153 lines
6.6 KiB
Python
153 lines
6.6 KiB
Python
import random
|
||
from http import HTTPStatus
|
||
from dashscope import Generation
|
||
import dashscope
|
||
import gradio as gr
|
||
|
||
# dashscope.api_key = 'sk-73e9b0452a7e40048495d8ac8ab1afe4' # Vincent's API key
|
||
dashscope.api_key = 'sk-83b8ed0ead0849ae9e63a2ae5bdbde0d' # Rayman's API key
|
||
|
||
|
||
def respond(prompt, chat_history, instruction, model, if_stream):
|
||
"""
|
||
与AI助手进行对话，并返回对话历史。
|
||
|
||
参数:
|
||
- prompt: 用户输入的文本。
|
||
- chat_history: 之前的聊天历史，列表，每个元素是二元组，包括用户输入和AI响应。
|
||
- instruction: 系统指令，作为对话的起始信息。
|
||
|
||
返回值:
|
||
- 生成器，每次产生一个二元组，包括空字符串和更新后的聊天历史。
|
||
"""
|
||
# 构建对话消息结构
|
||
messages = [{'role': 'system',
|
||
'content': instruction},
|
||
{'role': 'user',
|
||
'content': prompt}
|
||
]
|
||
full_response = "" # 初始化空字符串以聚合响应
|
||
# 调用AI模型生成响应
|
||
# -------非流式输出-------
|
||
if if_stream == 'Non-Stream':
|
||
response = Generation.call(model=model,
|
||
messages=messages,
|
||
# 设置随机数种子seed，如果没有设置，则随机数种子默认为1234
|
||
seed=1234,
|
||
# 将输出设置为"message"格式
|
||
result_format='message',
|
||
# 设置输出方式为非流式输出
|
||
stream=False,
|
||
# 设置输出方式为非增量式输出
|
||
incremental_output=False)
|
||
if not chat_history or chat_history[-1][0] != prompt:
|
||
chat_history.append((prompt, ""))
|
||
if response.status_code == HTTPStatus.OK:
|
||
# 获取响应中的消息内容
|
||
message = response.output.choices[0]['message']['content']
|
||
# 将消息内容添加到聊天历史中
|
||
chat_history.append((prompt, message))
|
||
# 返回更新后的聊天历史
|
||
return "", chat_history
|
||
elif if_stream == 'Stream':
|
||
# -------流式输出-------
|
||
responses = Generation.call(model=model,
|
||
messages=messages,
|
||
# 设置随机数种子seed，如果没有设置，则随机数种子默认为1234
|
||
seed=1234,
|
||
# 将输出设置为"message"格式
|
||
result_format='message',
|
||
stream=True, # 设置输出方式为流式输出
|
||
incremental_output=True, # 增量式流式输出
|
||
temperature=1.8,
|
||
top_p=0.9,
|
||
top_k=999)
|
||
# 确保聊天历史至少有当前会话的开始
|
||
if not chat_history or chat_history[-1][0] != prompt:
|
||
chat_history.append((prompt, ""))
|
||
# 循环处理每个流式响应
|
||
for response in responses:
|
||
if response.status_code == HTTPStatus.OK:
|
||
# 累加每次流式响应的内容
|
||
text = response.output.choices[0]['message']['content']
|
||
full_response += text
|
||
# 更新聊天历史的最后一项
|
||
last_turn = list(chat_history[-1])
|
||
last_turn[1] = full_response
|
||
chat_history[-1] = tuple(last_turn)
|
||
yield "", chat_history # 实时输出当前的聊天历史
|
||
else:
|
||
# 如果出错，构建错误信息并更新最后一项
|
||
full_response = 'Request id: {}, Status code: {}, error code: {}, error message: {}'.format(
|
||
response.request_id, response.status_code,
|
||
response.code, response.message
|
||
)
|
||
last_turn = list(chat_history[-1])
|
||
last_turn[1] = full_response
|
||
chat_history[-1] = tuple(last_turn)
|
||
yield "", chat_history
|
||
break # 出现错误时终止循环
|
||
|
||
|
||
def respond_nonStream(prompt, chat_history, instruction, model):
|
||
# 构建对话消息结构
|
||
messages = [{'role': 'system',
|
||
'content': instruction},
|
||
{'role': 'user',
|
||
'content': prompt}
|
||
]
|
||
full_response = "" # 初始化空字符串以聚合响应
|
||
# 调用AI模型生成响应
|
||
responses = Generation.call(model=model,
|
||
messages=messages,
|
||
# 设置随机数种子seed，如果没有设置，则随机数种子默认为1234
|
||
seed=1234,
|
||
# 将输出设置为"message"格式
|
||
result_format='message',
|
||
stream=True, # 设置输出方式为流式输出
|
||
incremental_output=True, # 增量式流式输出
|
||
temperature=1.8,
|
||
top_p=0.9,
|
||
top_k=999)
|
||
|
||
|
||
|
||
llm_model_list = ['qwen-turbo','qwen-plus', 'qwen-max']
|
||
init_llm = llm_model_list[0]
|
||
|
||
# 创建 Gradio 界面
|
||
with gr.Blocks() as demo:
|
||
gr.Markdown(
|
||
"""
|
||
# AI TestGenius
|
||
A simple LLM app for generating test cases from function design.
|
||
""")
|
||
chatbot = gr.Chatbot()
|
||
msg = gr.Textbox(label="Prompt")
|
||
with gr.Accordion(label="Advanced options", open=False):
|
||
system = gr.Textbox(label="System prompts", lines=2,
|
||
value="A conversation between a user and an LLM-based AI assistant. The assistant gives helpful and honest answers.")
|
||
llm = gr.Dropdown(
|
||
llm_model_list,
|
||
label='Choose LLM Model',
|
||
value=init_llm,
|
||
interactive=True
|
||
)
|
||
if_stream = gr.Dropdown(
|
||
["Stream", "Non-Stream"],
|
||
label='Choose Streaming',
|
||
value="Stream",
|
||
interactive=True
|
||
)
|
||
|
||
btn = gr.Button("Submit")
|
||
clear = gr.ClearButton(components=[msg, chatbot], value="Clear console")
|
||
btn.click(respond, inputs=[msg, chatbot, system, llm, if_stream], outputs=[msg, chatbot]) # click to submit
|
||
msg.submit(respond, inputs=[msg, chatbot, system, llm, if_stream], outputs=[msg, chatbot]) # Press enter to submit
|
||
|
||
|
||
# 运行界面
|
||
if __name__ == "__main__":
|
||
gr.close_all()
|
||
demo.launch()
|