safety-copilot / main.py
Asankhaya Sharma
add stats
cae23e1
raw
history blame
4.4 kB
# main.py
import os
import tempfile
import streamlit as st
from question import chat_with_doc
from langchain.embeddings import HuggingFaceInferenceAPIEmbeddings
from langchain.vectorstores import SupabaseVectorStore
from supabase import Client, create_client
from stats import add_usage
from langchain.llms import HuggingFaceEndpoint
from langchain.chains import ConversationalRetrievalChain
from langchain.memory import ConversationBufferMemory
supabase_url = st.secrets.SUPABASE_URL
supabase_key = st.secrets.SUPABASE_KEY
openai_api_key = st.secrets.openai_api_key
anthropic_api_key = st.secrets.anthropic_api_key
hf_api_key = st.secrets.hf_api_key
supabase: Client = create_client(supabase_url, supabase_key)
self_hosted = st.secrets.self_hosted
username = st.secrets.username
# embeddings = OpenAIEmbeddings(openai_api_key=openai_api_key)
embeddings = HuggingFaceInferenceAPIEmbeddings(
api_key=hf_api_key,
model_name="BAAI/bge-large-en-v1.5"
)
vector_store = SupabaseVectorStore(supabase, embeddings, query_name='match_documents', table_name="documents")
models = ["meta-llama/Llama-2-70b-chat-hf", "mistralai/Mixtral-8x7B-Instruct-v0.1"]
if openai_api_key:
models += ["gpt-3.5-turbo", "gpt-4"]
if anthropic_api_key:
models += ["claude-v1", "claude-v1.3",
"claude-instant-v1-100k", "claude-instant-v1.1-100k"]
if 'question' in st.query_params:
query = st.query_params['question']
model = "meta-llama/Llama-2-70b-chat-hf"
temp = 0.1
max_tokens = 500
add_usage(supabase, "api", "prompt" + query, {"model": model, "temperature": temp})
# print(st.session_state['max_tokens'])
endpoint_url = ("https://api-inference.huggingface.co/models/"+ model)
model_kwargs = {"temperature" : temp,
"max_new_tokens" : max_tokens,
"return_full_text" : False}
hf = HuggingFaceEndpoint(
endpoint_url=endpoint_url,
task="text-generation",
huggingfacehub_api_token=hf_api_key,
model_kwargs=model_kwargs
)
memory = ConversationBufferMemory(memory_key="chat_history", input_key='question', output_key='answer', return_messages=True)
qa = ConversationalRetrievalChain.from_llm(hf, retriever=vector_store.as_retriever(search_kwargs={"score_threshold": 0.8, "k": 4,"filter": {"user": username}}), memory=memory, return_source_documents=True)
model_response = qa({"question": query})
# print( model_response["answer"])
sources = model_response["source_documents"]
# print(sources)
if len(sources) > 0:
json = {"response": model_response["answer"]}
st.code(json, language="json")
else:
json = {"response": "I am sorry, I do not have enough information to provide an answer. If there is a public source of data that you would like to add, please email [email protected]."}
st.code(json, language="json")
memory.clear()
else:
# Set the theme
st.set_page_config(
page_title="Securade.ai - Safety Copilot",
page_icon="https://securade.ai/favicon.ico",
layout="centered",
initial_sidebar_state="collapsed",
menu_items={
"About": "# Securade.ai Safety Copilot v0.1\n [https://securade.ai](https://securade.ai)",
"Get Help" : "https://securade.ai",
"Report a Bug": "mailto:[email protected]"
}
)
st.title("👷‍♂️ Safety Copilot 🦺")
st.markdown("Chat with your personal safety assistant about any health & safety related queries.")
st.markdown("Up-to-date with latest OSH regulations for Singapore, Indonesia, Malaysia & other parts of Asia.")
st.markdown("---\n\n")
# Initialize session state variables
if 'model' not in st.session_state:
st.session_state['model'] = "meta-llama/Llama-2-70b-chat-hf"
if 'temperature' not in st.session_state:
st.session_state['temperature'] = 0.1
if 'chunk_size' not in st.session_state:
st.session_state['chunk_size'] = 500
if 'chunk_overlap' not in st.session_state:
st.session_state['chunk_overlap'] = 0
if 'max_tokens' not in st.session_state:
st.session_state['max_tokens'] = 500
if 'username' not in st.session_state:
st.session_state['username'] = username
chat_with_doc(st.session_state['model'], vector_store, stats_db=supabase)
st.markdown("---\n\n")