Upload 9 files

README.md

@@ -55,7 +55,12 @@ curl "https://huggingface.co/spaces/YOUR_USERNAME/YOUR_SPACE_NAME/v1/chat/comple
     -   **格式**: 使用逗号分隔。
     -   **示例**: `sk-mykey-1,sk-mykey-2`
 
-3.  **`DEBUG_MODE`** (可选)
+3.  **`HTTP_PROXY`** (可选)
+    -   **说明**: 为所有出站请求（例如到 Tenbin API）设置 HTTP/HTTPS 代理。
+    -   **格式**: 标准的代理 URL。
+    -   **示例**: `http://user:pass@host:port`
+
+4.  **`DEBUG_MODE`** (可选)
     -   **说明**: 设置为 `true` 可以启用详细的调试日志。
     -   **格式**: `true` 或 `false`。
     -   **示例**: `true`

getCaptcha.py

@@ -1,29 +1,29 @@
-import requests
-import time
-import os
-
-TURNSTILE_SOLVER_URL = os.environ.get("TURNSTILE_SOLVER_URL", "http://127.0.0.1:5000")
-
-def getTaskId():
-    url = f"{TURNSTILE_SOLVER_URL}/turnstile?url=https://tenbin.ai/workspace&sitekey=0x4AAAAAABGR2exxRproizri&action=issue_execution_token"
-
-    response = requests.get(url)
-    response.raise_for_status()
-    return response.json()['task_id']
-
-def getCaptcha(task_id):
-
-    url = f"{TURNSTILE_SOLVER_URL}/result?id={task_id}"
-    
-    while True:
-        try:
-            response = requests.get(url)
-            response.raise_for_status()
-            captcha = response.json().get('value', None)
-            if captcha:
-                return captcha
-            else:
-                time.sleep(1)
-        except Exception as e:
-            print(e)
+import requests
+import time
+import os
+
+TURNSTILE_SOLVER_URL = os.environ.get("TURNSTILE_SOLVER_URL", "http://127.0.0.1:5000")
+
+def getTaskId():
+    url = f"{TURNSTILE_SOLVER_URL}/turnstile?url=https://tenbin.ai/workspace&sitekey=0x4AAAAAABGR2exxRproizri&action=issue_execution_token"
+
+    response = requests.get(url)
+    response.raise_for_status()
+    return response.json()['task_id']
+
+def getCaptcha(task_id):
+
+    url = f"{TURNSTILE_SOLVER_URL}/result?id={task_id}"
+    
+    while True:
+        try:
+            response = requests.get(url)
+            response.raise_for_status()
+            captcha = response.json().get('value', None)
+            if captcha:
+                return captcha
+            else:
+                time.sleep(1)
+        except Exception as e:
+            print(e)
             time.sleep(1)

main.py

@@ -1,688 +1,699 @@
-import json
-import os
-import time
-import uuid
-import threading
-from typing import Any, Dict, List, Optional, TypedDict, Union
-
-import requests
-import websocket
-from fastapi import FastAPI, HTTPException, Depends, Query
-from fastapi.responses import StreamingResponse
-from fastapi.security import HTTPBearer, HTTPAuthorizationCredentials
-from pydantic import BaseModel, Field
-
-from getCaptcha import getCaptcha, getTaskId
-
-
-# Tenbin Account Management
-class TenbinAccount(TypedDict):
-    session_id: str
-    is_valid: bool
-    last_used: float
-    error_count: int
-
-
-# Global variables
-VALID_CLIENT_KEYS: set = set()
-TENBIN_ACCOUNTS: List[TenbinAccount] = []
-TENBIN_MODELS: Dict[str, str] = {}  # 模型映射表，key 是模型名称，value 是内部模型 ID
-account_rotation_lock = threading.Lock()
-MAX_ERROR_COUNT = 3
-ERROR_COOLDOWN = 300  # 5 minutes cooldown for accounts with errors
-DEBUG_MODE = os.environ.get("DEBUG_MODE", "false").lower() == "true"
-REQUEST_TIMEOUT = 120.0  # 请求超时时间，秒
-
-
-# Pydantic Models
-class ChatMessage(BaseModel):
-    role: str
-    content: Union[str, List[Dict[str, Any]]]
-    reasoning_content: Optional[str] = None
-
-
-class ChatCompletionRequest(BaseModel):
-    model: str
-    messages: List[ChatMessage]
-    stream: bool = True
-    temperature: Optional[float] = None
-    max_tokens: Optional[int] = None
-    top_p: Optional[float] = None
-    raw_response: bool = False  # 是否返回原始响应
-
-
-class ModelInfo(BaseModel):
-    id: str
-    object: str = "model"
-    created: int
-    owned_by: str = "tenbin"
-
-
-class ModelList(BaseModel):
-    object: str = "list"
-    data: List[ModelInfo]
-
-
-class ChatCompletionChoice(BaseModel):
-    message: ChatMessage
-    index: int = 0
-    finish_reason: str = "stop"
-
-
-class ChatCompletionResponse(BaseModel):
-    id: str = Field(default_factory=lambda: f"chatcmpl-{uuid.uuid4().hex}")
-    object: str = "chat.completion"
-    created: int = Field(default_factory=lambda: int(time.time()))
-    model: str
-    choices: List[ChatCompletionChoice]
-    usage: Dict[str, int] = Field(
-        default_factory=lambda: {
-            "prompt_tokens": 0,
-            "completion_tokens": 0,
-            "total_tokens": 0,
-        }
-    )
-
-
-class StreamChoice(BaseModel):
-    delta: Dict[str, Any] = Field(default_factory=dict)
-    index: int = 0
-    finish_reason: Optional[str] = None
-
-
-class StreamResponse(BaseModel):
-    id: str = Field(default_factory=lambda: f"chatcmpl-{uuid.uuid4().hex}")
-    object: str = "chat.completion.chunk"
-    created: int = Field(default_factory=lambda: int(time.time()))
-    model: str
-    choices: List[StreamChoice]
-
-
-# FastAPI App
-app = FastAPI(title="Tenbin OpenAI API Adapter")
-security = HTTPBearer(auto_error=False)
-
-
-def log_debug(message: str):
-    """Debug日志函数"""
-    if DEBUG_MODE:
-        print(f"[DEBUG] {message}")
-
-
-def load_client_api_keys():
-    """Load client API keys from environment variable (comma-separated) or client_api_keys.json"""
-    global VALID_CLIENT_KEYS
-    keys_str = os.environ.get("API_KEYS")
-    try:
-        if keys_str:
-            print("Loading client API keys from API_KEYS environment variable.")
-            keys = [key.strip() for key in keys_str.split(',')]
-        else:
-            print("Loading client API keys from file: client_api_keys.json")
-            with open("client_api_keys.json", "r", encoding="utf-8") as f:
-                keys = json.load(f)
-
-        VALID_CLIENT_KEYS = set(keys) if isinstance(keys, list) else set()
-        print(f"Successfully loaded {len(VALID_CLIENT_KEYS)} client API keys.")
-    except FileNotFoundError:
-        print("Error: client_api_keys.json not found and API_KEYS not set. Client authentication will fail.")
-        VALID_CLIENT_KEYS = set()
-    except Exception as e:
-        print(f"Error loading client API keys: {e}")
-        VALID_CLIENT_KEYS = set()
-
-
-def load_tenbin_accounts():
-    """Load Tenbin accounts from environment variable (comma-separated session_ids) or tenbin.json"""
-    global TENBIN_ACCOUNTS
-    TENBIN_ACCOUNTS = []
-    session_ids_str = os.environ.get("SESSION_IDS")
-    try:
-        accounts_to_process = []
-        if session_ids_str:
-            print("Loading Tenbin accounts from SESSION_IDS environment variable.")
-            session_ids = [sid.strip() for sid in session_ids_str.split(',')]
-            accounts_to_process = [{"session_id": sid} for sid in session_ids]
-        else:
-            print("Loading Tenbin accounts from file: tenbin.json")
-            with open("tenbin.json", "r", encoding="utf-8") as f:
-                accounts_to_process = json.load(f)
-
-        if not isinstance(accounts_to_process, list):
-            print("Warning: Account data should be a list of objects.")
-            return
-
-        for acc in accounts_to_process:
-            session_id = acc.get("session_id")
-            if session_id:
-                TENBIN_ACCOUNTS.append({
-                    "session_id": session_id,
-                    "is_valid": True,
-                    "last_used": 0,
-                    "error_count": 0
-                })
-        print(f"Successfully loaded {len(TENBIN_ACCOUNTS)} Tenbin accounts.")
-    except FileNotFoundError:
-        print("Error: tenbin.json not found and SESSION_IDS not set. API calls will fail.")
-    except Exception as e:
-        print(f"Error loading tenbin.json: {e}")
-
-
-def load_tenbin_models():
-    """Load Tenbin models from models.json"""
-    global TENBIN_MODELS
-    try:
-        with open("models.json", "r", encoding="utf-8") as f:
-            models_data = json.load(f)
-            if isinstance(models_data, dict):
-                TENBIN_MODELS = models_data
-                print(f"Successfully loaded {len(TENBIN_MODELS)} models.")
-            else:
-                print("Warning: models.json should contain a dictionary of model mappings.")
-                TENBIN_MODELS = {}
-    except FileNotFoundError:
-        print("Error: models.json not found. Model list will be empty.")
-        TENBIN_MODELS = {}
-    except Exception as e:
-        print(f"Error loading models.json: {e}")
-        TENBIN_MODELS = {}
-
-
-def get_best_tenbin_account() -> Optional[TenbinAccount]:
-    """Get the best available Tenbin account using a smart selection algorithm."""
-    with account_rotation_lock:
-        now = time.time()
-        valid_accounts = [
-            acc for acc in TENBIN_ACCOUNTS 
-            if acc["is_valid"] and (
-                acc["error_count"] < MAX_ERROR_COUNT or 
-                now - acc["last_used"] > ERROR_COOLDOWN
-            )
-        ]
-        
-        if not valid_accounts:
-            return None
-            
-        # Reset error count for accounts that have been in cooldown
-        for acc in valid_accounts:
-            if acc["error_count"] >= MAX_ERROR_COUNT and now - acc["last_used"] > ERROR_COOLDOWN:
-                acc["error_count"] = 0
-                
-        # Sort by last used (oldest first) and error count (lowest first)
-        valid_accounts.sort(key=lambda x: (x["last_used"], x["error_count"]))
-        account = valid_accounts[0]
-        account["last_used"] = now
-        return account
-
-
-def build_tenbin_prompt(messages: List[ChatMessage]) -> str:
-    """将 OpenAI 格式的消息列表转换为 Tenbin 格式的单个字符串"""
-    prompt = ""
-    for msg in messages:
-        role = msg.role
-        content = msg.content
-        if isinstance(content, list):
-            # 简单处理多模态内容，只提取文本部分
-            content = " ".join([
-                item.get("text", "") 
-                for item in content 
-                if item.get("type") == "text"
-            ])
-        
-        # 添加到提示中
-        if role == "system":
-            # 系统消息作为 Human 消息的前缀
-            prompt += f"\n\nHuman: <system>{content}</system>"
-        elif role == "user":
-            prompt += f"\n\nHuman: {content}"
-        elif role == "assistant":
-            prompt += f"\n\nAssistant: {content}"
-        # 忽略其他角色
-    
-    # 添加最后的 "Assistant:" 提示
-    prompt += "\n\nAssistant:"
-    return prompt
-
-
-async def authenticate_client(
-    auth: Optional[HTTPAuthorizationCredentials] = Depends(security),
-):
-    """Authenticate client based on API key in Authorization header"""
-    if not VALID_CLIENT_KEYS:
-        raise HTTPException(
-            status_code=503,
-            detail="Service unavailable: Client API keys not configured on server.",
-        )
-
-    if not auth or not auth.credentials:
-        raise HTTPException(
-            status_code=401,
-            detail="API key required in Authorization header.",
-            headers={"WWW-Authenticate": "Bearer"},
-        )
-
-    if auth.credentials not in VALID_CLIENT_KEYS:
-        raise HTTPException(status_code=403, detail="Invalid client API key.")
-
-
-@app.on_event("startup")
-async def startup():
-    """应用启动时初始化配置"""
-    print("Starting Tenbin OpenAI API Adapter server...")
-    load_client_api_keys()
-    load_tenbin_accounts()
-    load_tenbin_models()
-    print("Server initialization completed.")
-
-
-def get_models_list_response() -> ModelList:
-    """Helper to construct ModelList response from cached models."""
-    model_infos = [
-        ModelInfo(
-            id=model_id,
-            created=int(time.time()),
-            owned_by="tenbin"
-        )
-        for model_id in TENBIN_MODELS.keys()
-    ]
-    return ModelList(data=model_infos)
-
-
-@app.get("/v1/models", response_model=ModelList)
-async def list_v1_models(_: None = Depends(authenticate_client)):
-    """List available models - authenticated"""
-    return get_models_list_response()
-
-
-@app.get("/models", response_model=ModelList)
-async def list_models_no_auth():
-    """List available models without authentication - for client compatibility"""
-    return get_models_list_response()
-
-
-@app.get("/debug")
-async def toggle_debug(enable: bool = Query(None)):
-    """切换调试模式"""
-    global DEBUG_MODE
-    if enable is not None:
-        DEBUG_MODE = enable
-    return {"debug_mode": DEBUG_MODE}
-
-
-@app.post("/v1/chat/completions")
-async def chat_completions(
-    request: ChatCompletionRequest, _: None = Depends(authenticate_client)
-):
-    """创建聊天完成 - 使用 Tenbin API"""
-    # 检查模型是否存在
-    if request.model not in TENBIN_MODELS:
-        raise HTTPException(status_code=404, detail=f"Model '{request.model}' not found.")
-
-    # 获取内部模型 ID
-    internal_model_id = TENBIN_MODELS[request.model]
-    
-    if not request.messages:
-        raise HTTPException(status_code=400, detail="No messages provided in the request.")
-    
-    log_debug(f"Processing request for model: {request.model} (internal ID: {internal_model_id})")
-    
-    # 构建 Tenbin 格式的提示
-    prompt = build_tenbin_prompt(request.messages)
-    log_debug(f"Built prompt with length: {len(prompt)}")
-    
-    # 尝试所有账户
-    for attempt in range(len(TENBIN_ACCOUNTS)):
-        account = get_best_tenbin_account()
-        if not account:
-            raise HTTPException(
-                status_code=503, 
-                detail="No valid Tenbin accounts available."
-            )
-
-        session_id = account["session_id"]
-        log_debug(f"Using account with session_id ending in ...{session_id[-4:]}")
-        
-        try:
-            # 获取执行令牌
-            execution_token = get_tenbin_execution_token(internal_model_id, session_id)
-            
-            if request.stream:
-                log_debug("Returning stream response")
-                return StreamingResponse(
-                    tenbin_stream_generator(request.model, prompt, session_id, execution_token),
-                    media_type="text/event-stream",
-                    headers={
-                        "Cache-Control": "no-cache",
-                        "Connection": "keep-alive",
-                        "X-Accel-Buffering": "no",
-                    },
-                )
-            else:
-                log_debug("Building non-stream response")
-                return build_tenbin_non_stream_response(request.model, prompt, session_id, execution_token)
-
-        except Exception as e:
-            error_detail = str(e)
-            log_debug(f"Tenbin API error: {error_detail}")
-
-            with account_rotation_lock:
-                # 增加错误计数
-                account["error_count"] += 1
-                log_debug(f"Account ...{session_id[-4:]} error count: {account['error_count']}")
-                
-                # 如果错误看起来是认证问题，标记账户为无效
-                if "authentication" in error_detail.lower() or "unauthorized" in error_detail.lower():
-                    account["is_valid"] = False
-                    log_debug(f"Account ...{session_id[-4:]} marked as invalid due to auth error.")
-
-    # 所有尝试都失败
-    if request.stream:
-        return StreamingResponse(
-            error_stream_generator("All attempts to contact Tenbin API failed.", 503),
-            media_type="text/event-stream",
-            status_code=503,
-        )
-    else:
-        raise HTTPException(status_code=503, detail="All attempts to contact Tenbin API failed.")
-
-
-def get_tenbin_execution_token(model: str, session_id: str) -> str:
-    """获取 Tenbin 执行令牌"""
-    try:
-        task_id = getTaskId()
-        captcha = getCaptcha(task_id)
-        url = "https://graphql.tenbin.ai/graphql"
-
-        payload = {
-            "operationName": "IssueExecutionTokensMultiple",
-            "variables": {
-                "turnstileToken": captcha,
-                "models": [model],
-            },
-            "query": "query IssueExecutionTokensMultiple($turnstileToken: String!, $models: [ChatModel!]!) {\n  executionTokens: issueExecutionTokensMultiple(\n    turnstileToken: $turnstileToken\n    models: $models\n  )\n}",
-        }
-
-        headers = {
-            "User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/137.0.0.0 Safari/537.36",
-            "Accept-Encoding": "gzip, deflate, br, zstd",
-            "Content-Type": "application/json",
-            "Cookie": f"sessionId={session_id}",
-        }
-
-        log_debug(f"Getting execution token for model: {model}")
-        response = requests.post(url, data=json.dumps(payload), headers=headers, timeout=REQUEST_TIMEOUT)
-        response.raise_for_status()
-        
-        execution_token = response.json()["data"]["executionTokens"][0]
-        log_debug(f"Got execution token: {execution_token[:10]}...")
-        return execution_token
-    except Exception as e:
-        log_debug(f"Error getting execution token: {e}")
-        raise
-
-
-def tenbin_stream_generator(model: str, prompt: str, session_id: str, execution_token: str):
-    """Tenbin WebSocket 流式响应生成器"""
-    stream_id = f"chatcmpl-{uuid.uuid4().hex}"
-    created_time = int(time.time())
-    
-    # 发送初始角色增量
-    yield f"data: {StreamResponse(id=stream_id, created=created_time, model=model, choices=[StreamChoice(delta={'role': 'assistant'})]).json()}\n\n"
-    
-    # 连接 WebSocket
-    url = "wss://graphql.tenbin.ai/graphql"
-    headers = {
-        "Host": "graphql.tenbin.ai",
-        "Connection": "Upgrade",
-        "Pragma": "no-cache",
-        "Cache-Control": "no-cache",
-        "User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/137.0.0.0 Safari/537.36 Edg/137.0.0.0",
-        "Upgrade": "websocket",
-        "Origin": "https://tenbin.ai",
-        "Sec-WebSocket-Version": "13",
-        "Accept-Encoding": "gzip, deflate, br, zstd",
-        "Accept-Language": "zh-CN,zh;q=0.9",
-        "Cookie": f"sessionId={session_id}",
-        "Sec-WebSocket-Key": "I/tTy5psJkboWYQfCypjVA==",
-        "Sec-WebSocket-Extensions": "permessage-deflate; client_max_window_bits",
-        "Sec-WebSocket-Protocol": "graphql-transport-ws",
-    }
-    
-    ws = None
-    try:
-        log_debug("Connecting to WebSocket...")
-        ws = websocket.create_connection(url, header=headers)
-        ws.send(json.dumps({"type": "connection_init"}))
-        init_response = ws.recv()
-        log_debug(f"WebSocket init response: {init_response}")
-        
-        # 发送订阅请求
-        payload = {
-            "id": str(uuid.uuid4()),
-            "type": "subscribe",
-            "payload": {
-                "variables": {
-                    "prompt": prompt,
-                    "executionToken": execution_token,
-                    "stateToken": "",
-                },
-                "extensions": {},
-                "operationName": "StartConversation",
-                "query": "subscription StartConversation($executionToken: String!, $itemId: String, $itemDraftId: String, $systemPrompt: String, $prompt: String, $stateToken: String, $variables: [ConversationVariableInput!], $itemCallOption: ItemCallOption, $fileKey: String, $fileUploadIds: [String!], $selectedToolsByUser: [ToolType!]) {\n  startConversation(\n    executionToken: $executionToken\n    itemId: $itemId\n    itemDraftId: $itemDraftId\n    systemPrompt: $systemPrompt\n    prompt: $prompt\n    stateToken: $stateToken\n    variables: $variables\n    itemCallOption: $itemCallOption\n    fileKey: $fileKey\n    fileUploadIds: $fileUploadIds\n    selectedToolsByUser: $selectedToolsByUser\n  ) {\n    ...DeltaConversation\n    __typename\n  }\n}\n\nfragment DeltaConversation on AIConversationStreamResult {\n  seq\n  deltaToken\n  isFinished\n  newStateToken\n  error\n  fileUploadIds\n  toolResult {\n    id\n    title\n    url\n    faviconUrl\n    summary\n    __typename\n  }\n  action\n  activity\n  toolError\n  __typename\n}",
-            },
-        }
-        
-        log_debug("Sending subscription request...")
-        ws.send(json.dumps(payload))
-        
-        # 处理响应
-        accumulated_thinking = ""
-        thinking_mode = False
-        thinking_separator = "\n\n---\n\n"
-        is_thinking_model = model == "Claude-3.7-Sonnet-Extended"
-        
-        while True:
-            try:
-                msg = ws.recv()
-                log_debug(f"Received message: {msg[:100]}..." if len(msg) > 100 else msg)
-                
-                if msg.endswith('"type":"complete"}'):
-                    log_debug("Received complete message")
-                    break
-                
-                try:
-                    data = json.loads(msg)
-                    if data.get("type") != "next":
-                        continue
-                    
-                    payload_data = data.get("payload", {}).get("data", {})
-                    conversation = payload_data.get("startConversation", {})
-                    
-                    delta_token = conversation.get("deltaToken", "")
-                    is_finished = conversation.get("isFinished", False)
-                    
-                    if delta_token:
-                        if is_thinking_model:
-                            # 检查是否包含思考/回答分隔符
-                            if thinking_separator in accumulated_thinking + delta_token:
-                                # 找到分隔符，切换到回答模式
-                                if not thinking_mode:
-                                    # 如果之前没有发送过思考内容，先发送累积的思考内容
-                                    parts = (accumulated_thinking + delta_token).split(thinking_separator, 1)
-                                    thinking_content = parts[0]
-                                    answer_content = parts[1] if len(parts) > 1 else ""
-                                    
-                                    if thinking_content:
-                                        yield f"data: {StreamResponse(id=stream_id, created=created_time, model=model, choices=[StreamChoice(delta={'reasoning_content': thinking_content})]).json()}\n\n"
-                                    
-                                    if answer_content:
-                                        yield f"data: {StreamResponse(id=stream_id, created=created_time, model=model, choices=[StreamChoice(delta={'content': answer_content})]).json()}\n\n"
-                                    
-                                    thinking_mode = True
-                                    accumulated_thinking = ""
-                                else:
-                                    # 已经在回答模式，直接发送内容
-                                    yield f"data: {StreamResponse(id=stream_id, created=created_time, model=model, choices=[StreamChoice(delta={'content': delta_token})]).json()}\n\n"
-                            else:
-                                # 没有找到分隔符
-                                if thinking_mode:
-                                    # 已经在回答模式，直接发送内容
-                                    yield f"data: {StreamResponse(id=stream_id, created=created_time, model=model, choices=[StreamChoice(delta={'content': delta_token})]).json()}\n\n"
-                                else:
-                                    # 继续累积思考内容
-                                    accumulated_thinking += delta_token
-                        else:
-                            # 非思考模型，直接发送内容
-                            yield f"data: {StreamResponse(id=stream_id, created=created_time, model=model, choices=[StreamChoice(delta={'content': delta_token})]).json()}\n\n"
-                    
-                    if is_finished:
-                        # 如果还有未发送的思考内容，发送它
-                        if is_thinking_model and not thinking_mode and accumulated_thinking:
-                            yield f"data: {StreamResponse(id=stream_id, created=created_time, model=model, choices=[StreamChoice(delta={'reasoning_content': accumulated_thinking})]).json()}\n\n"
-                        
-                        # 发送完成信号
-                        log_debug("Stream finished")
-                        yield f"data: {StreamResponse(id=stream_id, created=created_time, model=model, choices=[StreamChoice(delta={}, finish_reason='stop')]).json()}\n\n"
-                        yield "data: [DONE]\n\n"
-                        break
-                        
-                except json.JSONDecodeError as e:
-                    log_debug(f"JSON decode error: {e}")
-                    continue
-                    
-            except websocket.WebSocketConnectionClosedException:
-                log_debug("WebSocket connection closed")
-                break
-                
-            except Exception as e:
-                log_debug(f"Error processing message: {e}")
-                yield f"data: {json.dumps({'error': str(e)})}\n\n"
-                break
-    
-    except Exception as e:
-        log_debug(f"WebSocket error: {e}")
-        yield f"data: {json.dumps({'error': str(e)})}\n\n"
-        yield "data: [DONE]\n\n"
-    
-    finally:
-        if ws:
-            try:
-                ws.close()
-                log_debug("WebSocket connection closed")
-            except:
-                pass
-
-
-def build_tenbin_non_stream_response(model: str, prompt: str, session_id: str, execution_token: str) -> ChatCompletionResponse:
-    """构建非流式响应"""
-    full_content = ""
-    full_reasoning_content = None
-    
-    # 使用流式生成器，但累积所有内容
-    for chunk in tenbin_stream_generator(model, prompt, session_id, execution_token):
-        if not chunk.startswith("data: ") or chunk.strip() == "data: [DONE]":
-            continue
-            
-        try:
-            data = json.loads(chunk[6:])  # 去掉 "data: " 前缀
-            if "choices" not in data:
-                continue
-                
-            delta = data["choices"][0].get("delta", {})
-            
-            if "content" in delta and delta["content"]:
-                full_content += delta["content"]
-                
-            if "reasoning_content" in delta and delta["reasoning_content"]:
-                if full_reasoning_content is None:
-                    full_reasoning_content = ""
-                full_reasoning_content += delta["reasoning_content"]
-                
-        except json.JSONDecodeError:
-            continue
-    
-    return ChatCompletionResponse(
-        model=model,
-        choices=[
-            ChatCompletionChoice(
-                message=ChatMessage(
-                    role="assistant",
-                    content=full_content,
-                    reasoning_content=full_reasoning_content,
-                )
-            )
-        ],
-    )
-
-
-async def error_stream_generator(error_detail: str, status_code: int):
-    """Generate error stream response"""
-    yield f'data: {json.dumps({"error": {"message": error_detail, "type": "tenbin_api_error", "code": status_code}})}\n\n'
-    yield "data: [DONE]\n\n"
-
-
-if __name__ == "__main__":
-    import uvicorn
-
-    # 从环境变量获取端口，默认为 8000
-    port = int(os.environ.get("PORT", 8000))
-
-    # 设置环境变量以启用调试模式
-    if os.environ.get("DEBUG_MODE", "").lower() == "true":
-        DEBUG_MODE = True
-        print("Debug mode enabled via environment variable")
-
-    if not os.path.exists("tenbin.json") and not os.environ.get("SESSION_IDS"):
-        print("Warning: tenbin.json not found and SESSION_IDS not set. Creating a dummy file.")
-        dummy_data = [
-            {
-                "session_id": "your_session_id_here",
-            }
-        ]
-        with open("tenbin.json", "w", encoding="utf-8") as f:
-            json.dump(dummy_data, f, indent=4)
-        print("Created dummy tenbin.json. Please replace with valid Tenbin data or set SESSION_IDS secret.")
-
-    if not os.path.exists("client_api_keys.json") and not os.environ.get("API_KEYS"):
-        print("Warning: client_api_keys.json not found and API_KEYS not set. Creating a dummy file.")
-        dummy_key = f"sk-dummy-{uuid.uuid4().hex}"
-        with open("client_api_keys.json", "w", encoding="utf-8") as f:
-            json.dump([dummy_key], f, indent=2)
-        print(f"Created dummy client_api_keys.json with key: {dummy_key}. Or set API_KEYS secret.")
-
-    if not os.path.exists("models.json"):
-        print("Warning: models.json not found. Creating a dummy file.")
-        dummy_models = {
-            "claude-3.7-sonnet": "AnthropicClaude37Sonnet",
-            "claude-3.7-sonnet-extended": "AnthropicClaude37SonnetExtended"
-        }
-        with open("models.json", "w", encoding="utf-8") as f:
-            json.dump(dummy_models, f, indent=4)
-        print("Created dummy models.json.")
-
-    load_client_api_keys()
-    load_tenbin_accounts()
-    load_tenbin_models()
-
-    print("\n--- Tenbin OpenAI API Adapter ---")
-    print(f"Debug Mode: {DEBUG_MODE}")
-    print("Endpoints:")
-    print("  GET  /v1/models (Client API Key Auth)")
-    print("  GET  /models (No Auth)")
-    print("  POST /v1/chat/completions (Client API Key Auth)")
-    print("  GET  /debug?enable=[true|false] (Toggle Debug Mode)")
-
-    print(f"\nClient API Keys: {len(VALID_CLIENT_KEYS)}")
-    if TENBIN_ACCOUNTS:
-        print(f"Tenbin Accounts: {len(TENBIN_ACCOUNTS)}")
-    else:
-        print("Tenbin Accounts: None loaded. Check tenbin.json.")
-    if TENBIN_MODELS:
-        models = sorted(list(TENBIN_MODELS.keys()))
-        print(f"Tenbin Models: {len(TENBIN_MODELS)}")
-        print(f"Available models: {', '.join(models[:5])}{'...' if len(models) > 5 else ''}")
-    else:
-        print("Tenbin Models: None loaded. Check models.json.")
-    print("------------------------------------")
-
+import json
+import os
+import time
+import uuid
+import threading
+from typing import Any, Dict, List, Optional, TypedDict, Union
+
+import requests
+import websocket
+from fastapi import FastAPI, HTTPException, Depends, Query
+from fastapi.responses import StreamingResponse
+from fastapi.security import HTTPBearer, HTTPAuthorizationCredentials
+from pydantic import BaseModel, Field
+
+from getCaptcha import getCaptcha, getTaskId
+
+
+# Tenbin Account Management
+class TenbinAccount(TypedDict):
+    session_id: str
+    is_valid: bool
+    last_used: float
+    error_count: int
+
+
+# Global variables
+VALID_CLIENT_KEYS: set = set()
+TENBIN_ACCOUNTS: List[TenbinAccount] = []
+TENBIN_MODELS: Dict[str, str] = {}  # 模型映射表，key 是模型名称，value 是内部模型 ID
+account_rotation_lock = threading.Lock()
+MAX_ERROR_COUNT = 3
+ERROR_COOLDOWN = 300  # 5 minutes cooldown for accounts with errors
+DEBUG_MODE = os.environ.get("DEBUG_MODE", "false").lower() == "true"
+REQUEST_TIMEOUT = 120.0  # 请求超时时间，秒
+HTTP_PROXY = os.environ.get("HTTP_PROXY")
+PROXIES = {"http": HTTP_PROXY, "https": HTTP_PROXY} if HTTP_PROXY else None
+
+if PROXIES:
+    print(f"Using HTTP proxy: {HTTP_PROXY}")
+
+
+# Pydantic Models
+class ChatMessage(BaseModel):
+    role: str
+    content: Union[str, List[Dict[str, Any]]]
+    reasoning_content: Optional[str] = None
+
+
+class ChatCompletionRequest(BaseModel):
+    model: str
+    messages: List[ChatMessage]
+    stream: bool = True
+    temperature: Optional[float] = None
+    max_tokens: Optional[int] = None
+    top_p: Optional[float] = None
+    raw_response: bool = False  # ��否返回原始响应
+
+
+class ModelInfo(BaseModel):
+    id: str
+    object: str = "model"
+    created: int
+    owned_by: str = "tenbin"
+
+
+class ModelList(BaseModel):
+    object: str = "list"
+    data: List[ModelInfo]
+
+
+class ChatCompletionChoice(BaseModel):
+    message: ChatMessage
+    index: int = 0
+    finish_reason: str = "stop"
+
+
+class ChatCompletionResponse(BaseModel):
+    id: str = Field(default_factory=lambda: f"chatcmpl-{uuid.uuid4().hex}")
+    object: str = "chat.completion"
+    created: int = Field(default_factory=lambda: int(time.time()))
+    model: str
+    choices: List[ChatCompletionChoice]
+    usage: Dict[str, int] = Field(
+        default_factory=lambda: {
+            "prompt_tokens": 0,
+            "completion_tokens": 0,
+            "total_tokens": 0,
+        }
+    )
+
+
+class StreamChoice(BaseModel):
+    delta: Dict[str, Any] = Field(default_factory=dict)
+    index: int = 0
+    finish_reason: Optional[str] = None
+
+
+class StreamResponse(BaseModel):
+    id: str = Field(default_factory=lambda: f"chatcmpl-{uuid.uuid4().hex}")
+    object: str = "chat.completion.chunk"
+    created: int = Field(default_factory=lambda: int(time.time()))
+    model: str
+    choices: List[StreamChoice]
+
+
+# FastAPI App
+app = FastAPI(title="Tenbin OpenAI API Adapter")
+security = HTTPBearer(auto_error=False)
+
+
+def log_debug(message: str):
+    """Debug日志函数"""
+    if DEBUG_MODE:
+        print(f"[DEBUG] {message}")
+
+
+def load_client_api_keys():
+    """Load client API keys from environment variable (comma-separated) or client_api_keys.json"""
+    global VALID_CLIENT_KEYS
+    keys_str = os.environ.get("API_KEYS")
+    try:
+        if keys_str:
+            print("Loading client API keys from API_KEYS environment variable.")
+            keys = [key.strip() for key in keys_str.split(',')]
+        else:
+            print("Loading client API keys from file: client_api_keys.json")
+            with open("client_api_keys.json", "r", encoding="utf-8") as f:
+                keys = json.load(f)
+
+        VALID_CLIENT_KEYS = set(keys) if isinstance(keys, list) else set()
+        print(f"Successfully loaded {len(VALID_CLIENT_KEYS)} client API keys.")
+    except FileNotFoundError:
+        print("Error: client_api_keys.json not found and API_KEYS not set. Client authentication will fail.")
+        VALID_CLIENT_KEYS = set()
+    except Exception as e:
+        print(f"Error loading client API keys: {e}")
+        VALID_CLIENT_KEYS = set()
+
+
+def load_tenbin_accounts():
+    """Load Tenbin accounts from environment variable (comma-separated session_ids) or tenbin.json"""
+    global TENBIN_ACCOUNTS
+    TENBIN_ACCOUNTS = []
+    session_ids_str = os.environ.get("SESSION_IDS")
+    try:
+        accounts_to_process = []
+        if session_ids_str:
+            print("Loading Tenbin accounts from SESSION_IDS environment variable.")
+            session_ids = [sid.strip() for sid in session_ids_str.split(',')]
+            accounts_to_process = [{"session_id": sid} for sid in session_ids]
+        else:
+            print("Loading Tenbin accounts from file: tenbin.json")
+            with open("tenbin.json", "r", encoding="utf-8") as f:
+                accounts_to_process = json.load(f)
+
+        if not isinstance(accounts_to_process, list):
+            print("Warning: Account data should be a list of objects.")
+            return
+
+        for acc in accounts_to_process:
+            session_id = acc.get("session_id")
+            if session_id:
+                TENBIN_ACCOUNTS.append({
+                    "session_id": session_id,
+                    "is_valid": True,
+                    "last_used": 0,
+                    "error_count": 0
+                })
+        print(f"Successfully loaded {len(TENBIN_ACCOUNTS)} Tenbin accounts.")
+    except FileNotFoundError:
+        print("Error: tenbin.json not found and SESSION_IDS not set. API calls will fail.")
+    except Exception as e:
+        print(f"Error loading tenbin.json: {e}")
+
+
+def load_tenbin_models():
+    """Load Tenbin models from models.json"""
+    global TENBIN_MODELS
+    try:
+        with open("models.json", "r", encoding="utf-8") as f:
+            models_data = json.load(f)
+            if isinstance(models_data, dict):
+                TENBIN_MODELS = models_data
+                print(f"Successfully loaded {len(TENBIN_MODELS)} models.")
+            else:
+                print("Warning: models.json should contain a dictionary of model mappings.")
+                TENBIN_MODELS = {}
+    except FileNotFoundError:
+        print("Error: models.json not found. Model list will be empty.")
+        TENBIN_MODELS = {}
+    except Exception as e:
+        print(f"Error loading models.json: {e}")
+        TENBIN_MODELS = {}
+
+
+def get_best_tenbin_account() -> Optional[TenbinAccount]:
+    """Get the best available Tenbin account using a smart selection algorithm."""
+    with account_rotation_lock:
+        now = time.time()
+        valid_accounts = [
+            acc for acc in TENBIN_ACCOUNTS 
+            if acc["is_valid"] and (
+                acc["error_count"] < MAX_ERROR_COUNT or 
+                now - acc["last_used"] > ERROR_COOLDOWN
+            )
+        ]
+        
+        if not valid_accounts:
+            return None
+            
+        # Reset error count for accounts that have been in cooldown
+        for acc in valid_accounts:
+            if acc["error_count"] >= MAX_ERROR_COUNT and now - acc["last_used"] > ERROR_COOLDOWN:
+                acc["error_count"] = 0
+                
+        # Sort by last used (oldest first) and error count (lowest first)
+        valid_accounts.sort(key=lambda x: (x["last_used"], x["error_count"]))
+        account = valid_accounts[0]
+        account["last_used"] = now
+        return account
+
+
+def build_tenbin_prompt(messages: List[ChatMessage]) -> str:
+    """将 OpenAI 格式的消息列表转换为 Tenbin 格式的单个字符串"""
+    prompt = ""
+    for msg in messages:
+        role = msg.role
+        content = msg.content
+        if isinstance(content, list):
+            # 简单处理多模态内容，只提取文本部分
+            content = " ".join([
+                item.get("text", "") 
+                for item in content 
+                if item.get("type") == "text"
+            ])
+        
+        # 添加到提示中
+        if role == "system":
+            # 系统消息作为 Human 消息的前缀
+            prompt += f"\n\nHuman: <system>{content}</system>"
+        elif role == "user":
+            prompt += f"\n\nHuman: {content}"
+        elif role == "assistant":
+            prompt += f"\n\nAssistant: {content}"
+        # 忽略其他角色
+    
+    # 添加最后的 "Assistant:" 提示
+    prompt += "\n\nAssistant:"
+    return prompt
+
+
+async def authenticate_client(
+    auth: Optional[HTTPAuthorizationCredentials] = Depends(security),
+):
+    """Authenticate client based on API key in Authorization header"""
+    if not VALID_CLIENT_KEYS:
+        raise HTTPException(
+            status_code=503,
+            detail="Service unavailable: Client API keys not configured on server.",
+        )
+
+    if not auth or not auth.credentials:
+        raise HTTPException(
+            status_code=401,
+            detail="API key required in Authorization header.",
+            headers={"WWW-Authenticate": "Bearer"},
+        )
+
+    if auth.credentials not in VALID_CLIENT_KEYS:
+        raise HTTPException(status_code=403, detail="Invalid client API key.")
+
+
+@app.on_event("startup")
+async def startup():
+    """应用启动时初始化配置"""
+    print("Starting Tenbin OpenAI API Adapter server...")
+    load_client_api_keys()
+    load_tenbin_accounts()
+    load_tenbin_models()
+    print("Server initialization completed.")
+
+
+def get_models_list_response() -> ModelList:
+    """Helper to construct ModelList response from cached models."""
+    model_infos = [
+        ModelInfo(
+            id=model_id,
+            created=int(time.time()),
+            owned_by="tenbin"
+        )
+        for model_id in TENBIN_MODELS.keys()
+    ]
+    return ModelList(data=model_infos)
+
+
+@app.get("/v1/models", response_model=ModelList)
+async def list_v1_models(_: None = Depends(authenticate_client)):
+    """List available models - authenticated"""
+    return get_models_list_response()
+
+
+@app.get("/models", response_model=ModelList)
+async def list_models_no_auth():
+    """List available models without authentication - for client compatibility"""
+    return get_models_list_response()
+
+
+@app.get("/debug")
+async def toggle_debug(enable: bool = Query(None)):
+    """切换调试模式"""
+    global DEBUG_MODE
+    if enable is not None:
+        DEBUG_MODE = enable
+    return {"debug_mode": DEBUG_MODE}
+
+
+@app.post("/v1/chat/completions")
+async def chat_completions(
+    request: ChatCompletionRequest, _: None = Depends(authenticate_client)
+):
+    """创建聊天完成 - 使用 Tenbin API"""
+    # 检查模型是否存在
+    if request.model not in TENBIN_MODELS:
+        raise HTTPException(status_code=404, detail=f"Model '{request.model}' not found.")
+
+    # 获取内部模型 ID
+    internal_model_id = TENBIN_MODELS[request.model]
+    
+    if not request.messages:
+        raise HTTPException(status_code=400, detail="No messages provided in the request.")
+    
+    log_debug(f"Processing request for model: {request.model} (internal ID: {internal_model_id})")
+    
+    # 构建 Tenbin 格式的提示
+    prompt = build_tenbin_prompt(request.messages)
+    log_debug(f"Built prompt with length: {len(prompt)}")
+    
+    # 尝试所有账户
+    for attempt in range(len(TENBIN_ACCOUNTS)):
+        account = get_best_tenbin_account()
+        if not account:
+            raise HTTPException(
+                status_code=503, 
+                detail="No valid Tenbin accounts available."
+            )
+
+        session_id = account["session_id"]
+        log_debug(f"Using account with session_id ending in ...{session_id[-4:]}")
+        
+        try:
+            # 获��执行令牌
+            execution_token = get_tenbin_execution_token(internal_model_id, session_id)
+            
+            if request.stream:
+                log_debug("Returning stream response")
+                return StreamingResponse(
+                    tenbin_stream_generator(request.model, prompt, session_id, execution_token),
+                    media_type="text/event-stream",
+                    headers={
+                        "Cache-Control": "no-cache",
+                        "Connection": "keep-alive",
+                        "X-Accel-Buffering": "no",
+                    },
+                )
+            else:
+                log_debug("Building non-stream response")
+                return build_tenbin_non_stream_response(request.model, prompt, session_id, execution_token)
+
+        except Exception as e:
+            error_detail = str(e)
+            log_debug(f"Tenbin API error: {error_detail}")
+
+            with account_rotation_lock:
+                # 增加错误计数
+                account["error_count"] += 1
+                log_debug(f"Account ...{session_id[-4:]} error count: {account['error_count']}")
+                
+                # 如果错误看起来是认证问题，标记账户为无效
+                if "authentication" in error_detail.lower() or "unauthorized" in error_detail.lower():
+                    account["is_valid"] = False
+                    log_debug(f"Account ...{session_id[-4:]} marked as invalid due to auth error.")
+
+    # 所有尝试都失败
+    if request.stream:
+        return StreamingResponse(
+            error_stream_generator("All attempts to contact Tenbin API failed.", 503),
+            media_type="text/event-stream",
+            status_code=503,
+        )
+    else:
+        raise HTTPException(status_code=503, detail="All attempts to contact Tenbin API failed.")
+
+
+def get_tenbin_execution_token(model: str, session_id: str) -> str:
+    """获取 Tenbin 执行令牌"""
+    try:
+        task_id = getTaskId()
+        captcha = getCaptcha(task_id)
+        url = "https://graphql.tenbin.ai/graphql"
+
+        payload = {
+            "operationName": "IssueExecutionTokensMultiple",
+            "variables": {
+                "turnstileToken": captcha,
+                "models": [model],
+            },
+            "query": "query IssueExecutionTokensMultiple($turnstileToken: String!, $models: [ChatModel!]!) {\n  executionTokens: issueExecutionTokensMultiple(\n    turnstileToken: $turnstileToken\n    models: $models\n  )\n}",
+        }
+
+        headers = {
+            "User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/137.0.0.0 Safari/537.36",
+            "Accept-Encoding": "gzip, deflate, br, zstd",
+            "Content-Type": "application/json",
+            "Cookie": f"sessionId={session_id}",
+        }
+
+        log_debug(f"Getting execution token for model: {model}")
+        response = requests.post(
+            url, 
+            data=json.dumps(payload), 
+            headers=headers, 
+            timeout=REQUEST_TIMEOUT,
+            proxies=PROXIES
+        )
+        response.raise_for_status()
+        
+        execution_token = response.json()["data"]["executionTokens"][0]
+        log_debug(f"Got execution token: {execution_token[:10]}...")
+        return execution_token
+    except Exception as e:
+        log_debug(f"Error getting execution token: {e}")
+        raise
+
+
+def tenbin_stream_generator(model: str, prompt: str, session_id: str, execution_token: str):
+    """Tenbin WebSocket 流式响应生成器"""
+    stream_id = f"chatcmpl-{uuid.uuid4().hex}"
+    created_time = int(time.time())
+    
+    # 发送初始角色增量
+    yield f"data: {StreamResponse(id=stream_id, created=created_time, model=model, choices=[StreamChoice(delta={'role': 'assistant'})]).json()}\n\n"
+    
+    # 连接 WebSocket
+    url = "wss://graphql.tenbin.ai/graphql"
+    headers = {
+        "Host": "graphql.tenbin.ai",
+        "Connection": "Upgrade",
+        "Pragma": "no-cache",
+        "Cache-Control": "no-cache",
+        "User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/137.0.0.0 Safari/537.36 Edg/137.0.0.0",
+        "Upgrade": "websocket",
+        "Origin": "https://tenbin.ai",
+        "Sec-WebSocket-Version": "13",
+        "Accept-Encoding": "gzip, deflate, br, zstd",
+        "Accept-Language": "zh-CN,zh;q=0.9",
+        "Cookie": f"sessionId={session_id}",
+        "Sec-WebSocket-Key": "I/tTy5psJkboWYQfCypjVA==",
+        "Sec-WebSocket-Extensions": "permessage-deflate; client_max_window_bits",
+        "Sec-WebSocket-Protocol": "graphql-transport-ws",
+    }
+    
+    ws = None
+    try:
+        log_debug("Connecting to WebSocket...")
+        ws = websocket.create_connection(url, header=headers)
+        ws.send(json.dumps({"type": "connection_init"}))
+        init_response = ws.recv()
+        log_debug(f"WebSocket init response: {init_response}")
+        
+        # 发送订���请求
+        payload = {
+            "id": str(uuid.uuid4()),
+            "type": "subscribe",
+            "payload": {
+                "variables": {
+                    "prompt": prompt,
+                    "executionToken": execution_token,
+                    "stateToken": "",
+                },
+                "extensions": {},
+                "operationName": "StartConversation",
+                "query": "subscription StartConversation($executionToken: String!, $itemId: String, $itemDraftId: String, $systemPrompt: String, $prompt: String, $stateToken: String, $variables: [ConversationVariableInput!], $itemCallOption: ItemCallOption, $fileKey: String, $fileUploadIds: [String!], $selectedToolsByUser: [ToolType!]) {\n  startConversation(\n    executionToken: $executionToken\n    itemId: $itemId\n    itemDraftId: $itemDraftId\n    systemPrompt: $systemPrompt\n    prompt: $prompt\n    stateToken: $stateToken\n    variables: $variables\n    itemCallOption: $itemCallOption\n    fileKey: $fileKey\n    fileUploadIds: $fileUploadIds\n    selectedToolsByUser: $selectedToolsByUser\n  ) {\n    ...DeltaConversation\n    __typename\n  }\n}\n\nfragment DeltaConversation on AIConversationStreamResult {\n  seq\n  deltaToken\n  isFinished\n  newStateToken\n  error\n  fileUploadIds\n  toolResult {\n    id\n    title\n    url\n    faviconUrl\n    summary\n    __typename\n  }\n  action\n  activity\n  toolError\n  __typename\n}",
+            },
+        }
+        
+        log_debug("Sending subscription request...")
+        ws.send(json.dumps(payload))
+        
+        # 处理响应
+        accumulated_thinking = ""
+        thinking_mode = False
+        thinking_separator = "\n\n---\n\n"
+        is_thinking_model = model == "Claude-3.7-Sonnet-Extended"
+        
+        while True:
+            try:
+                msg = ws.recv()
+                log_debug(f"Received message: {msg[:100]}..." if len(msg) > 100 else msg)
+                
+                if msg.endswith('"type":"complete"}'):
+                    log_debug("Received complete message")
+                    break
+                
+                try:
+                    data = json.loads(msg)
+                    if data.get("type") != "next":
+                        continue
+                    
+                    payload_data = data.get("payload", {}).get("data", {})
+                    conversation = payload_data.get("startConversation", {})
+                    
+                    delta_token = conversation.get("deltaToken", "")
+                    is_finished = conversation.get("isFinished", False)
+                    
+                    if delta_token:
+                        if is_thinking_model:
+                            # 检查是否包含思考/回答分隔符
+                            if thinking_separator in accumulated_thinking + delta_token:
+                                # 找到分隔符，切换到回答模式
+                                if not thinking_mode:
+                                    # 如果之前没有发送过思考内容，先发送累积的思考内容
+                                    parts = (accumulated_thinking + delta_token).split(thinking_separator, 1)
+                                    thinking_content = parts[0]
+                                    answer_content = parts[1] if len(parts) > 1 else ""
+                                    
+                                    if thinking_content:
+                                        yield f"data: {StreamResponse(id=stream_id, created=created_time, model=model, choices=[StreamChoice(delta={'reasoning_content': thinking_content})]).json()}\n\n"
+                                    
+                                    if answer_content:
+                                        yield f"data: {StreamResponse(id=stream_id, created=created_time, model=model, choices=[StreamChoice(delta={'content': answer_content})]).json()}\n\n"
+                                    
+                                    thinking_mode = True
+                                    accumulated_thinking = ""
+                                else:
+                                    # 已经在回答模式，直接发送内容
+                                    yield f"data: {StreamResponse(id=stream_id, created=created_time, model=model, choices=[StreamChoice(delta={'content': delta_token})]).json()}\n\n"
+                            else:
+                                # 没有找到分隔符
+                                if thinking_mode:
+                                    # 已经在回答模式，直接发送内容
+                                    yield f"data: {StreamResponse(id=stream_id, created=created_time, model=model, choices=[StreamChoice(delta={'content': delta_token})]).json()}\n\n"
+                                else:
+                                    # 继续累积思考内容
+                                    accumulated_thinking += delta_token
+                        else:
+                            # 非思��模型，直接发送内容
+                            yield f"data: {StreamResponse(id=stream_id, created=created_time, model=model, choices=[StreamChoice(delta={'content': delta_token})]).json()}\n\n"
+                    
+                    if is_finished:
+                        # 如果还有未发送的思考内容，发送它
+                        if is_thinking_model and not thinking_mode and accumulated_thinking:
+                            yield f"data: {StreamResponse(id=stream_id, created=created_time, model=model, choices=[StreamChoice(delta={'reasoning_content': accumulated_thinking})]).json()}\n\n"
+                        
+                        # 发送完成信号
+                        log_debug("Stream finished")
+                        yield f"data: {StreamResponse(id=stream_id, created=created_time, model=model, choices=[StreamChoice(delta={}, finish_reason='stop')]).json()}\n\n"
+                        yield "data: [DONE]\n\n"
+                        break
+                        
+                except json.JSONDecodeError as e:
+                    log_debug(f"JSON decode error: {e}")
+                    continue
+                    
+            except websocket.WebSocketConnectionClosedException:
+                log_debug("WebSocket connection closed")
+                break
+                
+            except Exception as e:
+                log_debug(f"Error processing message: {e}")
+                yield f"data: {json.dumps({'error': str(e)})}\n\n"
+                break
+    
+    except Exception as e:
+        log_debug(f"WebSocket error: {e}")
+        yield f"data: {json.dumps({'error': str(e)})}\n\n"
+        yield "data: [DONE]\n\n"
+    
+    finally:
+        if ws:
+            try:
+                ws.close()
+                log_debug("WebSocket connection closed")
+            except:
+                pass
+
+
+def build_tenbin_non_stream_response(model: str, prompt: str, session_id: str, execution_token: str) -> ChatCompletionResponse:
+    """构建非流式响应"""
+    full_content = ""
+    full_reasoning_content = None
+    
+    # 使用流式生成器，但累积所有内容
+    for chunk in tenbin_stream_generator(model, prompt, session_id, execution_token):
+        if not chunk.startswith("data: ") or chunk.strip() == "data: [DONE]":
+            continue
+            
+        try:
+            data = json.loads(chunk[6:])  # 去掉 "data: " 前缀
+            if "choices" not in data:
+                continue
+                
+            delta = data["choices"][0].get("delta", {})
+            
+            if "content" in delta and delta["content"]:
+                full_content += delta["content"]
+                
+            if "reasoning_content" in delta and delta["reasoning_content"]:
+                if full_reasoning_content is None:
+                    full_reasoning_content = ""
+                full_reasoning_content += delta["reasoning_content"]
+                
+        except json.JSONDecodeError:
+            continue
+    
+    return ChatCompletionResponse(
+        model=model,
+        choices=[
+            ChatCompletionChoice(
+                message=ChatMessage(
+                    role="assistant",
+                    content=full_content,
+                    reasoning_content=full_reasoning_content,
+                )
+            )
+        ],
+    )
+
+
+async def error_stream_generator(error_detail: str, status_code: int):
+    """Generate error stream response"""
+    yield f'data: {json.dumps({"error": {"message": error_detail, "type": "tenbin_api_error", "code": status_code}})}\n\n'
+    yield "data: [DONE]\n\n"
+
+
+if __name__ == "__main__":
+    import uvicorn
+
+    # 从环境变量获取端口，默认为 8000
+    port = int(os.environ.get("PORT", 8000))
+
+    # 设置环境变量以启用调试模式
+    if os.environ.get("DEBUG_MODE", "").lower() == "true":
+        DEBUG_MODE = True
+        print("Debug mode enabled via environment variable")
+
+    if not os.path.exists("tenbin.json") and not os.environ.get("SESSION_IDS"):
+        print("Warning: tenbin.json not found and SESSION_IDS not set. Creating a dummy file.")
+        dummy_data = [
+            {
+                "session_id": "your_session_id_here",
+            }
+        ]
+        with open("tenbin.json", "w", encoding="utf-8") as f:
+            json.dump(dummy_data, f, indent=4)
+        print("Created dummy tenbin.json. Please replace with valid Tenbin data or set SESSION_IDS secret.")
+
+    if not os.path.exists("client_api_keys.json") and not os.environ.get("API_KEYS"):
+        print("Warning: client_api_keys.json not found and API_KEYS not set. Creating a dummy file.")
+        dummy_key = f"sk-dummy-{uuid.uuid4().hex}"
+        with open("client_api_keys.json", "w", encoding="utf-8") as f:
+            json.dump([dummy_key], f, indent=2)
+        print(f"Created dummy client_api_keys.json with key: {dummy_key}. Or set API_KEYS secret.")
+
+    if not os.path.exists("models.json"):
+        print("Warning: models.json not found. Creating a dummy file.")
+        dummy_models = {
+            "claude-3.7-sonnet": "AnthropicClaude37Sonnet",
+            "claude-3.7-sonnet-extended": "AnthropicClaude37SonnetExtended"
+        }
+        with open("models.json", "w", encoding="utf-8") as f:
+            json.dump(dummy_models, f, indent=4)
+        print("Created dummy models.json.")
+
+    load_client_api_keys()
+    load_tenbin_accounts()
+    load_tenbin_models()
+
+    print("\n--- Tenbin OpenAI API Adapter ---")
+    print(f"Debug Mode: {DEBUG_MODE}")
+    print("Endpoints:")
+    print("  GET  /v1/models (Client API Key Auth)")
+    print("  GET  /models (No Auth)")
+    print("  POST /v1/chat/completions (Client API Key Auth)")
+    print("  GET  /debug?enable=[true|false] (Toggle Debug Mode)")
+
+    print(f"\nClient API Keys: {len(VALID_CLIENT_KEYS)}")
+    if TENBIN_ACCOUNTS:
+        print(f"Tenbin Accounts: {len(TENBIN_ACCOUNTS)}")
+    else:
+        print("Tenbin Accounts: None loaded. Check tenbin.json.")
+    if TENBIN_MODELS:
+        models = sorted(list(TENBIN_MODELS.keys()))
+        print(f"Tenbin Models: {len(TENBIN_MODELS)}")
+        print(f"Available models: {', '.join(models[:5])}{'...' if len(models) > 5 else ''}")
+    else:
+        print("Tenbin Models: None loaded. Check models.json.")
+    print("------------------------------------")
+
     uvicorn.run(app, host="0.0.0.0", port=port)

resigner.py

@@ -1,135 +1,141 @@
-import time
-import requests
-import uuid
-import random
-import string
-
-TURNSTILE_URL = "http://" # 自行搭建的turnstile solver服务
-API_KEY = "" # mail.eleme.uk的api
-
-def create_task():
-    url = f"{TURNSTILE_URL}/turnstile"
-    resp = requests.get(url, params={
-        "url": "https://oshiete.ai/email_confirmation",
-        "sitekey": "0x4AAAAAABGR2exxRproizri",
-        "action": "request_registration_link"
-    })
-    print(resp.json())
-    return resp.json()['task_id']
-
-
-def get_result(task_id):
-    url = f"{TURNSTILE_URL}/result"
-    resp = requests.get(url, params={
-        "id": task_id
-    })
-    if 'value' in resp.text:
-        return resp.json()['value']
-    return None
-
-def solve_turnstile():
-    task_id = create_task()
-    while True:
-        result = get_result(task_id)
-        if result:
-            return result
-            break
-        time.sleep(1)
-
-def generate_random_string(length=10):
-    characters = string.ascii_letters + string.digits
-    random_string = ''.join(random.choice(characters) for _ in range(length))
-    return random_string
-
-
-def get_email():
-    url = "https://mail.eleme.uk/api/emails/generate"
-    headers = {
-        "X-API-Key": API_KEY,
-        "Content-Type": "application/json"
-    }
-    data = {
-        "name": generate_random_string(8),
-        "expiryTime": 3600000,
-        "domain": "ele.edu.kg"
-    }
-    
-    try:
-        response = requests.post(url, headers=headers, json=data)
-        response.raise_for_status()  # 检查响应状态
-        email_data = response.json()
-        return email_data.get("email"), email_data.get("id")  # 假设API返回包含email字段的JSON
-    except requests.exceptions.RequestException as e:
-        print(f"获取邮箱时出错: {e}")
-        return None
-
-def get_code(id: str):
-    url = f"https://mail.eleme.uk/api/emails/{id}"
-    headers = {
-        "X-API-Key": API_KEY
-    }
-    
-    try:
-        response = requests.get(url, headers=headers)
-        response.raise_for_status()
-        emails_data = response.json()
-        # 提取验证码逻辑可能需要根据实际邮件内容调整
-        for email in emails_data['messages']:
-            if 'yGMO' in email['subject']:
-                # 获取邮件内容的详细信息
-                message_id = email['id']
-                url_message = f"https://mail.eleme.uk/api/emails/{id}/{message_id}"
-                message_response = requests.get(url_message, headers=headers)
-                message_response.raise_for_status()
-                message_data = message_response.json()
-                return message_data['message']['html'].split('code=')[1].split('<br>')[0]
-    except requests.exceptions.RequestException as e:
-        print(f"获取验证码时出错: {e}")
-        return None
-
-def send_email(email: str):
-    url = "https://graphql.oshiete.ai/graphql"
-    resp = requests.post(url, json={
-        "operationName": "RequestRegistrationLink",
-        "variables": {
-            "email": email,
-            "turnstileToken": solve_turnstile()
-        },
-        "query": """
-        mutation RequestRegistrationLink($email: String!, $turnstileToken: String!) {
-            requestRegistrationLink(email: $email, turnstileToken: $turnstileToken)
-        }
-        """
-    })
-
-def register(code: str):
-    url = "https://graphql.oshiete.ai/graphql"
-    resp = requests.post(url, json={
-        "operationName": "RegisterUser",
-        "variables": {
-            "code": code,
-            "dti": str(uuid.uuid4()),
-            "password": "Aa123321."
-        },
-        "query": """
-        mutation RegisterUser($code: String!, $password: String!, $dti: String) {
-            registerUser(code: $code, password: $password, dti: $dti) {
-                id
-                __typename
-            }
-        }
-        """
-    })
-    # 提取sessionId cookie
-    session_id = resp.cookies.get('sessionId')
-    return session_id
-
-if __name__ == "__main__":
-    email, id = get_email()
-    send_email(email)
-    code = None
-    while not code:
-        code = get_code(id)
-        time.sleep(1)
-    session_id = register(code)
+import time
+import requests
+import uuid
+import random
+import string
+import os
+
+TURNSTILE_URL = "http://" # 自行搭建的turnstile solver服务
+API_KEY = "" # mail.eleme.uk的api
+HTTP_PROXY = os.environ.get("HTTP_PROXY")
+PROXIES = {"http": HTTP_PROXY, "https": HTTP_PROXY} if HTTP_PROXY else None
+
+if PROXIES:
+    print(f"Using HTTP proxy: {HTTP_PROXY}")
+
+def create_task():
+    url = f"{TURNSTILE_URL}/turnstile"
+    resp = requests.get(url, params={
+        "url": "https://oshiete.ai/email_confirmation",
+        "sitekey": "0x4AAAAAABGR2exxRproizri",
+        "action": "request_registration_link"
+    })
+    print(resp.json())
+    return resp.json()['task_id']
+
+
+def get_result(task_id):
+    url = f"{TURNSTILE_URL}/result"
+    resp = requests.get(url, params={
+        "id": task_id
+    })
+    if 'value' in resp.text:
+        return resp.json()['value']
+    return None
+
+def solve_turnstile():
+    task_id = create_task()
+    while True:
+        result = get_result(task_id)
+        if result:
+            return result
+            break
+        time.sleep(1)
+
+def generate_random_string(length=10):
+    characters = string.ascii_letters + string.digits
+    random_string = ''.join(random.choice(characters) for _ in range(length))
+    return random_string
+
+
+def get_email():
+    url = "https://mail.eleme.uk/api/emails/generate"
+    headers = {
+        "X-API-Key": API_KEY,
+        "Content-Type": "application/json"
+    }
+    data = {
+        "name": generate_random_string(8),
+        "expiryTime": 3600000,
+        "domain": "ele.edu.kg"
+    }
+    
+    try:
+        response = requests.post(url, headers=headers, json=data, proxies=PROXIES)
+        response.raise_for_status()  # 检查响应状态
+        email_data = response.json()
+        return email_data.get("email"), email_data.get("id")  # 假设API返回包含email字段的JSON
+    except requests.exceptions.RequestException as e:
+        print(f"获取邮箱时出错: {e}")
+        return None
+
+def get_code(id: str):
+    url = f"https://mail.eleme.uk/api/emails/{id}"
+    headers = {
+        "X-API-Key": API_KEY
+    }
+    
+    try:
+        response = requests.get(url, headers=headers, proxies=PROXIES)
+        response.raise_for_status()
+        emails_data = response.json()
+        # 提取验证码逻辑可能需要根据实际邮件内容调整
+        for email in emails_data['messages']:
+            if 'yGMO' in email['subject']:
+                # 获取邮件内容的详细信息
+                message_id = email['id']
+                url_message = f"https://mail.eleme.uk/api/emails/{id}/{message_id}"
+                message_response = requests.get(url_message, headers=headers, proxies=PROXIES)
+                message_response.raise_for_status()
+                message_data = message_response.json()
+                return message_data['message']['html'].split('code=')[1].split('<br>')[0]
+    except requests.exceptions.RequestException as e:
+        print(f"获取验证码时出错: {e}")
+        return None
+
+def send_email(email: str):
+    url = "https://graphql.oshiete.ai/graphql"
+    resp = requests.post(url, json={
+        "operationName": "RequestRegistrationLink",
+        "variables": {
+            "email": email,
+            "turnstileToken": solve_turnstile()
+        },
+        "query": """
+        mutation RequestRegistrationLink($email: String!, $turnstileToken: String!) {
+            requestRegistrationLink(email: $email, turnstileToken: $turnstileToken)
+        }
+        """
+    }, proxies=PROXIES)
+
+def register(code: str):
+    url = "https://graphql.oshiete.ai/graphql"
+    resp = requests.post(url, json={
+        "operationName": "RegisterUser",
+        "variables": {
+            "code": code,
+            "dti": str(uuid.uuid4()),
+            "password": "Aa123321."
+        },
+        "query": """
+        mutation RegisterUser($code: String!, $password: String!, $dti: String) {
+            registerUser(code: $code, password: $password, dti: $dti) {
+                id
+                __typename
+            }
+        }
+        """
+    }, proxies=PROXIES)
+    # 提取sessionId cookie
+    session_id = resp.cookies.get('sessionId')
+    return session_id
+
+if __name__ == "__main__":
+    email, id = get_email()
+    send_email(email)
+    code = None
+    while not code:
+        code = get_code(id)
+        time.sleep(1)
+    session_id = register(code)
     print(session_id)