feat: support character description of ali qwen model

bot/tongyi/ali_qwen_session.py

@@ -0,0 +1,62 @@
+from bot.session_manager import Session
+from common.log import logger
+
+"""
+    e.g.
+    [
+        {"role": "system", "content": "You are a helpful assistant."},
+        {"role": "user", "content": "Who won the world series in 2020?"},
+        {"role": "assistant", "content": "The Los Angeles Dodgers won the World Series in 2020."},
+        {"role": "user", "content": "Where was it played?"}
+    ]
+"""
+
+class AliQwenSession(Session):
+    def __init__(self, session_id, system_prompt=None, model="qianwen"):
+        super().__init__(session_id, system_prompt)
+        self.model = model
+        self.reset()
+
+    def discard_exceeding(self, max_tokens, cur_tokens=None):
+        precise = True
+        try:
+            cur_tokens = self.calc_tokens()
+        except Exception as e:
+            precise = False
+            if cur_tokens is None:
+                raise e
+            logger.debug("Exception when counting tokens precisely for query: {}".format(e))
+        while cur_tokens > max_tokens:
+            if len(self.messages) > 2:
+                self.messages.pop(1)
+            elif len(self.messages) == 2 and self.messages[1]["role"] == "assistant":
+                self.messages.pop(1)
+                if precise:
+                    cur_tokens = self.calc_tokens()
+                else:
+                    cur_tokens = cur_tokens - max_tokens
+                break
+            elif len(self.messages) == 2 and self.messages[1]["role"] == "user":
+                logger.warn("user message exceed max_tokens. total_tokens={}".format(cur_tokens))
+                break
+            else:
+                logger.debug("max_tokens={}, total_tokens={}, len(messages)={}".format(max_tokens, cur_tokens, len(self.messages)))
+                break
+            if precise:
+                cur_tokens = self.calc_tokens()
+            else:
+                cur_tokens = cur_tokens - max_tokens
+        return cur_tokens
+
+    def calc_tokens(self):
+        return num_tokens_from_messages(self.messages, self.model)
+
+def num_tokens_from_messages(messages, model):
+    """Returns the number of tokens used by a list of messages."""
+    # 官方token计算规则："对于中文文本来说，1个token通常对应一个汉字；对于英文文本来说，1个token通常对应3至4个字母或1个单词"
+    # 详情请产看文档：https://help.aliyun.com/document_detail/2586397.html
+    # 目前根据字符串长度粗略估计token数，不影响正常使用
+    tokens = 0
+    for msg in messages:
+        tokens += len(msg["content"])
+    return tokens

bot/tongyi/tongyi_qwen_bot.py

@@ -10,7 +10,7 @@ import broadscope_bailian
 from broadscope_bailian import ChatQaMessage
 
 from bot.bot import Bot
-from bot.baidu.baidu_wenxin_session import BaiduWenxinSession
+from bot.tongyi.ali_qwen_session import AliQwenSession
 from bot.session_manager import SessionManager
 from bridge.context import ContextType
 from bridge.reply import Reply, ReplyType
@@ -27,7 +27,7 @@ class TongyiQwenBot(Bot):
         self.node_id = conf().get("qwen_node_id") or ""
         self.api_key_client = broadscope_bailian.AccessTokenClient(access_key_id=self.access_key_id, access_key_secret=self.access_key_secret)
         self.api_key_expired_time = self.set_api_key()
-        self.sessions = SessionManager(BaiduWenxinSession, model=conf().get("model") or "qwen")
+        self.sessions = SessionManager(AliQwenSession, model=conf().get("model") or "qwen")
         self.temperature = conf().get("temperature", 0.2) # 值在[0,1]之间，越大表示回复越具有不确定性
         self.top_p = conf().get("top_p", 1)
 
@@ -76,7 +76,7 @@ class TongyiQwenBot(Bot):
             reply = Reply(ReplyType.ERROR, "Bot不支持处理{}类型的消息".format(context.type))
             return reply
 
-    def reply_text(self, session: BaiduWenxinSession, retry_count=0) -> dict:
+    def reply_text(self, session: AliQwenSession, retry_count=0) -> dict:
         """
         call bailian's ChatCompletion to get the answer
         :param session: a conversation session
@@ -140,6 +140,7 @@ class TongyiQwenBot(Bot):
         history = []
         user_content = ''
         assistant_content = ''
+        system_content = ''
         for message in messages:
             role = message.get('role')
             if role == 'user':
@@ -149,8 +150,16 @@ class TongyiQwenBot(Bot):
                 history.append(ChatQaMessage(user_content, assistant_content))
                 user_content = ''
                 assistant_content = ''
+            elif role =='system':
+                system_content += message.get('content')
         if user_content == '':
             raise Exception('no user message')
+        if system_content != '':
+            # NOTE 模拟系统消息，测试发现人格描述以"你需要扮演ChatGPT"开头能够起作用，而以"你是ChatGPT"开头模型会直接否认
+            system_qa = ChatQaMessage(system_content, '好的，我会严格按照你的设定回答问题')
+            history.insert(0, system_qa)
+        logger.debug("[TONGYI] converted qa messages: {}".format([item.to_dict() for item in history]))
+        logger.debug("[TONGYI] user content as prompt: {}".format(user_content))
         return user_content, history
 
     def get_completion_content(self, response, node_id):