app.py

import streamlit as st
from openai import OpenAI
from app_config import *
from app_access_db import *


# model = "gpt-3.5-turbo"
# model = "gpt-4-turbo"
model = "gpt-4o"
gpt_base_url = None
# model = "lmstudio-community/Meta-Llama-3-8B-Instruct-GGUF"
# gpt_base_url = "http://localhost:1234/v1/"

# ------------------------------------------------------------------------------------------------
# SIDEBAR
# ------------------------------------------------------------------------------------------------
st.sidebar.title('OpenAI Business Chat')
st.sidebar.write('This chat bot is build with Tools and Function feature of OpenAI to be able to answer question regarding applications and performance of officers')
st.sidebar.markdown("""
    ### Having a sample database with a structure
        - application
            - app_number
            - amount
            - amount_paid
            - state. (APPROVED, REJECTED, PENDING_PAYMENT, PAID)
            - office_code [FK]
            - service_code [FK]
            - date_created
            - date_paid
            - date_processed
        - office
            - office_name
            - office_location_code [FK]
        - location
            - location_name
            - location_code
        - service
            - service_code
            - service_name
                    
    ### The chatbot  will provide answers from that database
        - The number of applications rejected is a location during the current month
        - The trend of applications in particular states, for a location
        - Any question you think relevant from this DB
""")

def onchange_openai_key():
    print(st.session_state.openai_key) 

openai_key = st.sidebar.text_input('OpenAI key', on_change=onchange_openai_key, key='openai_key')

def submit_openai_key(model=model):
    if(openai_key == None or openai_key==''):
        st.sidebar.write('Please provide the key before')
        return
    else:
        client = OpenAI(api_key=openai_key, base_url=gpt_base_url)
        model = model
        completion = client.chat.completions.create(
            model=model,  
            messages=[
                {"role": "system", "content": "You are an assistant giving simple and short answer for question of child. No questions, and no explanations"},
                {"role": "user", "content": "count from 0 to 10"}
            ],
            temperature=0
        )        
        st.sidebar.write(f'Simple count : {completion.choices[0].message.content}')

submit_key = st.sidebar.button(label='Submit', on_click=submit_openai_key)



# ------------------------------------------------------------------------------------------------
# CHAT
# ------------------------------------------------------------------------------------------------

st.title('OpenAI Business Chat')
st.write(f'Ask any question that can be answer by the LLM {model}.')


def askQuestion(model=model, question=''):
    if(openai_key == None or openai_key==''):
        print('Please provide the key before')
        return 'LLM API is not defined. Please provide the key before'
    else:
        client = OpenAI(api_key=openai_key, base_url=gpt_base_url)
        model = model
        completion = client.chat.completions.create(
            model=model,  
            messages=[
                {"role": "system", "content": f'{query_context}'},
                {"role": "user", "content": f'{question}'}
            ],
            temperature=0
        )        
        return completion.choices[0].message.content

class AssistantMessage:
    def __init__(self):
        self.sql : str
        self.response_data : DataFrame



def displayAssistantMessage( assistantMessage: AssistantMessage):
    with st.chat_message("assistant"):
        st.code(assistantMessage.sql, language='sql')
        st.code(assistantMessage.response_data, language='markdown')
        if assistantMessage.response_data.columns.size == 2:
            st.bar_chart(assistantMessage.response_data, x=assistantMessage.response_data.columns[0], y=assistantMessage.response_data.columns[1])
        if assistantMessage.response_data.columns.size == 1:
            st.metric(label=assistantMessage.response_data.columns[0], value=f'{assistantMessage.response_data.values[0][0]}')
            


# Initialize chat history
if "messages" not in st.session_state:
    st.session_state.messages = []

# Display chat messages from history on app rerun
for message in st.session_state.messages:
    if message["role"] == "user":
        with st.chat_message(message["role"]):
            st.markdown(message["content"])
    elif message["role"] == "assistant":
        displayAssistantMessage(message["content"])

# React to user input
if prompt := st.chat_input("What is up?"):
    with st.status('Running', expanded=True) as status:
        # Display user message in chat message container
        st.chat_message("user").markdown(prompt)
        # Add user message to chat history
        st.session_state.messages.append({"role": "user", "content": prompt})

        response = askQuestion(question=prompt)
        # st.code(response, language='sql')
        response_data = run_query(response.replace('```',''))
        # Display assistant response in chat message container
        assistanMsg = AssistantMessage()
        assistanMsg.sql = response.replace('```','')
        assistanMsg.response_data = response_data
        displayAssistantMessage(assistanMsg)
        # with st.chat_message("assistant"):
        #     st.code(response, language='sql')
        #     st.caption(response_data)
        #     st.bar_chart(response_data, x=response_data.columns[0], y=response_data.columns[1])
            
        # Add assistant response to chat history
        st.session_state.messages.append({"role": "assistant", "content": assistanMsg})
        status.update(label='Response of last question', state="complete", expanded=True)