streamlit_app.py

# streamlit_app.py
import streamlit as st
from src.chatbot import BioethicsChatbot
import time

st.set_page_config(page_title="Bioethics AI Assistant", page_icon="🧬", layout="wide")

st.title("🧬 Bioethics AI Assistant")
st.markdown("*Ask questions about medical ethics, informed consent, research ethics, and more*")

# CSS: fix input bar to bottom, style chat area and bubbles, and reserve vertical space
st.markdown(
    """
    <style>
      /* Chat-area wrapper */
      #chat-area {
        width: 100%;
        max-width: 800px;
        margin: 0;
      }

      /* User question */
      #user {
      top: 18px;
      z-index: 10;
      padding: 12px 16px;
      border-radius: 12px;
      background: var(--secondary-background-color);
      margin-bottom: 12px;
      will-change: transform;
      }

      /* Message bubbles */
      .msg {
      padding: 14px 16px;
      border-radius: 12px;
      margin: 10px 0;
      max-width: 65%;
      line-height: 1.45;
    }
    
    .msg.user {
      background: #101726;
      color: white;
      margin-left: auto;
    }
    
    .msg.assistant {
      background: #262730;
      color: white;
      margin-right: auto;
    }


      /* Fix the input form (id = input_form) at the bottom, centered */
      form#input_form {
      position: fixed;
      bottom: 18px;
      left: 50%;
      transform: translateX(-50%);
      width: 66.7%;
      z-index: 9999;
      background: transparent;
      }

      /* optional small visual tweak */
      form#input_form .stTextInput, form#input_form .stButton {
        margin: 0 6px;
      }
      
      form#input_form .stTextInput, form#input_form .stButton {
        margin: 0 6px;
      }

      /* Reserve space at the bottom equal to input form height */
      .block-container {
        padding-bottom: 120px;
      }

    </style>
    """,
    unsafe_allow_html=True,
)

# session state
if 'messages' not in st.session_state:
    # messages is a list of {"role": "user"|"assistant", "content": str}
    st.session_state.messages = []

if 'is_streaming' not in st.session_state:
    st.session_state.is_streaming = False

if 'show_sticky' not in st.session_state:
    st.session_state.show_sticky = False

# if a submission was made, store it as pending so next render can create placeholders nicely
if 'pending_question' not in st.session_state:
    st.session_state.pending_question = None

# cached chatbot resource
@st.cache_resource
def load_chatbot():
    return BioethicsChatbot("data/")

if 'bot' not in st.session_state:
    with st.spinner("🔄 Loading bioethics knowledge base..."):
        st.session_state.bot = load_chatbot()

#History
def build_history_pairs(messages, max_pairs=4):
    pairs = []
    i = 0
    while i < len(messages) - 1:
        if messages[i]['role'] == 'user' and messages[i+1]['role'] == 'assistant':
            user = messages[i]['content']
            assistant = messages[i+1]['content']
            if assistant is not None and assistant != "":
                pairs.append((user, assistant))
            i += 2
        else:
            i += 1
    return pairs[-max_pairs:]

# Layout columns
col_left, col1, col2 = st.columns([1, 4, 1])

with col1:
    st.markdown("### 💬 Conversation")

    # Input form (kept logically first so submit handling happens before rendering),
    # CSS will pin it visually to the bottom (form id = input_form).
    # Centered chat container for better visuals

    # render messages in strict chronological order (user -> assistant -> user -> assistant)
    # track the first in-flight assistant placeholder (content == "")
    response_placeholder = None
    inflight_index = None
    for idx, msg in enumerate(st.session_state.messages):
        if msg['role'] == 'user':
            # user bubble (dark)
            st.markdown(
                f"""
                <div style="display:flex; justify-content:center; width:100%;">
                  <div class="msg user">{msg['content']}</div>
                </div>
                """,
                unsafe_allow_html=True
            )

        else:
            # assistant bubble
            if msg['content'] == "":
                response_placeholder = st.empty()
                inflight_index = idx
                response_placeholder.markdown(
                    """
                    <div style="display:flex; justify-content:center; width:100%;">
                      <div id="assistant-inflight" class="msg assistant"></div>
                    </div>
                    """,
                    unsafe_allow_html=True
                )

            else:
                st.markdown(
                    f"""
                    <div style="display:flex; justify-content:center; width:100%;">
                      <div class="msg assistant">{msg['content']}</div>
                    </div>
                    """,
                    unsafe_allow_html=True
                )

    # If there's an inflight placeholder and we're not already streaming, start streaming now
    if response_placeholder is not None and not st.session_state.get("is_streaming", False):
        st.session_state.is_streaming = True
        # build history pairs from only completed assistant responses (exclude the in-flight assistant)
        history_pairs = build_history_pairs(st.session_state.messages[:inflight_index])
        st.session_state.bot.stream_handler.current_text = ""
        st.session_state.bot.stream_handler.placeholder = response_placeholder

        # perform the streaming call (this will update response_placeholder via the callback)
        try:
            user_question = st.session_state.messages[inflight_index - 1]['content']
            answer = st.session_state.bot.ask(user_question, history_pairs=history_pairs)
        except Exception as e:
            answer = f"❌ Error while generating response: {e}"

        # finalize UI and persist final assistant text
        try:
            # put final content into the same placeholder (removes streaming cursor)
            response_placeholder.markdown(
                f"<div style='background:#262730;color:#fff;padding:14px;border-radius:12px;margin:10px 0;max-width:85%;'>"
                f"{answer}</div>",
                unsafe_allow_html=True
            )
        except Exception:
            pass

        # save final answer back into session_state messages
        st.session_state.messages[inflight_index]['content'] = answer
        st.session_state.is_streaming = False

    question = st.chat_input("Your question:")

    if question:
        if st.session_state.get("is_streaming", False):
            st.warning("Please wait for the current response to finish.")
        elif st.session_state.get('query_count', 0) < 30:
            st.session_state.messages.append({"role": "user", "content": question})
            st.session_state.messages.append({"role": "assistant", "content": ""})
            st.session_state.query_count = st.session_state.get('query_count', 0) + 1
            st.rerun()  # Add this line
        else:
            st.error("📈 Demo limit reached for today. This prevents API abuse.")

with col2:
    if 'query_count' not in st.session_state:
        st.session_state.query_count = 0
    st.metric("Queries used in your session", f"{st.session_state.query_count}/30")
    st.markdown("---")
    with st.expander("📚 About the Sources"):
        st.markdown("""
        This assistant searches through open-access bioethics papers to find relevant information.

        **Search Process:**
        1. Your question is converted to embeddings
        2. Similar text chunks are found using FAISS vector search
        3. Only chunks with similarity score ≥ 0.65 are used for citations
        4. The language model synthesizes an answer from these sources
        """)

    st.markdown("**🔗 License**")
    st.markdown("- [Open Source Papers Used](https://huggingface.co/spaces/ciorant/bioethics-rag/blob/main/LICENSE_INFO.md)")

    st.markdown("**Tech Stack**")
    st.markdown("- Python & Streamlit")
    st.markdown("- OpenAI GPT-4o-mini")
    st.markdown("- FAISS Vector Search")
    st.markdown("- LangChain")

    st.markdown("**📊 Demo Stats**")
    if 'bot' in st.session_state and hasattr(st.session_state.bot, 'vector_store'):
        doc_count = len(st.session_state.bot.vector_store.documents)
        st.markdown(f"- {doc_count} text chunks indexed")
        st.markdown(f"- Vector dimension: {st.session_state.bot.vector_store.dimension}")
    st.markdown(f"- Queries today: {st.session_state.get('query_count', 0)}")

with col_left:
    st.markdown("### 💡 Sample Questions")
    sample_questions = [
        "What are the key principles of informed consent?",
        "What is the moral side of genomic testing?",
        "How should we approach clinical trial ethics?",
        "Should assistance in dying be legal?"
    ]

    for q in sample_questions:
        if st.button(q, key=f"sample_{hash(q)}", use_container_width=True):
            if not st.session_state.get("is_streaming", False) and st.session_state.get('query_count', 0) < 30:
                st.session_state.messages.append({"role": "user", "content": q})
                st.session_state.messages.append({"role": "assistant", "content": ""})
                st.session_state.query_count += 1
                st.rerun()