整理ollma接口，添加token个数限制接口

AIGC/AICore.py

@@ -0,0 +1,21 @@
+import requests
+from ollama import Client, ResponseError
+import tiktoken
+
+class AICore:
+    modelMaxTokens = 128000
+    # 初始化 DeepSeek 使用的 Tokenizer (cl100k_base)
+    encoder = tiktoken.get_encoding("cl100k_base")
+
+     
+
+    def __init__(self, model):
+         #初始化ollama客户端
+        ollamaClient = Client(host='http://localhost:11434', headers={'x-some-header': 'some-value'})
+        response = ollamaClient.show(model)
+        modelMaxTokens = response.modelinfo['qwen2.context_length']
+        
+    def getPromptToken(self, prompt)-> int: 
+        tokens = self.encoder.encode(prompt)
+        return len(tokens)
+   

AIGC/main.py

@@ -11,8 +11,9 @@ from fastapi import FastAPI, Request, HTTPException, WebSocket, WebSocketDisconn
 from fastapi.websockets import WebSocketState
 from h11 import ConnectionClosed
 import uvicorn
+from AICore import AICore
 from Utils.AIGCLog import AIGCLog
-from ollama import Client, ResponseError
+
 
 app = FastAPI(title = "AI 通信服务")
 logger = AIGCLog(name = "AIGC", log_file = "aigc.log")
@@ -27,8 +28,8 @@ logger.log(logging.INFO, f"使用的模型是 {args.model}")
 maxAIRegerateCount = 5
 lastPrompt = ""
 
-#初始化ollama客户端
+aicore = AICore(args.model)
-ollamaClient = Client(host='http://localhost:11434')
+
 
 async def heartbeat(websocket: WebSocket):
     pass
@@ -150,6 +151,7 @@ async def generateAIChat(promptStr: str, websocket: WebSocket| None = None):
             {"role": "system", "content": promptStr}
         ]
     try:
+
         # response = ollamaClient.chat(
         #     model = args.model,
         #     stream = False,
@@ -235,6 +237,7 @@ if __name__ == "__main__":
     server_thread.start()
 
     # Test
+    aicore.getPromptToken("测试功能")
     asyncio.run(
         generateAIChat(promptStr = f"""
             #你是一个游戏NPC对话生成器。请严格按以下要求生成两个角色的日常对话

AIGC/requirements.txt

@@ -1,3 +1,4 @@
 uvicorn[standard]
 fastapi
-ollama
+ollama
+tiktoken