fix: 修复速率限制bug并添加日志

- 修复每分钟最大请求次数限制失效的问题
- 将时间戳更新从API调用前移到调用后，确保重试请求也被正确计入速率限制
- 添加速率限制等待日志，方便调试
- 影响翻译器：OpenAI、OpenAI HQ、Gemini、Gemini HQ

Author: hgmzhn

doc/CHANGELOG_v1.8.9.md

@@ -15,6 +15,13 @@
 - 修复 web 模式缺少 nonce 和 start_instance 属性的错误
 - 修复 web 模式命令行参数未生效的问题
 
+### 翻译器修复
+- **修复速率限制bug**：修复每分钟最大请求次数限制失效的问题
+  - 问题：当API调用失败（网络错误、SSL错误等）时，时间戳在请求发送前就已更新，导致重试请求没有遵守速率限制
+  - 修复：将时间戳更新移到API调用成功后，确保只有真正完成的请求才计入速率限制
+  - 影响翻译器：OpenAI、OpenAI HQ、Gemini、Gemini HQ
+- **添加速率限制日志**：当触发速率限制等待时，会打印 `Ratelimit sleep: X.XXs` 日志，方便调试
+
 ## 🎯 新增功能
 
 ### API 功能增强

manga_translator/translators/gemini.py

@@ -300,13 +300,18 @@ def generate_content_with_logging(**kwargs):
                     delay = 60.0 / self._MAX_REQUESTS_PER_MINUTE
                     elapsed = now - GeminiTranslator._GLOBAL_LAST_REQUEST_TS[self._last_request_ts_key]
                     if elapsed < delay:
-                        await asyncio.sleep(delay - elapsed)
-                    GeminiTranslator._GLOBAL_LAST_REQUEST_TS[self._last_request_ts_key] = time.time()
+                        sleep_time = delay - elapsed
+                        self.logger.info(f'Ratelimit sleep: {sleep_time:.2f}s')
+                        await asyncio.sleep(sleep_time)
 
                 response = await asyncio.to_thread(
                     generate_content_with_logging,
                     **request_args
                 )
+                
+                # 在API调用成功后立即更新时间戳，确保所有请求（包括重试）都被计入速率限制
+                if self._MAX_REQUESTS_PER_MINUTE > 0:
+                    GeminiTranslator._GLOBAL_LAST_REQUEST_TS[self._last_request_ts_key] = time.time()
 
                 # 检查finish_reason，只有成功(1)才继续，其他都重试
                 if hasattr(response, 'candidates') and response.candidates:

manga_translator/translators/gemini_hq.py

@@ -325,14 +325,18 @@ def generate_content_with_logging(**kwargs):
                     delay = 60.0 / self._MAX_REQUESTS_PER_MINUTE
                     elapsed = now - GeminiHighQualityTranslator._GLOBAL_LAST_REQUEST_TS[self._last_request_ts_key]
                     if elapsed < delay:
-                        await asyncio.sleep(delay - elapsed)
-                    # 在请求前更新时间戳,确保下次计算的是从这次请求开始的间隔
-                    GeminiHighQualityTranslator._GLOBAL_LAST_REQUEST_TS[self._last_request_ts_key] = time.time()
+                        sleep_time = delay - elapsed
+                        self.logger.info(f'Ratelimit sleep: {sleep_time:.2f}s')
+                        await asyncio.sleep(sleep_time)
 
                 response = await asyncio.to_thread(
                     generate_content_with_logging,
                     **request_args
                 )
+                
+                # 在API调用成功后立即更新时间戳，确保所有请求（包括重试）都被计入速率限制
+                if self._MAX_REQUESTS_PER_MINUTE > 0:
+                    GeminiHighQualityTranslator._GLOBAL_LAST_REQUEST_TS[self._last_request_ts_key] = time.time()
 
                 # 检查finish_reason，只有成功(1)才继续，其他都重试
                 if hasattr(response, 'candidates') and response.candidates:
@@ -550,15 +554,19 @@ def generate_content_with_logging(**kwargs):
                 delay = 60.0 / self._MAX_REQUESTS_PER_MINUTE
                 elapsed = now - GeminiHighQualityTranslator._GLOBAL_LAST_REQUEST_TS[self._last_request_ts_key]
                 if elapsed < delay:
-                    await asyncio.sleep(delay - elapsed)
-                # 在请求前更新时间戳,确保下次计算的是从这次请求开始的间隔
-                GeminiHighQualityTranslator._GLOBAL_LAST_REQUEST_TS[self._last_request_ts_key] = time.time()
+                    sleep_time = delay - elapsed
+                    self.logger.info(f'Ratelimit sleep: {sleep_time:.2f}s')
+                    await asyncio.sleep(sleep_time)
 
             response = await asyncio.to_thread(
                 generate_content_with_logging,
                 **request_args
             )
 
+            # 在API调用成功后立即更新时间戳，确保所有请求（包括重试）都被计入速率限制
+            if self._MAX_REQUESTS_PER_MINUTE > 0:
+                GeminiHighQualityTranslator._GLOBAL_LAST_REQUEST_TS[self._last_request_ts_key] = time.time()
+            
             if response and response.text:
                 result = response.text.strip()
                 translations = result.split('\n')

manga_translator/translators/openai.py

@@ -228,15 +228,20 @@ async def _translate_batch(self, texts: List[str], source_lang: str, target_lang
                     delay = 60.0 / self._MAX_REQUESTS_PER_MINUTE
                     elapsed = now - OpenAITranslator._GLOBAL_LAST_REQUEST_TS[self._last_request_ts_key]
                     if elapsed < delay:
-                        await asyncio.sleep(delay - elapsed)
-                    OpenAITranslator._GLOBAL_LAST_REQUEST_TS[self._last_request_ts_key] = time.time()
+                        sleep_time = delay - elapsed
+                        self.logger.info(f'Ratelimit sleep: {sleep_time:.2f}s')
+                        await asyncio.sleep(sleep_time)
 
                 response = await self.client.chat.completions.create(
                     model=self.model,
                     messages=messages,
                     max_tokens=self.max_tokens,
                     temperature=self.temperature
                 )
+                
+                # 在API调用成功后立即更新时间戳，确保所有请求（包括重试）都被计入速率限制
+                if self._MAX_REQUESTS_PER_MINUTE > 0:
+                    OpenAITranslator._GLOBAL_LAST_REQUEST_TS[self._last_request_ts_key] = time.time()
 
                 # 检查成功条件
                 if response.choices and response.choices[0].message.content and response.choices[0].finish_reason != 'content_filter':

manga_translator/translators/openai_hq.py

@@ -296,16 +296,20 @@ async def _translate_batch_high_quality(self, texts: List[str], batch_data: List
                     delay = 60.0 / self._MAX_REQUESTS_PER_MINUTE
                     elapsed = now - OpenAIHighQualityTranslator._GLOBAL_LAST_REQUEST_TS[self._last_request_ts_key]
                     if elapsed < delay:
-                        await asyncio.sleep(delay - elapsed)
-                    # 在请求前更新时间戳,确保下次计算的是从这次请求开始的间隔
-                    OpenAIHighQualityTranslator._GLOBAL_LAST_REQUEST_TS[self._last_request_ts_key] = time.time()
+                        sleep_time = delay - elapsed
+                        self.logger.info(f'Ratelimit sleep: {sleep_time:.2f}s')
+                        await asyncio.sleep(sleep_time)
 
                 response = await self.client.chat.completions.create(
                     model=self.model,
                     messages=messages,
                     max_tokens=self.max_tokens,
                     temperature=self.temperature
                 )
+                
+                # 在API调用成功后立即更新时间戳，确保所有请求（包括重试）都被计入速率限制
+                if self._MAX_REQUESTS_PER_MINUTE > 0:
+                    OpenAIHighQualityTranslator._GLOBAL_LAST_REQUEST_TS[self._last_request_ts_key] = time.time()
 
                 # 检查成功条件
                 if response.choices and response.choices[0].message.content and response.choices[0].finish_reason != 'content_filter':
@@ -485,9 +489,9 @@ async def _translate(self, from_lang: str, to_lang: str, queries: List[str], ctx
                 delay = 60.0 / self._MAX_REQUESTS_PER_MINUTE
                 elapsed = now - OpenAIHighQualityTranslator._GLOBAL_LAST_REQUEST_TS[self._last_request_ts_key]
                 if elapsed < delay:
-                    await asyncio.sleep(delay - elapsed)
-                # 在请求前更新时间戳,确保下次计算的是从这次请求开始的间隔
-                OpenAIHighQualityTranslator._GLOBAL_LAST_REQUEST_TS[self._last_request_ts_key] = time.time()
+                    sleep_time = delay - elapsed
+                    self.logger.info(f'Ratelimit sleep: {sleep_time:.2f}s')
+                    await asyncio.sleep(sleep_time)
 
             response = await self.client.chat.completions.create(
                 model=self.model,
@@ -496,6 +500,10 @@ async def _translate(self, from_lang: str, to_lang: str, queries: List[str], ctx
                 temperature=self.temperature
             )
 
+            # 在API调用成功后立即更新时间戳，确保所有请求（包括重试）都被计入速率限制
+            if self._MAX_REQUESTS_PER_MINUTE > 0:
+                OpenAIHighQualityTranslator._GLOBAL_LAST_REQUEST_TS[self._last_request_ts_key] = time.time()
+            
             if response.choices and response.choices[0].message.content:
                 result = response.choices[0].message.content.strip()
                 translations = result.split('\n')