audrius/app-odoo
1
0
Fork 0
mirror of https://github.com/guohuadeng/app-odoo.git synced 2025-02-23 04:11:36 +02:00
Code Issues Packages Projects Releases Wiki Activity

fix #I6WHKN [app_chatgpt]敏感语处理有问题，要修 置顶

Browse Source
This commit is contained in:
ivan deng
2023-04-18 01:16:41 +08:00
parent 3072c7399b
commit 437da3d5b3
3 changed files with 48 additions and 46 deletions
Download Patch File Download Diff File Expand all files Collapse all files

69
app_chatgpt/models/ai_robot.py
View File

@@ -40,7 +40,7 @@ GPT-3 A set of models that can understand and generate natural language
# begin gpt 参数
# 1. stop：表示聊天机器人停止生成回复的条件，可以是一段文本或者一个列表，当聊天机器人生成的回复中包含了这个条件，就会停止继续生成回复。
# 2. temperature：控制回复的“新颖度”，值越高，聊天机器人生成的回复越不确定和随机，值越低，聊天机器人生成的回复会更加可预测和常规化。
# 3. top_p：与temperature有些类似，也是控制回复的“新颖度”。不同的是，top_p控制的是回复中概率最高的几个可能性的累计概率之和，值越小，生成的回复越保守，值越大，生成的回复越新颖。
# 3. top_p：言语连贯性，与temperature有些类似，也是控制回复的“新颖度”。不同的是，top_p控制的是回复中概率最高的几个可能性的累计概率之和，值越小，生成的回复越保守，值越大，生成的回复越新颖。
# 4. frequency_penalty：用于控制聊天机器人回复中出现频率过高的词汇的惩罚程度。聊天机器人会尝试避免在回复中使用频率较高的词汇，以提高回复的多样性和新颖度。
# 5. presence_penalty：与frequency_penalty相对，用于控制聊天机器人回复中出现频率较低的词汇的惩罚程度。聊天机器人会尝试在回复中使用频率较低的词汇，以提高回复的多样性和新颖度。
max_tokens = fields.Integer('Max response', default=600,
@@ -50,7 +50,7 @@ GPT-3 A set of models that can understand and generate natural language
(including system message, examples, message history, and user query) and the model's response.
One token is roughly 4 characters for typical English text.
""")
temperature = fields.Float(string='Temperature', default=0.9,
temperature = fields.Float(string='Temperature', default=0.8,
help="""
Controls randomness. Lowering the temperature means that the model will produce
more repetitive and deterministic responses.
@@ -104,39 +104,47 @@ GPT-3 A set of models that can understand and generate natural language
def action_disconnect(self):
requests.delete('https://chatgpt.com/v1/disconnect')
def get_ai(self, data, author_id=False, answer_id=False, param={}):
# 通用方法
# author_id: 请求的 partner_id 对象
# answer_id: 回答的 partner_id 对象
# kwargs，dict 形式的可变参数
self.ensure_one()
# 前置勾子，一般返回 False，有问题返回响应内容
res_pre = self.get_ai_pre(data, author_id, answer_id, param)
if res_pre:
return res_pre
if hasattr(self, 'get_%s' % self.provider):
res = getattr(self, 'get_%s' % self.provider)(data, author_id, answer_id, param)
else:
res = _('No robot provider found')
# 后置勾子，返回处理后的内容，用于处理敏感词等
res_post = self.get_ai_post(res, author_id, answer_id, param)
return res_post
def get_ai_pre(self, data, author_id=False, answer_id=False, param={}):
if self.is_filtering:
search = WordsSearch()
search.SetKeywords([])
content = data[0]['content']
if isinstance(data, list):
content = data[len(data)-1]['content']
else:
content = data
sensi = search.FindFirst(content)
if sensi is not None:
_logger.error('==========敏感词：%s' % sensi['Keyword'])
return _('温馨提示：您发送的内容含有敏感词，请修改内容后再向我发送。')
else:
return False
def get_ai(self, data, author_id=False, answer_id=False, param={}):
# 通用方法
# author_id: 请求的 partner_id 对象
# answer_id: 回答的 partner_id 对象
# param，dict 形式的参数
# 调整输出为2个参数：res_post详细内容，is_ai是否ai的响应
self.ensure_one()
# 前置勾子，一般返回 False，有问题返回响应内容，用于处理敏感词等
res_pre = self.get_ai_pre(data, author_id, answer_id, param)
if res_pre:
# 有错误内容，则返回上级内容及 is_ai为假
return res_pre, False
if not hasattr(self, 'get_%s' % self.provider):
res = _('No robot provider found')
return res, False
res = getattr(self, 'get_%s' % self.provider)(data, author_id, answer_id, param)
# 后置勾子，返回处理后的内容
res_post, is_ai = self.get_ai_post(res, author_id, answer_id, param)
return res_post, is_ai
def get_ai_post(self, res, author_id=False, answer_id=False, param={}):
if res and author_id and isinstance(res, openai.openai_object.OpenAIObject) or isinstance(res, list):
# 返回是个对象，那么就是ai
usage = json.loads(json.dumps(res['usage']))
content = json.loads(json.dumps(res['choices'][0]['message']['content']))
data = content.replace(' .', '.').strip()
@@ -169,9 +177,10 @@ GPT-3 A set of models that can understand and generate natural language
'first_ask_time': ask_date
})
ai_use.write(vals)
return data, True
else:
data = res
return data
# 直接返回错误语句，那么就是非ai
return res, False
def get_ai_system(self, content=None):
# 获取基础ai角色设定, role system
@@ -221,7 +230,7 @@ GPT-3 A set of models that can understand and generate natural language
# 处理传参，传过来的优先于 robot 默认的
max_tokens = param.get('max_tokens') or self.max_tokens or 600,
temperature = param.get('temperature') or self.temperature or 0.9,
temperature = param.get('temperature') or self.temperature or 0.8,
top_p = param.get('top_p') or self.top_p or 0.6,
frequency_penalty = param.get('frequency_penalty') or self.frequency_penalty or 0.5,
presence_penalty = param.get('presence_penalty') or self.presence_penalty or 0.5,
@@ -250,7 +259,7 @@ GPT-3 A set of models that can understand and generate natural language
model=self.ai_model,
messages=messages,
n=1,
temperature=self.temperature or 0.9,
temperature=self.temperature or 0.8,
max_tokens=self.max_tokens or 600,
top_p=self.top_p or 0.6,
frequency_penalty=self.frequency_penalty or 0.5,
@@ -276,7 +285,7 @@ GPT-3 A set of models that can understand and generate natural language
pdata = {
"model": self.ai_model,
"prompt": data,
"temperature": 0.9,
"temperature": 0.8,
"max_tokens": max_tokens,
"top_p": 1,
"frequency_penalty": 0.0,
@@ -311,9 +320,9 @@ GPT-3 A set of models that can understand and generate natural language
else:
messages = [{"role": "user", "content": data}]
# 处理传参，传过来的优先于 robot 默认的
# todo: 处理传参，传过来的优先于 robot 默认的，当前有问题，无法做tuple转换
max_tokens = param.get('max_tokens') or self.max_tokens or 600,
temperature = param.get('temperature') or self.temperature or 0.9,
temperature = param.get('temperature') or self.temperature or 0.8,
top_p = param.get('top_p') or self.top_p or 0.6,
frequency_penalty = param.get('frequency_penalty') or self.frequency_penalty or 0.5,
presence_penalty = param.get('presence_penalty') or self.presence_penalty or 0.5,
@@ -330,7 +339,7 @@ GPT-3 A set of models that can understand and generate natural language
messages=messages,
# 返回的回答数量
n=1,
temperature=self.temperature or 0.9,
temperature=self.temperature or 0.8,
max_tokens=self.max_tokens or 600,
top_p=self.top_p or 0.6,
frequency_penalty=self.frequency_penalty or 0.5,