newchatbot.py

import streamlit as st
import os
import chatFromLangchain
from chatFromLangchain import LLm
from dotenv import load_dotenv
from langchain_core.messages import HumanMessage
from langchain_core.messages import AIMessage
import re
import actions
import pandas as pd
 
# Load environment variables from the .env file
load_dotenv()
token = os.getenv('AZURE_OPENAI_KEY')
llm_model = 'gpt-3.5-turbo-1106'
temperature = 1.0
 
# Add the logo beside the title
st.image("logo.png")
st.title("Metaminds")
 
# Add custom CSS for chat bubbles with alignment to the right and left
st.markdown("""
    <style>
    .chat-container {
        display: flex;
        flex-direction: column;
        gap: 10px; /* Space between messages */
        padding: 10px;
    }
    .chat-bubble {
        max-width: 70%;
        padding: 10px;
        border-radius: 15px;
        margin: 5px 0; /* Space around each bubble */
        line-height: 1.5;
        word-wrap: break-word; /* Ensure long words wrap */
    }
    .human-message-container {
        display: flex;
        justify-content: flex-start; /* Align human messages to the left */
    }
    .ai-message-container {
        display: flex;
        justify-content: flex-end; /* Align AI messages to the right */
    }
    .human-message {
        background-color: #DCF8C6;
        color: black;
    }
    .ai-message {
        background-color: #E3E3E3;
        color: black;
    }
    </style>
    """, unsafe_allow_html=True)
 
# Dev Mode Checkbox
dev_mode = st.checkbox("Dev Mode")
 
# Setting the LLM
if dev_mode:
    with st.expander("Setting the LLM"):
        st.markdown("This page is used to have a chat with the uploaded documents")
        with st.form("setting"):
            row_1 = st.columns(3)
            with row_1[0]:
                token = st.text_input("Open AI Key", type="password")
 
            with row_1[1]:
                llm_model = st.text_input("LLM model", value="openai-gpt-4o")
 
            with row_1[2]:
                temperature = st.number_input("Temperature", value=1.0, step=0.1)
 
            create_chatbot = st.form_submit_button("Create chatbot")
    
    with st.expander("Chat History LLm"):
        chathistplace = st.empty()
 
# Prepare the LLM model
if "conversation" not in st.session_state:
    st.session_state.conversation = None
 
if token:
    st.session_state.conversation = LLm()
 
# Chat history
if "history" not in st.session_state:
    st.session_state.history = []
 
# Chat messages 
if "messages" not in st.session_state:
    st.session_state.messages = [{'role': 'system', 'content': '''You are a Helpful chatbot also a truthful one who will be chatting with user Ragrding some connections to the database .
You have access to these tools
tableFindinRedshift, this tools find if the table is present in the Redshift database, this tools takes only one input 'tableName': str, and returns either None or a list representing multiple tables present in table or a list of tuples where each tuple representing column name and the column type
tableFindinImpala, this tools find if the table is present in the Impala database, this tools takes only one input 'tableName': str, and returns either None or a list representing multiple tables present in table or a list of tuples where each tuple representing column name and the column type
 
if a user have not mentioned the Database connection in his connection confirm it from the user.
 
Here is as example 
user: Hi
as: Hi I am an assistant who can help you with details of any table in Datalake.
user: Ok tell about the Mara table in Redshift
assistant: ```{'action': 'tableFindinRedshift', 'input': 'Mara'}```
function: ['Mara1', 'Mara2', 'Maraxyz']
assistant: There Are multiple table present in the Reshift similar to Mara please Provide which one to choose 
user: Mara1
assistant:  ```{'action': 'tableFindinRedshift', 'input': 'Mara1'}```
function: [('client_id', int), ('purchase', int)]
assistant: Mara1 has 2 columns client_id: int, purchase: int
user: Thank you now write a Query which will give me all the clients which has purchase value more than 500
assistant: Certainly ```SELECT client_id FROM Mara1 WHERE purchase > 500```
 
Refuse to answer any query from user where it deviates from the above mentioned capabilites
Example:
user: Tell me about the weather.
assistant: Hi I am still MetaMinds 1.0 maybe I can answer your questions better when I am MetaMinds 2.0 I can only answer related to Datalakes.
'''}]
 
 
def getAction(text):
    # Regular expression to extract the content within the backticks
    pattern = r"```(.*?)```"
 
    # Search for the pattern
    match = re.search(pattern, text)
 
    # Extract the matched content if found
    if match:
        extracted_content = match.group(1)
        return eval(extracted_content)
    else:
        return None
 
# Display chats - OLD
st.markdown('<div class="chat-container">', unsafe_allow_html=True)
for message in st.session_state.history:
    if isinstance(message, pd.DataFrame):
        st.dataframe(message)
        continue
    message = eval(message.json(), {'false': False, 'true': True})
    role = message["type"]
    content = message["content"]
    if role == "human":
        st.markdown(f'<div class="human-message-container"><div class="chat-bubble human-message">{content}</div></div>', unsafe_allow_html=True)
    else:
        st.markdown(f'<div class="ai-message-container"><div class="chat-bubble ai-message">{content}</div></div>', unsafe_allow_html=True)
st.markdown('</div>', unsafe_allow_html=True)
 
if dev_mode:
    chathistplace.json(st.session_state.messages)
 
 
# Ask a question
if question := st.chat_input("Ask a question"):
    # Append user question to history
    st.session_state.history.append(
        HumanMessage(content=question)
    )
    # Add user question
    st.markdown(f'<div class="human-message-container"><div class="chat-bubble human-message">{question}</div></div>', unsafe_allow_html=True)
 
    # Answer the question
    answer = st.session_state.conversation.generate_answer(question)
 
    # When some action is getting performed
    if getAction(answer) is not None:
        action = getAction(answer)
        print('\n\n\n\n------------------------ Performing Action -----------------------')
        print(action, '\n\n\n')
        
        if action['action'] == 'tableFindinRedshift':
            action_statement = f"Searcing the Table {action['input'].lower()} in Redshift Hang on a sec "
            st.markdown(f'<div class="ai-message-container"><div class="chat-bubble ai-message">{action_statement}</div></div>', unsafe_allow_html=True)
            st.session_state.history.append(
                AIMessage(content=action_statement)
            )
            answer = actions.tableFindinRedshift(action['input'].lower())
        
            if isinstance(answer, pd.DataFrame) and 'schema' in answer.columns:
                table_list = []
 
                for i, row in answer.iterrows():
                    table_list.append(f'{row["schema"]}.{row["table_name"]}')
                
 
                st.session_state.messages.append(
                    {
                        'role': 'assistant', 'content': str(table_list)
                    }
                )
                st.session_state.messages.append(
                    {
                        'role': 'assistant', 'content': f'These are all the Tables I could find in Redshift that has name {action["input"].lower()} in it please select among these'
                    }
                )
 
                st.markdown(f'''<div class="ai-message-container"><div class="chat-bubble ai-message">These are all the Tables I could find in Redshift that has name {action["input"].lower()} in it please select among these</div></div>''', unsafe_allow_html=True)
 
                st.session_state.history.append(
                    AIMessage(content=f'These are all the Tables I could find in Redshift that has name {action["input"].lower()} in it please select among these')
                ) 
 
            
            if isinstance(answer, pd.DataFrame) and 'datatype' in answer.columns:
                column_detail = []
 
                for i, row in answer.iterrows():
                    column_detail.append([row['columnname'], row['datatype']])
                
 
                st.session_state.messages.append(
                    {
                        'role': 'assistant', 'content': str(column_detail)
                    }
                )
 
                st.session_state.messages.append(
                    {
                        'role': 'assistant', 'content': f'Here is the column details of Table {action["input"].lower()} from Redshift'
                    }
                )
 
                st.markdown(f'''<div class="ai-message-container"><div class="chat-bubble ai-message">Here is the column details of Table {action["input"].lower()} from Redshift</div></div>''', unsafe_allow_html=True) 
 
                st.session_state.history.append(
                    AIMessage(content=f'Here is the column details of Table {action["input"].lower()} from Redshift')
                )
                
        if isinstance(answer, str):
            st.markdown(f'<div class="ai-message-container"><div class="chat-bubble ai-message">{answer}</div></div>', unsafe_allow_html=True) 
        if isinstance(answer, pd.DataFrame):
            st.dataframe(answer)
 
    else:
        st.markdown(f'<div class="ai-message-container"><div class="chat-bubble ai-message">{answer}</div></div>', unsafe_allow_html=True)
    
    # Append assistant answer to history
    if isinstance(answer, pd.DataFrame):
        st.session_state.history.append(
            answer
        )
    else:
        st.session_state.history.append(
            AIMessage(content=answer)
        )
 
    if dev_mode:
        chathistplace.json(st.session_state.messages)