Create app.py

app.py

@@ -0,0 +1,92 @@
+import streamlit as st
+import os
+
+# --- LANGCHAIN IMPORTS ---
+from langchain_community.document_loaders import PyPDFLoader
+from langchain_experimental.text_splitter import SemanticChunker
+from langchain_huggingface import HuggingFaceEmbeddings
+from langchain_community.vectorstores import FAISS
+from langchain.chains import ConversationalRetrievalChain
+from langchain_community.chat_models import ChatOllama
+from langchain.memory import ConversationBufferMemory
+
+# 1) SET UP PAGE
+st.title("💬 Conversational Chat - Data Management & Personal Data Protection")
+local_file = "Policies001.pdf"
+st.write(f"Loading local file: {local_file}")
+index_folder = "faiss_index"
+
+# 2) LOAD OR BUILD VECTORSTORE
+embeddings = HuggingFaceEmbeddings(
+    model_name="CAMeL-Lab/bert-base-arabic-camelbert-mix",
+    model_kwargs={"trust_remote_code": True}
+)
+
+if os.path.exists(index_folder):
+    st.write("Loading existing FAISS index from disk...")
+    vectorstore = FAISS.load_local(index_folder, embeddings, allow_dangerous_deserialization=True)
+else:
+    st.write("Building a new FAISS index...")
+    loader = PyPDFLoader(local_file)
+    documents = loader.load()
+
+    text_splitter = SemanticChunker(
+        embeddings=embeddings,
+        breakpoint_threshold_type='percentile',
+        breakpoint_threshold_amount=90
+    )
+    chunked_docs = text_splitter.split_documents(documents)
+    st.write(f"Document split into {len(chunked_docs)} chunks.")
+
+    vectorstore = FAISS.from_documents(chunked_docs, embeddings)
+    vectorstore.save_local(index_folder)
+
+# 3) CREATE RETRIEVER
+retriever = vectorstore.as_retriever(search_type="similarity", search_kwargs={"k": 5})
+
+# 4) SET UP LLM + CONVERSATIONAL RETRIEVAL CHAIN
+llm = ChatOllama(model="command-r7b-arabic")
+
+# IMPORTANT: Memory object to store conversation
+memory = ConversationBufferMemory(
+    memory_key="chat_history",    # key used internally by the chain
+    return_messages=True          # ensures we get the entire message history
+)
+
+qa_chain = ConversationalRetrievalChain.from_llm(
+    llm=llm,
+    retriever=retriever,
+    memory=memory,
+    verbose=True  # optional: prints chain's internal logs in the console
+)
+
+# 5) MANAGE SESSION STATE FOR UI CHAT
+if "messages" not in st.session_state:
+    st.session_state["messages"] = [
+        {"role": "assistant", "content": "👋 Hello! Ask me anything about Data Management & Personal Data Protection!"}
+    ]
+
+# Display existing messages in chat format
+for msg in st.session_state["messages"]:
+    with st.chat_message(msg["role"]):
+        st.markdown(msg["content"])
+
+# 6) CHAT INPUT
+user_input = st.chat_input("Type your question...")
+
+# 7) PROCESS NEW USER MESSAGE
+if user_input:
+    # a) Display user message in UI
+    st.session_state["messages"].append({"role": "user", "content": user_input})
+    with st.chat_message("user"):
+        st.markdown(user_input)
+
+    # b) Run chain with conversation memory
+    #    We pass user_input as the "question" to the chain.
+    response_dict = qa_chain({"question": user_input})
+    answer = response_dict["answer"]
+
+    # c) Display assistant response
+    st.session_state["messages"].append({"role": "assistant", "content": answer})
+    with st.chat_message("assistant"):
+        st.markdown(answer)