main.py

# main.py
import os
import tempfile

import streamlit as st
from files import file_uploader, url_uploader
from question import chat_with_doc
from brain import brain
from langchain.embeddings import HuggingFaceInferenceAPIEmbeddings
from langchain.vectorstores import SupabaseVectorStore
from supabase import Client, create_client
from explorer import view_document
from stats import get_usage_today
from st_login_form import login_form

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

# 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"]

# Set the theme
st.set_page_config(
    page_title="meraKB",
    layout="wide",
    initial_sidebar_state="expanded",
)


st.title("🧠 meraKB - Your digital brain 🧠")
st.markdown("Store your knowledge in a vector store and chat with it.")
if self_hosted == "false":
    st.markdown('**📢 Note: In the public demo, access to functionality is restricted. You can only use the GPT-3.5-turbo model and upload files up to 1Mb. To use more models and upload larger files, consider self-hosting meraKB.**')

st.markdown("---\n\n")

st.session_state["overused"] = False
if self_hosted == "false":
    usage = get_usage_today(supabase)
    if usage > st.secrets.usage_limit:
        st.markdown(
            f"<span style='color:red'>You have used {usage} tokens today, which is more than your daily limit of {st.secrets.usage_limit} tokens. Please come back later or consider self-hosting.</span>", unsafe_allow_html=True)
        st.session_state["overused"] = True
    else:
        st.markdown(f"<span style='color:blue'>Usage today: {usage} tokens out of {st.secrets.usage_limit}</span>", unsafe_allow_html=True)
    st.write("---")

client = login_form()

if st.session_state["authenticated"]:
    if st.session_state["username"]:
        st.success(f"Welcome {st.session_state['username']}")
    else:
        st.session_state["username"] = 'guest'
        st.success("Welcome guest")

    # 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

    # Create a radio button for user to choose between adding knowledge or asking a question
    user_choice = st.radio(
        "Choose an action", ('Add Knowledge', 'Chat with your Brain', 'Forget', "Explore"))

    st.markdown("---\n\n")

    if user_choice == 'Add Knowledge':
        # Display chunk size and overlap selection only when adding knowledge
        st.sidebar.title("Configuration")
        st.sidebar.markdown(
            "Choose your chunk size and overlap for adding knowledge.")
        st.session_state['chunk_size'] = st.sidebar.slider(
            "Select Chunk Size", 100, 1000, st.session_state['chunk_size'], 50)
        st.session_state['chunk_overlap'] = st.sidebar.slider(
            "Select Chunk Overlap", 0, 100, st.session_state['chunk_overlap'], 10)
        
        # Create two columns for the file uploader and URL uploader
        col1, col2 = st.columns(2)
        
        with col1:
            file_uploader(supabase, vector_store)
        with col2:
            url_uploader(supabase, vector_store)
    elif user_choice == 'Chat with your Brain':
        # Display model and temperature selection only when asking questions
        st.sidebar.title("Configuration")
        st.sidebar.markdown(
            "Choose your model and temperature for asking questions.")
        if self_hosted != "false":
            st.session_state['model'] = st.sidebar.selectbox(
            "Select Model", models, index=(models).index(st.session_state['model']))
        else:
            st.sidebar.write("**Model**: gpt-3.5-turbo")
            st.sidebar.write("**Self Host to unlock more models such as claude-v1 and GPT4**")
            st.session_state['model'] = "gpt-3.5-turbo"
        st.session_state['temperature'] = st.sidebar.slider(
            "Select Temperature", 0.1, 1.0, st.session_state['temperature'], 0.1)
        if st.secrets.self_hosted != "false":
            st.session_state['max_tokens'] = st.sidebar.slider(
                "Select Max Tokens", 500, 4000, st.session_state['max_tokens'], 500)
        else:
            st.session_state['max_tokens'] = 500
        
        chat_with_doc(st.session_state['model'], vector_store, stats_db=supabase)
    elif user_choice == 'Forget':
        st.sidebar.title("Configuration")

        brain(supabase)
    elif user_choice == 'Explore':
        st.sidebar.title("Configuration")
        view_document(supabase)

    st.markdown("---\n\n")
else:
    st.error("Not authenticated")