Update app.py
Browse files
app.py
CHANGED
@@ -0,0 +1,126 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
import streamlit as st
|
3 |
+
from dotenv import load_dotenv
|
4 |
+
from langchain.chains import create_history_aware_retriever, create_retrieval_chain
|
5 |
+
from langchain.chains.combine_documents import create_stuff_documents_chain
|
6 |
+
from langchain_community.vectorstores import Chroma
|
7 |
+
from langchain_core.messages import HumanMessage, SystemMessage
|
8 |
+
from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder
|
9 |
+
from langchain_openai import ChatOpenAI, OpenAIEmbeddings
|
10 |
+
import pandas as pd
|
11 |
+
from huggingface_hub import login
|
12 |
+
|
13 |
+
# Load environment variables
|
14 |
+
openai_api_key = os.getenv("OPENAI_API_KEY")
|
15 |
+
|
16 |
+
hf_key = os.getenv("huggingface")
|
17 |
+
login(hf_key)
|
18 |
+
|
19 |
+
# Load product data from CSV
|
20 |
+
product_data_path = "./db/catalog_chatbot_2024-07-08.csv"
|
21 |
+
df = pd.read_csv(product_data_path, encoding='ISO-8859-1', sep=';')
|
22 |
+
|
23 |
+
# Define the embedding model
|
24 |
+
embeddings = OpenAIEmbeddings(model="text-embedding-3-small")
|
25 |
+
|
26 |
+
# Create a persistent directory for ChromaDB
|
27 |
+
persistent_directory = os.path.join("./","db", "chroma_open_ai")
|
28 |
+
|
29 |
+
# Check if the vector store already exists
|
30 |
+
if not os.path.exists(os.path.join(persistent_directory, 'chroma.sqlite3')):
|
31 |
+
db = Chroma(persist_directory=persistent_directory, embedding_function=embeddings)
|
32 |
+
for index, row in df.iterrows():
|
33 |
+
product_info = (
|
34 |
+
f"Nom du produit: {row['Nom du produit']} - "
|
35 |
+
f"Catégorie: {row['Catégorie par défaut']} - "
|
36 |
+
f"Caractéristiques: {row['Caractéristiques']} - "
|
37 |
+
f"Prix: {row['Prix de vente TTC']} - "
|
38 |
+
f"Description: {row['Description sans HTML']}"
|
39 |
+
)
|
40 |
+
metadata = {
|
41 |
+
"reference": row['Référence interne'],
|
42 |
+
"name": row['Nom du produit'],
|
43 |
+
"price": row['Prix de vente TTC'],
|
44 |
+
"product_url": row['URL Produit']
|
45 |
+
}
|
46 |
+
db.add_texts(texts=[product_info], metadatas=[metadata])
|
47 |
+
db.persist()
|
48 |
+
else:
|
49 |
+
db = Chroma(persist_directory=persistent_directory, embedding_function=embeddings)
|
50 |
+
|
51 |
+
# Create a retriever
|
52 |
+
retriever = db.as_retriever(search_type="similarity", search_kwargs={"k": 10})
|
53 |
+
|
54 |
+
# Create a ChatOpenAI model
|
55 |
+
llm = ChatOpenAI(model="gpt-4o")
|
56 |
+
|
57 |
+
# Function to format the products
|
58 |
+
def format_retrieved_products(retrieved_docs):
|
59 |
+
recommendations = []
|
60 |
+
seen_products = set()
|
61 |
+
for doc in retrieved_docs:
|
62 |
+
metadata = doc.metadata
|
63 |
+
product_name = metadata.get("name", "Produit inconnu")
|
64 |
+
price = metadata.get("price", "Prix non disponible")
|
65 |
+
product_url = metadata.get("product_url", "#")
|
66 |
+
|
67 |
+
if product_name not in seen_products:
|
68 |
+
recommendation = f"**{product_name}** - {price} €\n[Voir produit]({product_url})"
|
69 |
+
recommendations.append(recommendation)
|
70 |
+
seen_products.add(product_name)
|
71 |
+
|
72 |
+
return "\n".join(recommendations)
|
73 |
+
|
74 |
+
# Update the system prompt
|
75 |
+
qa_system_prompt = (
|
76 |
+
"You are a sales assistant helping customers purchase wine. "
|
77 |
+
"Use the retrieved context from the Chroma DB to answer the question. "
|
78 |
+
"Recommend 3 different items and provide the URLs of the 3 products from Calais Vins."
|
79 |
+
)
|
80 |
+
|
81 |
+
qa_prompt = ChatPromptTemplate.from_messages(
|
82 |
+
[
|
83 |
+
("system", qa_system_prompt),
|
84 |
+
MessagesPlaceholder("chat_history"),
|
85 |
+
("human", "{input}"),
|
86 |
+
("system", "{context}")
|
87 |
+
]
|
88 |
+
)
|
89 |
+
|
90 |
+
question_answer_chain = create_stuff_documents_chain(llm, qa_prompt)
|
91 |
+
|
92 |
+
# Define a retrieval chain
|
93 |
+
def create_custom_retrieval_chain(retriever, llm_chain):
|
94 |
+
def invoke(inputs):
|
95 |
+
query = inputs["input"]
|
96 |
+
retrieved_docs = retriever.get_relevant_documents(query)
|
97 |
+
formatted_response = format_retrieved_products(retrieved_docs)
|
98 |
+
return {"answer": formatted_response}
|
99 |
+
|
100 |
+
return invoke
|
101 |
+
|
102 |
+
rag_chain = create_custom_retrieval_chain(retriever, question_answer_chain)
|
103 |
+
|
104 |
+
# Streamlit App Interface
|
105 |
+
def run_streamlit_chatbot():
|
106 |
+
st.title("Wine Sales Assistant")
|
107 |
+
|
108 |
+
chat_history = []
|
109 |
+
|
110 |
+
# User input area
|
111 |
+
user_query = st.text_input("Posez une question au chatbot (e.g., je recherche un vin blanc fruité):")
|
112 |
+
|
113 |
+
if user_query:
|
114 |
+
result = rag_chain({"input": user_query, "chat_history": chat_history})
|
115 |
+
|
116 |
+
# Display chatbot response
|
117 |
+
st.write("### Chatbot's Recommendations:")
|
118 |
+
st.write(result["answer"])
|
119 |
+
|
120 |
+
# Display recommendations in a pop-up like fashion
|
121 |
+
with st.expander("Voir les recommandations"):
|
122 |
+
st.write(result["answer"])
|
123 |
+
|
124 |
+
# Main function to run the Streamlit app
|
125 |
+
if __name__ == "__main__":
|
126 |
+
run_streamlit_chatbot()
|