") # --- 返回结果 (统一格式) --- @@ -1111,7 +1111,7 @@ class LLMRequest: else: if not api_key: logger.error(f"尝试使用无效 (空) 的 API key 为模型 {self.model_name} 构建请求头。") - raise ValueError(f"无效的 API key 提供给 _build_headers。") + raise ValueError("无效的 API key 提供给 _build_headers。") if self.is_gemini: return {"x-goog-api-key": api_key, "Content-Type": "application/json"}