app.py

#!/usr/bin/env python3
"""
Pathology RAG System - Streamlit Version
Query existing FAISS database
"""

import os
import sys
from pathlib import Path
from datetime import datetime

import streamlit as st

# Force CPU
os.environ["CUDA_VISIBLE_DEVICES"] = ""

# Add src folder
sys.path.append("src")

DB_PATH = "output/biomedbert_vector_db"

if not Path(DB_PATH).exists():
    st.error("Vector database not found. Upload output/biomedbert_vector_db.")
    st.stop()

# Import RAG pipeline & Updater
try:
    from retriever import CompleteRAGPipeline
    from document_processor import DynamicRAGUpdater
except ImportError as e:
    st.error(f"Import error: {e}")
    st.stop()


# -----------------------------
# Load Pipeline (cached)
# -----------------------------

@st.cache_resource
def load_pipeline():
    # Cache busted to pick up the new ask method return dictionary
    pipeline = CompleteRAGPipeline(
        faiss_db_path=DB_PATH,
        embedding_model="microsoft/BiomedNLP-BiomedBERT-base-uncased-abstract-fulltext",
    )

    return pipeline


pipeline = load_pipeline()


# -----------------------------
# Page Config
# -----------------------------

st.set_page_config(
    page_title="Pathology RAG",
    layout="wide"
)

st.title("🔬 Pathology Report Analysis System")

st.markdown(
"""
AI-powered search and question answering over pathology reports  
Vector database powered by **BiomedBERT + FAISS**
"""
)


# -----------------------------
# Session State
# -----------------------------

if "query_count" not in st.session_state:
    st.session_state.query_count = 0


# -----------------------------
# Sidebar
# -----------------------------

st.sidebar.header("System Info")

st.sidebar.write(f"Queries: {st.session_state.query_count}")

st.sidebar.write("Embedding Model:")
st.sidebar.write("BiomedBERT")

st.sidebar.write("Vector DB:")
st.sidebar.write("FAISS")


# -----------------------------
# Document Upload
# -----------------------------

st.sidebar.divider()
st.sidebar.header("📄 Upload Report")

with st.sidebar.form(key='upload_form', clear_on_submit=True):
    uploaded_file = st.file_uploader("Upload PDF Pathology Report", type=["pdf"])
    submit_btn = st.form_submit_button("Process Document")

if submit_btn and uploaded_file is not None:
    with st.spinner("Processing Document... this may take a while."):
        
        # Save file
        upload_dir = Path("uploaded_reports")
        upload_dir.mkdir(exist_ok=True)
        pdf_path = upload_dir / uploaded_file.name
        
        with open(pdf_path, "wb") as f:
            f.write(uploaded_file.getbuffer())
        
        # Instantiate updater
        updater = DynamicRAGUpdater(
            vector_db_path=DB_PATH,
            embedding_model="microsoft/BiomedNLP-BiomedBERT-base-uncased-abstract-fulltext",
            upload_dir=str(upload_dir)
        )
        
        # Process and add to vector database
        try:
            stats = updater.process_and_add_pdf(str(pdf_path))
            st.sidebar.success(f"Successfully processed `{uploaded_file.name}`")
            st.sidebar.json(stats)
            
            # Clear pipeline cache to reflect new db index
            load_pipeline.clear()
            
        except Exception as e:
            st.sidebar.error(f"Error during processing: {e}")

st.sidebar.divider()

# -----------------------------
# Query Input
# -----------------------------

st.header("🔎 Ask a Question")

question = st.text_area(
    "Enter your medical query",
    placeholder="What are common findings in breast cancer pathology?",
)

num_sources = st.slider(
    "Number of sources",
    min_value=1,
    max_value=10,
    value=5
)


# -----------------------------
# Search Button
# -----------------------------

if st.button("Search"):

    if question.strip() == "":
        st.warning("Please enter a question.")

    else:

        with st.spinner("Running RAG pipeline..."):

            st.session_state.query_count += 1

            result = pipeline.ask(
                question,
                top_k=num_sources
            )

        answer = result["answer"]

        st.subheader("Answer")

        st.markdown(answer)


        # Metadata
        st.subheader("Query Info")

        st.write({
            "query_number": st.session_state.query_count,
            "timestamp": datetime.now().strftime("%Y-%m-%d %H:%M:%S"),
            "sources_used": result["num_sources"]
        })


        # Sources
        st.subheader("Sources")

        sources = result["sources"]

        if not sources:
            st.write("No sources retrieved.")

        for i, source in enumerate(sources, 1):

            chunk = source["chunk"]

            with st.expander(f"Source {i} | {chunk['filename']}"):

                st.write(chunk["text"][:600])