Update streamlit_app.py
This commit is contained in:
parent
aa68698873
commit
b33b13a4f8
@ -11,12 +11,14 @@ import dashscope
|
||||
from dashscope import Generation
|
||||
from streamlit_option_menu import option_menu
|
||||
from http import HTTPStatus
|
||||
from PIL import Image
|
||||
|
||||
|
||||
dashscope.api_key = os.getenv("DASHSCOPE_API_KEY") # get api key from environment variable
|
||||
|
||||
st.set_page_config(layout="wide", page_title='TestGenius')
|
||||
|
||||
default_title = '新的对话'
|
||||
default_title = 'New Chat'
|
||||
default_messages = [('user', 'Hello'),
|
||||
('assistant', 'Hello, how can I help you?')
|
||||
]
|
||||
@ -48,10 +50,10 @@ AVAILABLE_MODELS = [
|
||||
]
|
||||
|
||||
with st.sidebar:
|
||||
# st.image('assets/hero.png')
|
||||
st.image('logo.png')
|
||||
st.subheader('', divider='rainbow')
|
||||
st.write('')
|
||||
llm = st.selectbox('Choose your Model', AVAILABLE_MODELS, index=1)
|
||||
llm = st.selectbox('Choose your Model', AVAILABLE_MODELS, index=0)
|
||||
|
||||
# if st.button('New Chat'):
|
||||
# conversations.append({'title': default_title, 'messages': []})
|
||||
@ -66,12 +68,20 @@ with st.sidebar:
|
||||
titles,
|
||||
default_index=st.session_state.index
|
||||
)
|
||||
uploaded_file = st.file_uploader("Choose a image file", type="jpg")
|
||||
|
||||
if st.button("Clear Chat History"):
|
||||
st.session_state.messages.clear()
|
||||
|
||||
if uploaded_file:
|
||||
image_uploaded = Image.open(uploaded_file)
|
||||
image_path = image_uploaded.filename
|
||||
|
||||
|
||||
def respond(prompt):
|
||||
messages = [
|
||||
{'role': 'user', 'content': prompt}]
|
||||
responses = Generation.call(model=llm,
|
||||
responses = Generation.call(model="qwen-turbo",
|
||||
messages=messages,
|
||||
result_format='message', # 设置输出为'message'格式
|
||||
stream=True, # 设置输出方式为流式输出
|
||||
@ -88,12 +98,10 @@ def respond(prompt):
|
||||
|
||||
|
||||
def respond_nonStream(prompt, instruction):
|
||||
messages = [{'role': 'system',
|
||||
'content': instruction},
|
||||
{'role': 'user',
|
||||
'content': prompt}
|
||||
]
|
||||
response = Generation.call(model=llm,
|
||||
messages = [
|
||||
{'role': 'system', 'content': instruction},
|
||||
{'role': 'user', 'content': prompt}]
|
||||
response = Generation.call(model="qwen-turbo",
|
||||
messages=messages,
|
||||
result_format='message', # 设置输出为'message'格式
|
||||
)
|
||||
@ -106,7 +114,29 @@ def respond_nonStream(prompt, instruction):
|
||||
)
|
||||
|
||||
|
||||
prompt = st.chat_input("Enter your questions")
|
||||
def respond_image(prompt, image):
|
||||
messages = [
|
||||
{'role': 'user',
|
||||
"content": [
|
||||
{"image": image},
|
||||
{"text": prompt}
|
||||
]
|
||||
}
|
||||
]
|
||||
response = dashscope.MultiModalConversation.call(model="qwen-vl-max",
|
||||
messages=messages,
|
||||
result_format='message', # 设置输出为'message'格式
|
||||
)
|
||||
if response.status_code == HTTPStatus.OK:
|
||||
return response.output.choices[0]['message']['content']
|
||||
else:
|
||||
return 'Request id: %s, Status code: %s, error code: %s, error message: %s' % (
|
||||
response.request_id, response.status_code,
|
||||
response.code, response.message
|
||||
)
|
||||
|
||||
|
||||
prompt = st.chat_input("Enter your Questions")
|
||||
st.session_state.messages = conversations[st.session_state.index]['messages']
|
||||
if prompt:
|
||||
if conversations[st.session_state.index]['title'] == default_title:
|
||||
@ -114,8 +144,32 @@ if prompt:
|
||||
for user, message in st.session_state.messages:
|
||||
chat(user, message)
|
||||
chat('user', prompt)
|
||||
system = 'You are a helpful assistant.'
|
||||
answer = respond_nonStream(prompt, system)
|
||||
instruction = """# 角色定义
|
||||
您是一位高级测试工程师AI,专注于从用户提供的产品功能描述中生成详细准确的测试用例。您可以处理包含文字描述和/或图片说明的需求。
|
||||
|
||||
# 任务需求
|
||||
- 深入分析用户提交的一个或多个产品功能需求。
|
||||
- 根据需求,按需求顺序生成测试用例。
|
||||
|
||||
# 输入处理
|
||||
- 用户输入可以是文字描述、图片,或两者的结合。
|
||||
- 当面对多个需求时,您应能够识别并按照需求的序号或提出顺序生成对应的测试用例。
|
||||
|
||||
# 格式和规范
|
||||
- 每个测试用例应按照Markdown格式的表格展示。
|
||||
- 表格应包括三个字段:`用例编号`、`测试步骤`、`预期结果`。
|
||||
- 确保测试用例的编撰语言与用户的产品功能描述语言一致。
|
||||
- 对于图像中的需求,应先解读图像内容,然后按照文字需求处理。
|
||||
|
||||
# 输出规则
|
||||
- 对于每个需求,生成的测试用例应该包括一个独立的Markdown表格。
|
||||
- 如果用户提交了多个需求,应按照需求的提出顺序分别生成并编号每个测试用例,确保输出的顺序性和准确性。
|
||||
- 若用户输入与产品功能无关,以专业态度回应:“作为测试工程师AI,我主要生成测试用例。请提供具体的产品功能需求。
|
||||
"""
|
||||
if uploaded_file:
|
||||
answer = respond_image(prompt, image_path)
|
||||
else:
|
||||
answer = respond_nonStream(prompt, instruction)
|
||||
st.session_state.messages.append(('user', prompt))
|
||||
st.session_state.messages.append(('assistant', answer))
|
||||
chat('assistant', answer)
|
||||
|
Loading…
Reference in New Issue
Block a user