IRS-chat / app.py
github-actions
Sync updates from source repository
a6b4aa3
raw
history blame
7.22 kB
from omegaconf import OmegaConf
from query import VectaraQuery
import os
from PIL import Image
import uuid
import streamlit as st
from streamlit_pills import pills
from streamlit_feedback import streamlit_feedback
from utils import thumbs_feedback, send_amplitude_data, escape_dollars_outside_latex
max_examples = 6
languages = {'English': 'eng', 'Spanish': 'spa', 'French': 'fra', 'Chinese': 'zho', 'German': 'deu', 'Hindi': 'hin', 'Arabic': 'ara',
'Portuguese': 'por', 'Italian': 'ita', 'Japanese': 'jpn', 'Korean': 'kor', 'Russian': 'rus', 'Turkish': 'tur', 'Persian (Farsi)': 'fas',
'Vietnamese': 'vie', 'Thai': 'tha', 'Hebrew': 'heb', 'Dutch': 'nld', 'Indonesian': 'ind', 'Polish': 'pol', 'Ukrainian': 'ukr',
'Romanian': 'ron', 'Swedish': 'swe', 'Czech': 'ces', 'Greek': 'ell', 'Bengali': 'ben', 'Malay (or Malaysian)': 'msa', 'Urdu': 'urd'}
# Setup for HTTP API Calls to Amplitude Analytics
if 'device_id' not in st.session_state:
st.session_state.device_id = str(uuid.uuid4())
if "feedback_key" not in st.session_state:
st.session_state.feedback_key = 0
def isTrue(x) -> bool:
if isinstance(x, bool):
return x
return x.strip().lower() == 'true'
def launch_bot():
def reset():
st.session_state.messages = [{"role": "assistant", "content": "How may I help you?"}]
st.session_state.ex_prompt = None
st.session_state.first_turn = True
def generate_response(question):
response = vq.submit_query(question, languages[st.session_state.language])
return response
def generate_streaming_response(question):
response = vq.submit_query_streaming(question, languages[st.session_state.language])
return response
def show_example_questions():
if len(st.session_state.example_messages) > 0 and st.session_state.first_turn:
selected_example = pills("Queries to Try:", st.session_state.example_messages, index=None)
if selected_example:
st.session_state.ex_prompt = selected_example
st.session_state.first_turn = False
return True
return False
if 'cfg' not in st.session_state:
corpus_keys = str(os.environ['corpus_keys']).split(',')
cfg = OmegaConf.create({
'corpus_keys': corpus_keys,
'api_key': str(os.environ['api_key']),
'title': os.environ['title'],
'source_data_desc': os.environ['source_data_desc'],
'streaming': isTrue(os.environ.get('streaming', False)),
'prompt_name': os.environ.get('prompt_name', None),
'examples': os.environ.get('examples', None),
'language': 'English'
})
st.session_state.cfg = cfg
st.session_state.ex_prompt = None
st.session_state.first_turn = True
st.session_state.language = cfg.language
example_messages = [example.strip() for example in cfg.examples.split(",")]
st.session_state.example_messages = [em for em in example_messages if len(em)>0][:max_examples]
st.session_state.vq = VectaraQuery(cfg.api_key, cfg.corpus_keys, cfg.prompt_name)
cfg = st.session_state.cfg
vq = st.session_state.vq
st.set_page_config(page_title=cfg.title, layout="wide")
# left side content
with st.sidebar:
image = Image.open('Vectara-logo.png')
st.image(image, width=175)
st.markdown(f"## About\n\n"
f"This demo uses Retrieval Augmented Generation to ask questions about {cfg.source_data_desc}\n")
cfg.language = st.selectbox('Language:', languages.keys())
if st.session_state.language != cfg.language:
st.session_state.language = cfg.language
reset()
st.rerun()
st.markdown("\n")
bc1, _ = st.columns([1, 1])
with bc1:
if st.button('Start Over'):
reset()
st.rerun()
st.markdown("---")
st.markdown(
"## How this works?\n"
"This app was built with [Vectara](https://vectara.com).\n"
"Vectara's [Indexing API](https://docs.vectara.com/docs/api-reference/indexing-apis/indexing) was used to ingest the data into a Vectara corpus (or index).\n\n"
"This app uses Vectara [Chat API](https://docs.vectara.com/docs/console-ui/vectara-chat-overview) to query the corpus and present the results to you, answering your question.\n\n"
)
st.markdown("---")
st.markdown(f"<center> <h2> Vectara AI Assistant: {cfg.title} </h2> </center>", unsafe_allow_html=True)
if "messages" not in st.session_state.keys():
reset()
# Display chat messages
for message in st.session_state.messages:
with st.chat_message(message["role"]):
st.write(message["content"])
example_container = st.empty()
with example_container:
if show_example_questions():
example_container.empty()
st.rerun()
# select prompt from example question or user provided input
if st.session_state.ex_prompt:
prompt = st.session_state.ex_prompt
else:
prompt = st.chat_input()
if prompt:
st.session_state.messages.append({"role": "user", "content": prompt})
with st.chat_message("user"):
st.write(prompt)
st.session_state.ex_prompt = None
# Generate a new response if last message is not from assistant
if st.session_state.messages[-1]["role"] != "assistant":
with st.chat_message("assistant"):
if cfg.streaming:
stream = generate_streaming_response(prompt)
response = st.write_stream(stream)
else:
with st.spinner("Thinking..."):
response = generate_response(prompt)
st.write(response)
response = escape_dollars_outside_latex(response)
message = {"role": "assistant", "content": response}
st.session_state.messages.append(message)
# Send query and response to Amplitude Analytics
send_amplitude_data(
user_query=st.session_state.messages[-2]["content"],
chat_response=st.session_state.messages[-1]["content"],
demo_name=cfg["title"],
language=st.session_state.language
)
st.rerun()
if (st.session_state.messages[-1]["role"] == "assistant") & (st.session_state.messages[-1]["content"] != "How may I help you?"):
streamlit_feedback(feedback_type="thumbs", on_submit = thumbs_feedback, key = st.session_state.feedback_key,
kwargs = {"user_query": st.session_state.messages[-2]["content"],
"chat_response": st.session_state.messages[-1]["content"],
"demo_name": cfg["title"],
"response_language": st.session_state.language})
if __name__ == "__main__":
launch_bot()