app.py

import os
import streamlit as st
from dotenv import load_dotenv
from langchain.chains import create_history_aware_retriever, create_retrieval_chain
from langchain.chains.combine_documents import create_stuff_documents_chain
from langchain_community.vectorstores import Chroma
from langchain_core.messages import HumanMessage, SystemMessage
from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder
from langchain_openai import ChatOpenAI, OpenAIEmbeddings
import pandas as pd
from huggingface_hub import login

# Load environment variables
openai_api_key = os.getenv("OPENAI_API_KEY")

hf_key = os.getenv("huggingface")
login(hf_key)

# Load product data from CSV
product_data_path = "./db/catalog_chatbot_2024-07-08.csv"
df = pd.read_csv(product_data_path, encoding='ISO-8859-1', sep=';')

# Define the embedding model
embeddings = OpenAIEmbeddings(model="text-embedding-3-small")

# Create a persistent directory for ChromaDB
persistent_directory = os.path.join("./","db", "chroma_open_ai")

# Check if the vector store already exists
if not os.path.exists(os.path.join(persistent_directory, 'chroma.sqlite3')):
    db = Chroma(persist_directory=persistent_directory, embedding_function=embeddings)
    for index, row in df.iterrows():
        product_info = (
            f"Nom du produit: {row['Nom du produit']} - "
            f"Catégorie: {row['Catégorie par défaut']} - "
            f"Caractéristiques: {row['Caractéristiques']} - "
            f"Prix: {row['Prix de vente TTC']} - "
            f"Description: {row['Description sans HTML']}"
        )
        metadata = {
            "reference": row['Référence interne'],
            "name": row['Nom du produit'],
            "price": row['Prix de vente TTC'],
            "product_url": row['URL Produit']
        }
        db.add_texts(texts=[product_info], metadatas=[metadata])
    db.persist()
else:
    db = Chroma(persist_directory=persistent_directory, embedding_function=embeddings)

# Create a retriever
retriever = db.as_retriever(search_type="similarity", search_kwargs={"k": 10})

# Create a ChatOpenAI model
llm = ChatOpenAI(model="gpt-4o")

# Function to format the products
def format_retrieved_products(retrieved_docs):
    recommendations = []
    seen_products = set()
    for doc in retrieved_docs:
        metadata = doc.metadata
        product_name = metadata.get("name", "Produit inconnu")
        price = metadata.get("price", "Prix non disponible")
        product_url = metadata.get("product_url", "#")

        if product_name not in seen_products:
            recommendation = f"**{product_name}** - {price} €\n[Voir produit]({product_url})"
            recommendations.append(recommendation)
            seen_products.add(product_name)
    
    return "\n".join(recommendations)

# Update the system prompt
qa_system_prompt = (
    "You are a sales assistant helping customers purchase wine. "
    "Use the retrieved context from the Chroma DB to answer the question. "
    "Recommend 3 different items and provide the URLs of the 3 products from Calais Vins."
)

qa_prompt = ChatPromptTemplate.from_messages(
    [
        ("system", qa_system_prompt),
        MessagesPlaceholder("chat_history"),
        ("human", "{input}"),
        ("system", "{context}")
    ]
)

question_answer_chain = create_stuff_documents_chain(llm, qa_prompt)

# Define a retrieval chain
def create_custom_retrieval_chain(retriever, llm_chain):
    def invoke(inputs):
        query = inputs["input"]
        retrieved_docs = retriever.get_relevant_documents(query)
        formatted_response = format_retrieved_products(retrieved_docs)
        return {"answer": formatted_response}
    
    return invoke

rag_chain = create_custom_retrieval_chain(retriever, question_answer_chain)

# Streamlit App Interface
def run_streamlit_chatbot():
    st.title("Wine Sales Assistant")
    
    chat_history = []
    
    # User input area
    user_query = st.text_input("Posez une question au chatbot (e.g., je recherche un vin blanc fruité):")

    if user_query:
        result = rag_chain({"input": user_query, "chat_history": chat_history})
        
        # Display chatbot response
        st.write("### Chatbot's Recommendations:")
        st.write(result["answer"])

        # Display recommendations in a pop-up like fashion
        with st.expander("Voir les recommandations"):
            st.write(result["answer"])

# Main function to run the Streamlit app
if __name__ == "__main__":
    run_streamlit_chatbot()