Upload app.py

app.py

@@ -0,0 +1,106 @@
+import asyncio
+import os
+
+try:
+    asyncio.get_running_loop()
+except RuntimeError:
+    loop = asyncio.new_event_loop()
+    asyncio.set_event_loop(loop)
+
+import streamlit as st
+from langchain_community.vectorstores import FAISS
+from langchain_community.document_loaders import PyMuPDFLoader, Docx2txtLoader
+from langchain_huggingface import HuggingFaceEmbeddings
+from langchain.text_splitter import RecursiveCharacterTextSplitter
+from langchain_ollama import OllamaLLM
+from langchain.docstore.document import Document
+
+os.environ["STREAMLIT_WATCHER_TYPE"] = "none"
+
+# 🧠 Cache Ollama model
+@st.cache_resource
+def load_llm():
+    return OllamaLLM(model="phi3:mini", 
+                temperature=0.2,
+                top_p=0.95,
+                num_ctx=2048,
+                max_tokens=200)
+
+# 🧠 Cache embedder
+@st.cache_resource
+def load_embedder():
+    return HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2")
+
+llm = load_llm()
+embedder = load_embedder()
+
+# Sidebar Upload
+st.sidebar.title("📄 Upload Terms & Conditions")
+input_mode = st.sidebar.radio("Choose Input Method", ["📋 Paste Text", "📁 Upload File"])
+
+uploaded_text = ""
+if input_mode == "📋 Paste Text":
+    uploaded_text = st.sidebar.text_area("Paste your T&C text here")
+elif input_mode == "📁 Upload File":
+    uploaded_file = st.sidebar.file_uploader("Upload a .txt, .pdf, or .docx file", type=["txt", "pdf", "docx"])
+    if uploaded_file:
+        if uploaded_file.type == "text/plain":
+            uploaded_text = uploaded_file.read().decode("utf-8")
+        elif uploaded_file.type == "application/pdf":
+            with open("temp.pdf", "wb") as f:
+                f.write(uploaded_file.read())
+            docs = PyMuPDFLoader("temp.pdf").load()
+            uploaded_text = "\n".join([d.page_content for d in docs])
+        elif uploaded_file.type == "application/vnd.openxmlformats-officedocument.wordprocessingml.document":
+            with open("temp.docx", "wb") as f:
+                f.write(uploaded_file.read())
+            docs = Docx2txtLoader("temp.docx").load()
+            uploaded_text = "\n".join([d.page_content for d in docs])
+
+# ✅ Vectorstore setup
+if uploaded_text:
+    st.success("✅ Document loaded and processed!")
+
+    if "db" not in st.session_state:
+        text_splitter = RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=50)
+        documents = text_splitter.create_documents([uploaded_text])
+        db = FAISS.from_documents(documents, embedder)
+        st.session_state.db = db
+        st.session_state.chat_history = []
+
+# 💬 Chat section
+if "db" in st.session_state:
+    st.title("🧾 Legal Assistant Chat")
+    st.markdown("Ask anything about the uploaded document.")
+
+    user_input = st.chat_input("Type your question here...")
+
+    if user_input:
+        with st.spinner("🤖 Thinking..."):
+            retriever = st.session_state.db.as_retriever(search_kwargs={"k": 3})
+            docs = retriever.invoke(user_input)
+            context = "\n\n".join([doc.page_content for doc in docs])
+
+            prompt = f"""You are a helpful legal assistant.
+Based on the following contract, answer the user's question, This application built by Vighnesh.
+
+Context:
+{context}
+
+Question:
+{user_input}
+
+Answer:"""
+
+            answer = llm.invoke(prompt)
+
+            # Save chat history
+            st.session_state.chat_history.append(("user", user_input))
+            st.session_state.chat_history.append(("assistant", answer))
+
+    # Display chat history
+    for role, message in st.session_state.chat_history:
+        if role == "user":
+            st.chat_message("user").write(message)
+        else:
+            st.chat_message("assistant").write(message)