app.py

import streamlit as st
from time import sleep
from retriever import load_vector_store
from langgraph_graph import generate_answer

st.set_page_config("MedMCQA Chatbot", page_icon="🩺")

# 🌗 Sidebar
with st.sidebar:
    st.title("🩺 MedMCQA Chatbot")
    theme_mode = st.radio("🌓 Theme", ["Light", "Dark"], horizontal=True)

# 🌓 Theme
if theme_mode == "Dark":
    st.markdown("""
        <style>
        body, .stApp { background-color: #1e1e1e !important; color: #eee !important; }
        .stTextInput input { background-color: #333 !important; color: #eee !important; }
        input::placeholder { color: #bbb !important; }
        .stButton>button { background-color: #444 !important; color: #eee !important; }
        </style>
    """, unsafe_allow_html=True)
else:
    st.markdown("""
        <style>
        body, .stApp { background-color: #ffffff !important; color: #111 !important; }
        .stTextInput input { background-color: #f0f0f0 !important; color: #111 !important; }
        input::placeholder { color: #444 !important; }
        .stButton>button { background-color: #e0e0e0 !important; color: #111 !important; }
        </style>
    """, unsafe_allow_html=True)

st.header("🩺 MedMCQA Chatbot")
st.caption("Ask a medical question and get answers from the MedMCQA dataset only. If not found, it will respond gracefully.")

# ✅ Cache FAISS loading to avoid blocking
@st.cache_resource(show_spinner="Loading FAISS Index... Please wait ⏳")
def get_vector_store():
    return load_vector_store()

# Load vector store in the background
db = get_vector_store()

# ✏️ Query
query = st.text_input("🔍 Enter your medical question:",
                      placeholder="e.g., What is the mechanism of Aspirin?")

# 🚀 Answer generation
if query:
    results = db.similarity_search(query, k=3)
    context = "\n\n".join([doc.page_content for doc in results])

    with st.spinner("🧠 Generating answer..."):
        response = generate_answer(query, context)

    # Animated answer
    st.markdown("<h4>🧠 Answer:</h4>", unsafe_allow_html=True)
    answer_placeholder = st.empty()
    final_text = ""
    for char in response:
        final_text += char
        answer_placeholder.markdown(final_text)
        sleep(0.01)

    with st.expander("🔎 Top Matches"):
        for i, doc in enumerate(results, 1):
            st.markdown(f"**Result {i}:**\n\n{doc.page_content}")

# 📬 Sidebar Contact
with st.sidebar:
    st.markdown("---")
    st.markdown("### 📬 Contact")
    st.markdown("[📧 Email](mailto:[email protected])")
    st.markdown("[🔗 LinkedIn](https://linkedin.com/in/sankethhonavar)")
    st.markdown("[💻 GitHub](https://github.com/SankethHonavar)")