Upload 4 files

app.py

@@ -90,10 +90,14 @@ class ResponseWrapper:
         self._data = data
         self._text = self._extract_text()
         self._finish_reason = self._extract_finish_reason()
+        if self.finish_reason != "STOP":
+            # 抛出错误
+            raise StopCandidateException(f"生成文本失败: {self.finish_reason}")
         self._prompt_token_count = self._extract_prompt_token_count()
         self._candidates_token_count = self._extract_candidates_token_count()
         self._total_token_count = self._extract_total_token_count()
         self._thoughts = self._extract_thoughts()
+        self._json_dumps = json.dumps(self._data, indent=4, ensure_ascii=False)
 
     def _extract_thoughts(self) -> Optional[str]:
         try:
@@ -157,6 +161,14 @@ class ResponseWrapper:
     def total_token_count(self) -> Optional[int]:
         return self._total_token_count
 
+    @property
+    def thoughts(self) -> Optional[str]:
+        return self._thoughts
+
+    @property
+    def json_dumps(self) -> str:
+        return self._json_dumps
+
 class APIKeyManager:
     def __init__(self):
         self.api_keys = re.findall(r"AIzaSy[a-zA-Z0-9_-]{33}", os.environ.get('KeyArray'))
@@ -392,7 +404,7 @@ def chat_completions():
     response_type = 'streamGenerateContent' if stream else 'generateContent'
     is_SSE = '&alt=sse' if stream else ''
 
-    gemini_history, system_instruction, error_response = func.process_messages_for_gemini(messages)
+    contents, system_instruction, error_response = func.process_messages_for_gemini(messages)
 
     if error_response:
         logger.error(f"处理输入消息时出错↙\n {error_response}")
@@ -414,7 +426,7 @@ def chat_completions():
         }
 
         data = {
-            "contents": gemini_history,
+            "contents": contents,
             "generationConfig": {
                 "temperature": temperature,
                 "maxOutputTokens": max_tokens,
@@ -429,15 +441,16 @@ def chat_completions():
             response.raise_for_status()
 
             if stream:
-
                 return 1, response
             else:
-
                 return 1, ResponseWrapper(response.json())
         except requests.exceptions.RequestException as e:
             return handle_api_error(e, attempt)
+        except StopCandidateException as e:
+            return handle_api_error(e, attempt)
 
     def generate_stream(response):
+        logger.info(f"流式开始 →")
         buffer = b""
         try:
             for line in response.iter_lines():
@@ -486,9 +499,10 @@ def chat_completions():
                     yield f"data: {json.dumps({'error': str(e)})}\n\n"
 
             yield f"data: {json.dumps({'choices': [{'delta': {}, 'finish_reason': 'stop', 'index': 0}]})}\n\n"
-
+            logger.info(f"流式结束 ←")
+            logger.info(f"200!")
         except Exception as e:
-            logger.error(f"Stream error: {e}")
+            logger.error(f"流式处理错误↙\n{e}")
             yield f"data: {json.dumps({'error': str(e)})}\n\n"
 
     attempt = 0
@@ -498,8 +512,8 @@ def chat_completions():
         logger.info(f"第 {attempt}/{MAX_RETRIES} 次尝试 ...")
         success, response = do_request(current_api_key, attempt)
 
-        if success == 1:
-            break
+        if success == 0:
+            continue
         elif success == 2:
 
             logger.error(f"{model} 很可能暂时不可用，请更换模型或未来一段时间再试")
@@ -511,6 +525,48 @@ def chat_completions():
             }
             return jsonify(response), 503
 
+        if stream:
+            return Response(
+                    stream_with_context(generate_stream(response)),
+                    mimetype='text/event-stream'
+                )
+        else:
+            try:
+                text_content = response.text
+                prompt_tokens = response.prompt_token_count
+                completion_tokens = response.candidates_token_count
+                total_tokens = response.total_token_count
+                finish_reason = response.finish_reason
+                json_dumps = response.json_dumps
+                logger.info(f"AI响应处理成功↓\n{json_dumps}")
+                if is_thinking and show_thoughts:
+                    text_content = response.thoughts + '\n' + text_content
+            except StopCandidateException as e:
+                logger.error(f"生成内容失败↙\n{e}")
+                continue
+
+            response_data = {
+                'id': 'chatcmpl-xxxxxxxxxxxx',  
+                'object': 'chat.completion',
+                'created': int(datetime.now().timestamp()),
+                'model': model,
+                'choices': [{
+                    'index': 0,
+                    'message': {
+                        'role': 'assistant',
+                        'content': text_content
+                    },
+                    'finish_reason': finish_reason
+                }],
+                'usage': {
+                    'prompt_tokens': prompt_tokens,
+                    'completion_tokens': completion_tokens,
+                    'total_tokens': total_tokens
+                }
+            }
+            logger.info(f"200!")
+            return jsonify(response_data)
+
     else:
         logger.error(f"{MAX_RETRIES} 次尝试均失败，请调整配置，或等待官方恢复，或向Moonfanz反馈")
         response = {
@@ -521,52 +577,6 @@ def chat_completions():
         }
         return jsonify(response), 500 if response is not None else 503
 
-    if stream:
-        return Response(
-                stream_with_context(generate_stream(response)),
-                mimetype='text/event-stream'
-            )
-    else:
-        try:
-            text_content = response.text
-            prompt_tokens = response.prompt_token_count
-            completion_tokens = response.candidates_token_count
-            total_tokens = response.total_token_count
-            finish_reason = response.finish_reason
-            if is_thinking and show_thoughts:
-                # 把thoughts加到text_content的前面再加一个回车
-                text_content = response.thoughts + '\n' + text_content
-            logger.info(f"finish_reason: {finish_reason}")
-        except AttributeError as e:
-            return jsonify({
-                    'error': {
-                        'message': 'AI响应处理失败',
-                        'type': 'response_processing_error'
-                    }
-                }), 500
-
-        response_data = {
-            'id': 'chatcmpl-xxxxxxxxxxxx',  
-            'object': 'chat.completion',
-            'created': int(datetime.now().timestamp()),
-            'model': model,
-            'choices': [{
-                'index': 0,
-                'message': {
-                    'role': 'assistant',
-                    'content': text_content
-                },
-                'finish_reason': finish_reason
-            }],
-            'usage': {
-                'prompt_tokens': prompt_tokens,
-                'completion_tokens': completion_tokens,
-                'total_tokens': total_tokens
-            }
-        }
-        logger.info(f"200!")
-        return jsonify(response_data)
-
 @app.route('/hf/v1/models', methods=['GET'])
 def list_models():
     response = {"object": "list", "data": GEMINI_MODELS}

func.py

@@ -35,7 +35,7 @@ def authenticate_request(request):
     return True, None, None
 
 def process_messages_for_gemini(messages):
-    gemini_history = []
+    contents = []
     errors = []
     system_instruction_text = ""
     is_system_phase = True
@@ -53,11 +53,11 @@ def process_messages_for_gemini(messages):
                 is_system_phase = False
 
                 if role == 'user':
-                    gemini_history.append({"role": "user", "parts": [{"text": content}]})
+                    contents.append({"role": "user", "parts": [{"text": content}]})
                 elif role == 'system':
-                    gemini_history.append({"role": "user", "parts": [{"text": content}]})
+                    contents.append({"role": "user", "parts": [{"text": content}]})
                 elif role == 'assistant':
-                    gemini_history.append({"role": "model", "parts": [{"text": content}]})
+                    contents.append({"role": "model", "parts": [{"text": content}]})
                 else:
                     errors.append(f"Invalid role: {role}")
         elif isinstance(content, list):
@@ -100,13 +100,13 @@ def process_messages_for_gemini(messages):
 
             if parts: 
                 if role in ['user', 'system']:
-                    gemini_history.append({"role": "user", "parts": parts})
+                    contents.append({"role": "user", "parts": parts})
                 elif role in ['assistant']:
-                    gemini_history.append({"role": "model", "parts": parts})
+                    contents.append({"role": "model", "parts": parts})
                 else:
                     errors.append(f"Invalid role: {role}")
 
     if errors:
-        return gemini_history, {"parts": [{"text": system_instruction_text}]}, (jsonify({'error': errors}), 400)
+        return contents, {"parts": [{"text": system_instruction_text}]}, (jsonify({'error': errors}), 400)
     else:
-        return gemini_history, {"parts": [{"text": system_instruction_text}]}, None
+        return contents, {"parts": [{"text": system_instruction_text}]}, None