Update app.py

app.py

@@ -108,19 +108,14 @@ def generate_completion(prompt, max_tokens=256, temperature=0.7):
         # 记录文本类型和长度
         logger.info(f"输入文本类型: {type(text)}, 长度: {len(text)}")
         
-        # 使用更安全的编码方式
-        # 确保输入是字符串类型
-        input_text = str(text)
-        
-        # 编码输入
-        inputs = tokenizer(
-            input_text,
-            return_tensors="pt",
-            truncation=True,
-            max_length=1024,
-            padding=True
-        )
-        inputs = inputs.to(DEVICE)
+        # 使用更基础的编码方式
+        input_ids = tokenizer.encode(text, truncation=True, max_length=1024, return_tensors="pt")
+        attention_mask = torch.ones_like(input_ids)
+        
+        inputs = {
+            "input_ids": input_ids.to(DEVICE),
+            "attention_mask": attention_mask.to(DEVICE)
+        }
         
         # 生成响应
         with torch.no_grad():
@@ -153,9 +148,9 @@ def generate_completion(prompt, max_tokens=256, temperature=0.7):
 
 # 创建极简FastAPI应用
 app = FastAPI(
-    title="OpenAI Completions API兼容服务",
+    title="OpenAI API兼容服务",
     version="1.0",
-    description="专为OpenClaw优化的OpenAI Completions API兼容服务"
+    description="专为OpenClaw优化的OpenAI API兼容服务"
 )
 
 # 启动时加载模型
@@ -180,95 +175,49 @@ async def health_check():
 @app.get("/")
 async def root():
     return {
-        "message": "OpenAI Completions API兼容服务运行中",
+        "message": "OpenAI API兼容服务运行中",
         "model_loaded": model is not None,
         "api_auth_enabled": API_AUTH_ENABLED,
         "endpoints": {
-            "completions": "/v1/completions",
+            "v1": "/v1",
             "chat_completions": "/v1/chat/completions"
         }
     }
 
-# OpenAI Completions端点（OpenClaw主要使用这个）
-@app.post("/v1/completions")
-async def create_completion(
-    request: Request,
-    auth_valid: bool = Depends(verify_api_key)
-):
-    """OpenAI Completions API兼容端点"""
-    try:
-        # 解析请求
-        data = await request.json()
-        prompt = data.get("prompt", "")
-        model_name = data.get("model", "qwen1.5-0.5b-chat")
-        max_tokens = data.get("max_tokens", MAX_TOKENS)
-        temperature = data.get("temperature", 0.7)
-        
-        if not prompt:
-            return JSONResponse(
-                status_code=400,
-                content={
-                    "error": {
-                        "message": "缺少必需的参数: prompt",
-                        "type": "invalid_request_error"
-                    }
-                }
-            )
-        
-        # 生成响应
-        result = generate_completion(prompt, max_tokens, temperature)
-        
-        if "error" in result:
-            return JSONResponse(
-                status_code=500,
-                content={
-                    "error": {
-                        "message": result["error"],
-                        "type": "internal_error"
-                    }
-                }
-            )
-        
-        # 返回OpenAI Completions兼容格式
-        return {
-            "id": f"cmpl-{int(time.time())}",
-            "object": "text_completion",
-            "created": int(time.time()),
-            "model": model_name,
-            "choices": [
-                {
-                    "text": result["text"],
-                    "index": 0,
-                    "logprobs": None,
-                    "finish_reason": "stop"
-                }
-            ],
-            "usage": {
-                "prompt_tokens": len(tokenizer.encode(prompt)) if tokenizer else 0,
-                "completion_tokens": len(tokenizer.encode(result["text"])) if tokenizer else 0,
-                "total_tokens": len(tokenizer.encode(prompt)) + len(tokenizer.encode(result["text"])) if tokenizer else 0
-            }
+# 添加/v1端点（OpenClaw可能需要）
+@app.get("/v1")
+async def v1_root():
+    return {
+        "message": "OpenAI v1 API端点",
+        "endpoints": {
+            "models": "/v1/models",
+            "chat_completions": "/v1/chat/completions"
         }
-        
-    except Exception as e:
-        logger.error(f"Completions API错误: {e}")
-        return JSONResponse(
-            status_code=500,
-            content={
-                "error": {
-                    "message": f"内部服务器错误: {str(e)}",
-                    "type": "internal_error"
-                }
+    }
+
+# 添加模型列表端点（OpenAI兼容）
+@app.get("/v1/models")
+async def list_models():
+    """返回可用的模型列表"""
+    return {
+        "object": "list",
+        "data": [
+            {
+                "id": "qwen1.5-0.5b-chat",
+                "object": "model",
+                "created": int(time.time()),
+                "owned_by": "qwen"
             }
-        )
+        ]
+    }
 
-# OpenAI Chat Completions端点
+# OpenAI Chat Completions端点（主要端点）
 @app.post("/v1/chat/completions")
 async def create_chat_completion(
     request: Request,
     auth_valid: bool = Depends(verify_api_key)
 ):
-    """OpenAI Chat Completions API兼容端点"""
+    """OpenAI Chat Completions API兼容端点 - 这是OpenClaw使用的主要端点"""
     try:
         # 解析请求
         data = await request.json()
@@ -277,6 +226,9 @@ async def create_chat_completion(
         max_tokens = data.get("max_tokens", MAX_TOKENS)
         temperature = data.get("temperature", 0.7)
         
+        # 记录请求详情
+        logger.info(f"收到Chat Completions请求: model={model_name}, messages_count={len(messages)}")
+        
         # 检查消息格式
         if not messages or not isinstance(messages, list):
             return JSONResponse(
@@ -322,7 +274,7 @@ async def create_chat_completion(
             )
         
         # 返回OpenAI Chat Completions兼容格式
-        return {
+        response_data = {
             "id": f"chatcmpl-{int(time.time())}",
             "object": "chat.completion",
             "created": int(time.time()),
@@ -344,6 +296,9 @@ async def create_chat_completion(
             }
         }
         
+        logger.info(f"成功生成响应: {len(result['text'])} 字符")
+        return response_data
+        
     except Exception as e:
         logger.error(f"Chat Completions API错误: {e}")
         return JSONResponse(
@@ -356,13 +311,23 @@ async def create_chat_completion(
             }
         )
 
-# 添加OpenClaw专用端点（简化版）
+# 添加兼容性端点（为不同版本的OpenClaw提供支持）
+@app.post("/chat/completions")
+async def legacy_chat_completion(
+    request: Request,
+    auth_valid: bool = Depends(verify_api_key)
+):
+    """兼容旧版本OpenClaw的端点"""
+    # 直接转发到/v1/chat/completions
+    return await create_chat_completion(request, auth_valid)
+
+# 添加通用聊天端点
 @app.post("/api/chat")
-async def openclaw_chat_api(
+async def generic_chat_api(
     request: Request,
     auth_valid: bool = Depends(verify_api_key)
 ):
-    """专为OpenClaw设计的API端点"""
+    """通用聊天API端点"""
     try:
         # 解析请求
         data = await request.json()
@@ -403,7 +368,7 @@ async def openclaw_chat_api(
                 }
             )
         
-        # 返回OpenClaw专用格式
+        # 返回通用格式
         return {
             "choices": [{
                 "message": {
@@ -413,7 +378,7 @@ async def openclaw_chat_api(
         }
         
     except Exception as e:
-        logger.error(f"OpenClaw API错误: {e}")
+        logger.error(f"通用聊天API错误: {e}")
         return JSONResponse(
             status_code=500,
             content={