app.py

import streamlit as st

# Set the page config immediately after importing streamlit
st.set_page_config(page_title="Climate Adaptation & Resilience Analyzer", layout="wide")
st.title("Climate Adaptation & Resilience Analyzer")

import os
import re
import base64
from tempfile import NamedTemporaryFile
from huggingface_hub import hf_hub_download
from typing import Optional
import pandas as pd
import folium
from streamlit_folium import st_folium
import anthropic

# Import all helper functions from app_helpers.py
from app_helpers import *

# Import necessary modules from LangChain and its community extensions
from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder
from langchain_core.messages import HumanMessage, AIMessage
from langchain_openai import ChatOpenAI, OpenAIEmbeddings
from langchain.chains import create_history_aware_retriever, create_retrieval_chain
from langchain.chains.combine_documents import create_stuff_documents_chain
from langchain.docstore.document import Document
from langchain_text_splitters import RecursiveCharacterTextSplitter
from langchain_community.document_loaders import PyPDFLoader, TextLoader
from langchain_community.vectorstores import FAISS
# ------------------------------------------------------------------------------
# LOAD DATA FILES
# ------------------------------------------------------------------------------
state_df = load_pickle_data("./maps_helpers/state_df.pkl")
county_df = load_pickle_data("./maps_helpers/county_df.pkl")
states_gdf = load_pickle_data("./maps_helpers/states_gdf_caps.pkl")
counties_gdf = load_pickle_data("./maps_helpers/counties_gdf_caps.pkl")
city_mapping_df = load_pickle_data("./maps_helpers/city_mapping_df.pkl")
city_plans_df = load_pickle_data("./maps_helpers/city_plans_df.pkl")

# ------------------------------------------------------------------------------
# CONSTANTS & CONFIGURATIONS
# ------------------------------------------------------------------------------
REGION_COLORS = {
    1: "#e41a1c",
    2: "#377eb8",
    3: "#4daf4a",
    4: "#984ea3",
    5: "#ff7f00",
    6: "#ffff33",
    7: "#a65628",
    8: "#f781bf",
    9: "#999999",
    10: "#66c2a5"
}

from huggingface_hub import hf_hub_download
import tempfile
import os
import streamlit as st
from typing import Optional

class HFHubDatasetManager:
    """Manager for accessing files directly from HF Hub"""
    
    def __init__(self, repo_id: str, token: Optional[str] = None):
        self.repo_id = repo_id
        self.token = token
        self._cache_dir = tempfile.mkdtemp()
        
    def download_pdf(self, filename: str, subfolder: str = "CAPS") -> str:
        """Download a PDF file from the HF repository."""
        file_path = f"{subfolder}/{filename}" if subfolder else filename
        
        try:
            local_path = hf_hub_download(
                repo_id=self.repo_id,
                filename=file_path,
                repo_type="dataset",
                cache_dir=self._cache_dir,
                token=self.token
            )
            return local_path
        except Exception as e:
            raise FileNotFoundError(f"Could not download {filename}: {str(e)}")

@st.cache_resource
def get_hf_hub_manager():
    """Get or create the HF Hub manager (cached)"""
    # TODO: REPLACE WITH YOUR ACTUAL REPOSITORY ID
    DATASET_REPO_ID = "your-username/climate-action-plans"  # CHANGE THIS!
    
    # Optional: for private repos, add your token to .streamlit/secrets.toml
    HF_TOKEN = st.secrets.get("HF_TOKEN", None) if "HF_TOKEN" in st.secrets else None
    
    return HFHubDatasetManager(DATASET_REPO_ID, token=HF_TOKEN)

def set_viewing_pdf_from_hf(pdf_filename):
    """Fetch PDF from HF dataset and set it for viewing"""
    try:
        manager = get_hf_hub_manager()
        pdf_path = manager.download_pdf(pdf_filename)
        st.session_state["viewing_pdf"] = pdf_path
    except FileNotFoundError as e:
        st.error(f"Could not load PDF: {str(e)}")

# ------------------------------------------------------------------------------
# HELPER FUNCTIONS FOR PDF VIEWER
# ------------------------------------------------------------------------------
def show_pdf(file_path):
    """
    Reads a PDF file and displays it in an embedded iframe.
    """
    try:
        with open(file_path, "rb") as f:
            base64_pdf = base64.b64encode(f.read()).decode("utf-8")
        pdf_display = (
            f'<iframe src="data:application/pdf;base64,{base64_pdf}" '
            f'width="700" height="1000" type="application/pdf"></iframe>'
        )
        st.markdown(pdf_display, unsafe_allow_html=True)
    except Exception as e:
        st.error(f"Could not load PDF: {e}")

def display_pdf_links(plan_list, state_abbr="CA", folder="CAPS"):
    """
    For each plan in plan_list, display a button. When clicked, the PDF is fetched
    from HuggingFace dataset and displayed.
    """
    if plan_list:
        for idx, plan in enumerate(plan_list):
            file_name = plan
            if not file_name.lower().endswith(".pdf"):
                parts = [part.strip() for part in plan.split(",")]
                if len(parts) == 3:
                    # Expected format: City, Year, Plan Type.
                    file_name = f"{parts[0]}, {state_abbr} {parts[2]} {parts[1]}.pdf"
                else:
                    # Fallback: simply append the .pdf extension.
                    file_name = plan + ".pdf"
            
            # CHANGED: Now uses set_viewing_pdf_from_hf instead of set_viewing_pdf
            st.button(
                plan, 
                key=f"pdf_{plan}_{idx}", 
                on_click=set_viewing_pdf_from_hf,  # Changed to HF version
                args=(file_name,)  # Only passing filename
            )
    else:
        st.write("None")

def set_viewing_pdf(pdf_path):
    st.session_state["viewing_pdf"] = pdf_path

#def display_pdf_links(plan_list, state_abbr="CA", folder="CAPS"):
#    """
#    For each plan in plan_list, display a button. When clicked, the PDF file name is constructed
#    and its full filepath (from `folder`) is saved into session state using an on_click callback.
#    """
#    if plan_list:
#        for idx, plan in enumerate(plan_list):
#            file_name = plan
#            if not file_name.lower().endswith(".pdf"):
#               parts = [part.strip() for part in plan.split(",")]
#               if len(parts) == 3:
#                    # Expected format: City, Year, Plan Type.
#                    file_name = f"{parts[0]}, {state_abbr} {parts[2]} {parts[1]}.pdf"
#                else:
#                    # Fallback: simply append the .pdf extension.
#                    file_name = plan + ".pdf"
#            pdf_path = os.path.join(folder, file_name)
#            # Append the index to the key to ensure uniqueness
#            st.button(plan, key=f"pdf_{plan}_{idx}", on_click=set_viewing_pdf, args=(pdf_path,))
#    else:
#        st.write("None")

# ------------------------------------------------------------------------------
# INTRO TEXT
# ------------------------------------------------------------------------------
st.markdown("""
### Welcome to the Climate Adaptation & Resilience Analyzer

This platform is a collection of integrated tools designed to support evidence-based climate planning, policy evaluation, and strategy development across the United States. It leverages artificial intelligence (AI), geospatial analysis, and document understanding to help users navigate and interpret a large corpus of local climate action and adaptation plans.

The Climate Adaptation & Resilience Analyzer enables users to:

- Explore national, state, and county-level **climate risk data and policy coverage** through interactive geospatial maps
- **Generate structured reports** from uploaded or existing climate action plans
- **Query** individual plans or entire document collections using natural language
- **Compare plans** across jurisdictions to evaluate strategic differences and shared approaches
- Access a **structured dataset** summarizing responses to standardized analytical questions

This suite of tools is intended for climate researchers, urban planners, policymakers, and practitioners seeking to understand and advance local climate resilience efforts. It supports both retrospective review of existing plans and the design of forward-looking strategies grounded in real-world policy examples.

To enable AI-powered features such as document querying and report generation, an **OpenAI API key** is required. You can obtain one [here](https://platform.openai.com/api-keys).
""")

# ------------------------------------------------------------------------------
# API KEYS INPUT
# ------------------------------------------------------------------------------
openai_api_key = st.text_input("OpenAI API Key", type="password")
anthropic_api_key = st.text_input("Anthropic API Key (Optional, required if using Long Context Models for Plan Comparison)", type="password")

# ------------------------------------------------------------------------------
# TABS SETUP (Added a new tab for Plan Insights)
# ------------------------------------------------------------------------------
(maps_tab, summary_tab, multi_plan_qa_tab, 
 document_qa_tab, plan_comparison_tab, plan_insights_tab) = st.tabs([
    "Geospatial Policy Explorer",           # State & county-level maps with risk + plan querying
    "Climate Plan Report Generator",        # Generate or view structured reports
    "Cross-Plan Knowledge Query",           # Ask questions across the full corpus
    "Single-Plan Query Assistant",          # Ask detailed questions about one specific plan
    "Comparative Policy Analysis",          # Compare focus plan vs. others
    "Dataset Overview & Insights"           # Tabular view of plan content and metadata
])
# ------------------------------------------------------------------------------
# TAB 1: MAPS
# ------------------------------------------------------------------------------
with maps_tab:
    # --------------------------------------------------------------------------
    # STATE-LEVEL POLICY TRACKER
    # --------------------------------------------------------------------------
    st.markdown("""
        # State and County Level Policy Trackers
        
        These two interactive tools allow users to explore the geographic distribution of climate action plans across the United States and gain deep insights into both plan content and place-based climate data.

        Using the built-in maps, you can:

        - Identify which states and counties have climate action plans.
        - Access and view those plans directly in the platform.
        - Query plans and locations using natural language powered by GPT-4o.
        - Compare local risks (e.g., wildfire, flooding, heat, drought), FEMA risk profiles, and environmental justice indicators.
        - Analyze areas even if they do not yet have a formal climate plan, using external data integrated into the map layers.
        - The QA tool is powered by GPT-4o. It has access to detailed information about each plan in the selected locale, high level information about climate action plans in the EPA region, external data, higher-level information about climate action plans acrouss the United States

        How to Use

        - Click on any state or county on the map to bring up its associated data.
        - View available Climate Action Plans (CAPs) for that region and open them in the built-in PDF viewer.
        - View the distribution of climate action plans across the United States using the City Markers.
        - Enter your OpenAI API key to enable the query engine.
        - Type in a question — the more specific and detailed, the better.
        - Example: “What climate hazards is this county most at risk for in the mid-century under a high emissions scenario?”
        - Submit your query to receive an AI-generated response grounded in both the policy documents and the location’s climate data.""")


    state_tab, county_tab = st.tabs(["State-Level Policy Tracker", "County-Level Policy Tracker"])

    with state_tab:

        # Initialize state map with no default tiles; add an OpenStreetMap layer.
        m_state = folium.Map(location=[35.3, -97.6], zoom_start=4, tiles=None)
        folium.TileLayer("OpenStreetMap", control=False).add_to(m_state)

        # Add state boundaries with tooltips.
        state_boundaries = folium.FeatureGroup(name="State Boundaries", control=False)
        tooltip_state = folium.GeoJsonTooltip(
            fields=["NAME", "POP_TT", "EPA_REGION"],
            aliases=["State:", "Population:", "EPA Region:"],
            localize=True,
            sticky=False,
            labels=True,
            style="""
                background-color: #F0EFEF;
                border: 2px solid black;
                border-radius: 3px;
                box-shadow: 3px;
            """,
            max_width=800,
        )
        folium.GeoJson(
            states_gdf,
            style_function=lambda x: {
                "fillColor": REGION_COLORS.get(x["properties"].get("EPA_REGION"), "transparent"),
                "color": "black",
                "fillOpacity": 0.4,
                "weight": 1
            },
            tooltip=tooltip_state,
            highlight_function=lambda x: {"weight": 2, "color": "blue"}
        ).add_to(state_boundaries)
        state_boundaries.add_to(m_state)

        # Add city markers to the map.
        add_city_markers(m_state)
        folium.LayerControl(collapsed=False).add_to(m_state)

        # Define a three-column layout for additional info, the map, and the right column.
        cols_state = st.columns([3, 6, 1])
        with cols_state[1]:
            st.subheader("US State Map")
            st_data_state = st_folium(m_state, width=900, height=650)
            if st.session_state.get("viewing_pdf"):
                with st.expander("PDF Viewer", expanded=True):
                    pdf_file = st.session_state["viewing_pdf"]
                    st.write("Viewing:", os.path.basename(pdf_file))
                    show_pdf(pdf_file)
        
        with cols_state[0]:
            st.markdown("### Additional Information")
            if st_data_state.get("last_active_drawing"):
                props = st_data_state["last_active_drawing"].get("properties", {})
                state_name = props.get("NAME", "N/A")
                population = props.get("POP_TT", "N/A")
                fips = props.get("STATE_FIPS", "N/A")
                state_abbr = props.get("STATE_ABBR", "CA")
                n_caps = props.get("n_caps", 0)
                epa_region = props.get("EPA_REGION", "N/A")
                plan_list = props.get("plan_list", [])

                st.write("**State:**", state_name)
                st.write("**Population:**", population)
                st.write("**FIPS:**", fips)
                st.write("**EPA Region:**", f"{int(epa_region)}")
                st.write("**Number of Climate Action Plans:**", f"{int(n_caps):,}")
                
                with st.expander("Cities with Climate Action Plans:"):
                    display_pdf_links(plan_list, state_abbr=state_abbr)
                
                # (Additional risk index and FEMA risk info displayed in expanders)
                with st.expander("NRI Future Risk Index (Higher Warming Pathway):"):
                    st.write("**Mid-Century Coastal Flooding Risk (Percentile):**", props.get("CFLD_MID_HIGHER_PRISKS", "N/A"))
                    st.write("**Late-Century Coastal Flooding Risk (Percentile):**", props.get("CFLD_LATE_HIGHER_PRISKS", "N/A"))
                    st.write("**Mid-Century Coastal Flooding Hazard Multiplier:**", props.get("CFLD_MID_HIGHER_HM", "N/A"))
                    st.write("**Late-Century Coastal Flooding Hazard Multiplier:**", props.get("CFLD_LATE_HIGHER_HM", "N/A"))
                    st.write("**Mid-Century Wildfire Risk (Percentile):**", props.get("WFIR_MID_HIGHER_PRISKS", "N/A"))
                    st.write("**Late-Century Wildfire Risk (Percentile):**", props.get("WFIR_LATE_HIGHER_PRISKS", "N/A"))
                    st.write("**Mid-Century Wildfire Hazard Multiplier:**", props.get("WFIR_MID_HIGHER_HM", "N/A"))
                    st.write("**Late-Century Wildfire Hazard Multiplier:**", props.get("WFIR_LATE_HIGHER_HM", "N/A"))
                    st.write("**Mid-Century Drought Risk (Percentile):**", props.get("DRGT_MID_HIGHER_PRISKS", "N/A"))
                    st.write("**Late-Century Drought Risk (Percentile):**", props.get("DRGT_LATE_HIGHER_PRISKS", "N/A"))
                    st.write("**Mid-Century Drought Hazard Multiplier:**", props.get("DRGT_MID_HIGHER_HM", "N/A"))
                    st.write("**Late-Century Drought Hazard Multiplier:**", props.get("DRGT_LATE_HIGHER_HM", "N/A"))
                    st.write("**Mid-Century Hurricane Risk (Percentile):**", props.get("HRCN_MID_HIGHER_PRISKS", "N/A"))
                    st.write("**Late-Century Hurricane Risk (Percentile):**", props.get("HRCN_LATE_HIGHER_PRISKS", "N/A"))
                    st.write("**Mid-Century Hurricane Hazard Multiplier:**", props.get("HRCN_MID_HIGHER_HM", "N/A"))
                    st.write("**Late-Century Hurricane Hazard Multiplier:**", props.get("HRCN_LATE_HIGHER_HM", "N/A"))
                    st.write("**Mid-Century Extreme Heat Risk (Percentile):**", props.get("EXHT_L95_MID_HIGHER_PRISKS", "N/A"))
                    st.write("**Late-Century Extreme Heat Risk (Percentile):**", props.get("EXHT_L95_LATE_HIGHER_PRISKS", "N/A"))
                    st.write("**Mid-Century Extreme Heat Hazard Multiplier:**", props.get("EXHT_L95_MID_HIGHER_HM", "N/A"))
                    st.write("**Late-Century Extreme Heat Hazard Multiplier:**", props.get("EXHT_L95_LATE_HIGHER_HM", "N/A"))

                with st.expander("FEMA Risk Profile:"):
                    st.write("**Disaster Risk (Percentile):**", props.get("RISK_SCORE", "N/A"))
                    st.write("**Disaster Loss ($/year):**", props.get("EAL_VALT", "N/A"))
                    st.write("**Social Vulnerability (Percentile):**", props.get("SOVI_SCORE", "N/A"))
                    st.write("**Community Resilience (Percentile):**", props.get("RESL_SCORE", "N/A"))
                    st.write("**Annual Avalanche Loss ($/year):**", props.get("AVLN_EALT", "N/A"))
                    st.write("**Annual Avalanche Loss (Percentile):**", props.get("AVLN_EALS", "N/A"))
                    st.write("**Annual Coastal Flooding Loss ($/year):**", props.get("CFLD_EALT", "N/A"))
                    st.write("**Annual Coastal Flooding Loss (Percentile):**", props.get("CFLD_EALS", "N/A"))
                    st.write("**Annual Cold Wave Loss ($/year):**", props.get("CWAV_EALT", "N/A"))
                    st.write("**Annual Cold Wave Loss (Percentile):**", props.get("CWAV_EALS", "N/A"))
                    st.write("**Annual Drought Loss ($/year):**", props.get("DRGT_EALT", "N/A"))
                    st.write("**Annual Drought Loss (Percentile):**", props.get("DRGT_EALS", "N/A"))
                    st.write("**Annual Hail Loss ($/year):**", props.get("HAIL_EALT", "N/A"))
                    st.write("**Annual Hail Loss (Percentile):**", props.get("HAIL_EALS", "N/A"))
                    st.write("**Annual Heat Wave Loss ($/year):**", props.get("HWAV_EALT", "N/A"))
                    st.write("**Annual Heat Wave Loss (Percentile):**", props.get("HWAV_EALS", "N/A"))
                    st.write("**Annual Hurricane Loss ($/year):**", props.get("HRCN_EALT", "N/A"))
                    st.write("**Annual Hurricane Loss (Percentile):**", props.get("HRCN_EALS", "N/A"))
                    st.write("**Annual Ice Storm Loss ($/year):**", props.get("ISTM_EALT", "N/A"))
                    st.write("**Annual Ice Storm Loss (Percentile):**", props.get("ISTM_EALS", "N/A"))
                    st.write("**Annual Landslide Loss ($/year):**", props.get("LNDS_EALT", "N/A"))
                    st.write("**Annual Landslide Loss (Percentile):**", props.get("LNDS_EALS", "N/A"))
                    st.write("**Annual River Flooding Loss ($/year):**", props.get("RFLD_EALT", "N/A"))
                    st.write("**Annual River Flooding Loss (Percentile):**", props.get("RFLD_EALS", "N/A"))
                    st.write("**Annual Wind Loss ($/year):**", props.get("SWND_EALT", "N/A"))
                    st.write("**Annual Wind Loss (Percentile):**", props.get("SWND_EALS", "N/A"))
                    st.write("**Annual Tornado Loss ($/year):**", props.get("TRND_EALT", "N/A"))
                    st.write("**Annual Tornado Loss (Percentile):**", props.get("TRND_EALS", "N/A"))
                    st.write("**Annual Winter Weather Loss ($/year):**", props.get("WNTW_EALT", "N/A"))
                    st.write("**Annual Winter Weather Loss (Percentile):**", props.get("WNTW_EALS", "N/A"))
                
                with st.expander("CEJST Data:"):
                    st.write("**Share of properties at risk of flood in 30 years (percentile):**", props.get("Share of properties at risk of flood in 30 years (percentile)", "N/A"))
                    st.write("**Share of properties at risk of flood in 30 years:**", props.get("Share of properties at risk of flood in 30 years", "N/A"))
                    st.write("**Share of properties at risk of fire in 30 years (percentile):**", props.get("Share of properties at risk of fire in 30 years (percentile)", "N/A"))
                    st.write("**Share of properties at risk of fire in 30 years:**", props.get("Share of properties at risk of fire in 30 years", "N/A"))
                    st.write("**Energy burden (percentile):**", props.get("Energy burden (percentile)", "N/A"))
                    st.write("**PM2.5 (percentile):**", props.get("PM2.5 in the air (percentile)", "N/A"))
                    st.write("**PM2.5 (Volume):**", props.get("PM2.5 in the air", "N/A"))
                    st.write("**Impervious surface or cropland:**", props.get("Share of the tract's land area that is covered by impervious surface or cropland as a percent", "N/A"))
                    st.write("**Asthma Prevalence (Percentile):**", props.get("Current asthma among adults aged greater than or equal to 18 years", "N/A"))
                
                # Build extra context for the QA chain
                extra_context = (
                    f"State: {state_name}\n"
                    f"Population: {population}\n"
                    f"FIPS: {fips}\n"
                    f"EPA Region: {epa_region}\n"
                    f"Climate Action Plans: {', '.join(plan_list) if plan_list else 'No climate action plans'}\n"
                    f"NRI Future Risk Index (Higher Warming Pathway):\n"
                    f"Mid-Century Coastal Flooding Risk (Percentile): {props.get('CFLD_MID_HIGHER_PRISKS', 'N/A')}\n"
                    f"Late-Century Coastal Flooding Risk (Percentile): {props.get('CFLD_LATE_HIGHER_PRISKS', 'N/A')}\n"
                    f"Mid-Century Coastal Flooding Hazard Multiplier: {props.get('CFLD_MID_HIGHER_HM', 'N/A')}\n"
                    f"Late-Century Coastal Flooding Hazard Multiplier: {props.get('CFLD_LATE_HIGHER_HM', 'N/A')}\n"
                    f"Mid-Century Wildfire Risk (Percentile): {props.get('WFIR_MID_HIGHER_PRISKS', 'N/A')}\n"
                    f"Late-Century Wildfire Risk (Percentile): {props.get('WFIR_LATE_HIGHER_PRISKS', 'N/A')}\n"
                    f"Mid-Century Wildfire Hazard Multiplier: {props.get('WFIR_MID_HIGHER_HM', 'N/A')}\n"
                    f"Late-Century Wildfire Hazard Multiplier: {props.get('WFIR_LATE_HIGHER_HM', 'N/A')}\n"
                    f"Mid-Century Drought Risk (Percentile): {props.get('DRGT_MID_HIGHER_PRISKS', 'N/A')}\n"
                    f"Late-Century Drought Risk (Percentile): {props.get('DRGT_LATE_HIGHER_PRISKS', 'N/A')}\n"
                    f"Mid-Century Drought Hazard Multiplier: {props.get('DRGT_MID_HIGHER_HM', 'N/A')}\n"
                    f"Late-Century Drought Hazard Multiplier: {props.get('DRGT_LATE_HIGHER_HM', 'N/A')}\n"
                    f"FEMA Risk Profile:\n"
                    f"Disaster Risk (Percentile): {props.get('RISK_SCORE', 'N/A')}\n"
                    f"Disaster Loss ($/year): {props.get('EAL_VALT', 'N/A')}\n"
                    f"Social Vulnerability (Percentile): {props.get('SOVI_SCORE', 'N/A')}\n"
                    f"Community Resilience (Percentile): {props.get('RESL_SCORE', 'N/A')}\n"
                    f"Annual Avalanche Loss ($/year): {props.get('AVLN_EALT', 'N/A')}\n"
                    f"Annual Avalanche Loss (Percentile): {props.get('AVLN_EALS', 'N/A')}\n"
                    f"Annual Coastal Flooding Loss ($/year): {props.get('CFLD_EALT', 'N/A')}\n"
                    f"Annual Coastal Flooding Loss (Percentile): {props.get('CFLD_EALS', 'N/A')}\n"
                    f"Annual Cold Wave Loss ($/year): {props.get('CWAV_EALT', 'N/A')}\n"
                    f"Annual Cold Wave Loss (Percentile): {props.get('CWAV_EALS', 'N/A')}\n"
                    f"Annual Drought Loss ($/year): {props.get('DRGT_EALT', 'N/A')}\n"
                    f"Annual Drought Loss (Percentile): {props.get('DRGT_EALS', 'N/A')}\n"
                    f"Annual Hail Loss ($/year): {props.get('HAIL_EALT', 'N/A')}\n"
                    f"Annual Hail Loss (Percentile): {props.get('HAIL_EALS', 'N/A')}\n"
                    f"Annual Heat Wave Loss ($/year): {props.get('HWAV_EALT', 'N/A')}\n"
                    f"Annual Heat Wave Loss (Percentile): {props.get('HWAV_EALS', 'N/A')}\n"
                    f"Annual Hurricane Loss ($/year): {props.get('HRCN_EALT', 'N/A')}\n"
                    f"Annual Hurricane Loss (Percentile): {props.get('HRCN_EALS', 'N/A')}\n"
                    f"Annual Ice Storm Loss ($/year): {props.get('ISTM_EALT', 'N/A')}\n"
                    f"Annual Ice Storm Loss (Percentile): {props.get('ISTM_EALS', 'N/A')}\n"
                    f"Annual Landslide Loss ($/year): {props.get('LNDS_EALT', 'N/A')}\n"
                    f"Annual Landslide Loss (Percentile): {props.get('LNDS_EALS', 'N/A')}\n"
                    f"Annual River Flooding Loss ($/year): {props.get('RFLD_EALT', 'N/A')}\n"
                    f"Annual River Flooding Loss (Percentile): {props.get('RFLD_EALS', 'N/A')}\n"   
                    f"Annual Wind Loss ($/year): {props.get('SWND_EALT', 'N/A')}\n"
                    f"Annual Wind Loss (Percentile): {props.get('SWND_EALS', 'N/A')}\n"
                    f"Annual Tornado Loss ($/year): {props.get('TRND_EALT', 'N/A')}\n"
                    f"Annual Tornado Loss (Percentile): {props.get('TRND_EALS', 'N/A')}\n"
                    f"Annual Winter Weather Loss ($/year): {props.get('WNTW_EALT', 'N/A')}\n"
                    f"Annual Winter Weather Loss (Percentile): {props.get('WNTW_EALS', 'N/A')}\n"
                    f"CEJST Data:\n"
                    f"Share of properties at risk of flood in 30 years (percentile): {props.get('Share of properties at risk of flood in 30 years (percentile)', 'N/A')}\n"
                    f"Share of properties at risk of flood in 30 years: {props.get('Share of properties at risk of flood in 30 years', 'N/A')}\n"
                    f"Share of properties at risk of fire in 30 years (percentile): {props.get('Share of properties at risk of fire in 30 years (percentile)', 'N/A')}\n"
                    f"Share of properties at risk of fire in 30 years: {props.get('Share of properties at risk of fire in 30 years', 'N/A')}\n"
                    f"Energy burden (percentile): {props.get('Energy burden (percentile)', 'N/A')}\n"
                    f"PM2.5 (percentile): {props.get('PM2.5 in the air (percentile)', 'N/A')}\n"
                    f"PM2.5 (Volume): {props.get('PM2.5 in the air', 'N/A')}\n"
                    f"Impervious surface or cropland: {props.get('Share of the tracts land area that is covered by impervious surface or cropland as a percent', 'N/A')}\n"
                    f"Asthma Prevalence (Percentile): {props.get('Current asthma among adults aged greater than or equal to 18 years', 'N/A')}\n"
                )

                user_question = st.text_input("Ask a Question about the selected State:", key="state_question")
                if st.button("Submit State Query", key="state_submit"):
                    if openai_api_key and user_question:
                        # Call the new maps_qa function (formerly answer_question)
                        result = maps_qa(openai_api_key, user_question, extra_context, plan_list, state_abbr, epa_region)
                        st.write(result)
                    else:
                        st.error("Please provide both an API key and a question.")
            else:
                st.info("Click on a state to view details.")
        
        with cols_state[2]:
            legend_html = generate_legend_html(REGION_COLORS)
            st.markdown(legend_html, unsafe_allow_html=True)

    # --------------------------------------------------------------------------
    # COUNTY-LEVEL POLICY TRACKER
    # --------------------------------------------------------------------------
    with county_tab:
        
        # Initialize county map with no default tiles; add an OpenStreetMap layer.
        m_county = folium.Map(location=[35.3, -97.6], zoom_start=4, tiles=None)
        folium.TileLayer("OpenStreetMap", control=False).add_to(m_county)

        # Add county boundaries with tooltips.
        county_boundaries = folium.FeatureGroup(name="County Boundaries", control=False)
        tooltip_county = folium.GeoJsonTooltip(
            fields=["NAME", "POP_TT", "FIPS_TT", "EPA_REGION"],
            aliases=["County:", "Population:", "FIPS:", "EPA Region:"],
            localize=True,
            sticky=False,
            labels=True,
            style="""
                background-color: #F0EFEF;
                border: 2px solid black;
                border-radius: 3px;
                box-shadow: 3px;
            """,
            max_width=800,
        )
        folium.GeoJson(
            counties_gdf,
            style_function=lambda x: {
                "fillColor": REGION_COLORS.get(x["properties"].get("EPA_REGION"), "transparent"),
                "color": "black",
                "fillOpacity": 0.4,
                "weight": 1
            },
            tooltip=tooltip_county,
            highlight_function=lambda x: {"weight": 2, "color": "blue"}
        ).add_to(county_boundaries)
        county_boundaries.add_to(m_county)

        # Add city markers.
        add_city_markers(m_county)
        folium.LayerControl(collapsed=False).add_to(m_county)

        # Define a three-column layout for county info, map, and the right column.
        cols_county = st.columns([3, 6, 1])
        with cols_county[1]:
            st.subheader("United States County Map")
            st_data_county = st_folium(m_county, width=900, height=650)
            if st.session_state.get("viewing_pdf"):
                with st.expander("PDF Viewer", expanded=True):
                    pdf_file = st.session_state["viewing_pdf"]
                    st.write("Viewing:", os.path.basename(pdf_file))
                    show_pdf(pdf_file)
        
        with cols_county[0]:
            st.markdown("### Additional Information")
            if st_data_county.get("last_active_drawing"):
                props = st_data_county["last_active_drawing"].get("properties", {})
                county_name = props.get("NAME", "N/A")
                epa_region = props.get("EPA_REGION", "N/A")
                population = props.get("POP_TT", "N/A")
                fips = props.get("FIPS_TT", "N/A")
                n_caps = props.get("n_caps", 0)
                state_abbr = props.get("STATE_ABBR", "CA")
                plan_list = props.get("plan_list", [])
                
                st.write("**County:**", county_name)
                st.write("**Population:**", population)
                st.write("**FIPS:**", fips)
                st.write("**EPA Region:**", epa_region)
                st.write("**Number of Climate Action Plans:**", f"{int(n_caps):,}")
                
                with st.expander("Cities with Climate Action Plans:"):
                    display_pdf_links(plan_list, state_abbr=state_abbr)
                
                with st.expander("#### NRI Future Risk Index (Higher Warming Pathway):"):
                    st.write("**Coastal Flooding Mid-Century Projected Risk:**", props.get("CFLD_MID_HIGHER_PRISKS", "N/A"))
                    st.write("**Coastal Flooding Late-Century Projected Risk:**", props.get("CFLD_LATE_HIGHER_PRISKS", "N/A"))
                    st.write("**Coastal Flooding Mid-Century Hazard Multiplier:**", props.get("CFLD_MID_HIGHER_HM", "N/A"))
                    st.write("**Coastal Flooding Late-Century Hazard Multiplier:**", props.get("CFLD_LATE_HIGHER_HM", "N/A"))
                    st.write("**Wildfire Mid-Century Projected Risk:**", props.get("WFIR_MID_HIGHER_PRISKS", "N/A"))
                    st.write("**Wildfire Late-Century Projected Risk:**", props.get("WFIR_LATE_HIGHER_PRISKS", "N/A"))
                    st.write("**Wildfire Mid-Century Hazard Multiplier:**", props.get("WFIR_MID_HIGHER_HM", "N/A"))
                    st.write("**Wildfire Late-Century Hazard Multiplier:**", props.get("WFIR_LATE_HIGHER_HM", "N/A"))
                    st.write("**Drought Mid-Century Projected Risk:**", props.get("DRGT_MID_HIGHER_PRISKS", "N/A"))
                    st.write("**Drought Late-Century Projected Risk:**", props.get("DRGT_LATE_HIGHER_PRISKS", "N/A"))
                    st.write("**Drought Mid-Century Hazard Multiplier:**", props.get("DRGT_MID_HIGHER_HM", "N/A"))
                    st.write("**Drought Late-Century Hazard Multiplier:**", props.get("DRGT_LATE_HIGHER_HM", "N/A"))
                    st.write("**Hurricane Mid-Century Projected Risk:**", props.get("HRCN_MID_HIGHER_PRISKS", "N/A"))
                    st.write("**Hurricane Late-Century Projected Risk:**", props.get("HRCN_LATE_HIGHER_PRISKS", "N/A"))
                    st.write("**Hurricane Mid-Century Hazard Multiplier:**", props.get("HRCN_MID_HIGHER_HM", "N/A"))
                    st.write("**Hurricane Late-Century Hazard Multiplier:**", props.get("HRCN_LATE_HIGHER_HM", "N/A"))
                    st.write("**Extreme Heat Mid-Century Projected Risk:**", props.get("EXHT_L95_MID_HIGHER_PRISKS", "N/A"))
                    st.write("**Extreme Heat Late-Century Projected Risk:**", props.get("EXHT_L95_LATE_HIGHER_PRISKS", "N/A"))
                    st.write("**Extreme Heat Mid-Century Hazard Multiplier:**", props.get("EXHT_L95_MID_HIGHER_HM", "N/A"))
                    st.write("**Extreme Heat Late-Century Hazard Multiplier:**", props.get("EXHT_L95_LATE_HIGHER_HM", "N/A"))
                
                with st.expander("#### FEMA Risk Profile:"):
                    st.write("**Disaster Risk (Percentile):**", props.get("RISK_SCORE", "N/A"))
                    st.write("**Disaster Risk (Percentile, relative to state):**", props.get("RISK_SPCTL", "N/A"))
                    st.write("**Disaster Loss ($/year):**", props.get("EAL_VALT", "N/A"))
                    st.write("**Social Vulnerability (Percentile):**", props.get("SOVI_SCORE", "N/A"))
                    st.write("**Community Resilience (Percentile):**", props.get("RESL_SCORE", "N/A"))
                    st.write("**Annual Avalanche Loss ($/year):**", props.get("AVLN_EALT", "N/A"))
                    st.write("**Annual Avalanche Loss (Percentile):**", props.get("AVLN_EALS", "N/A"))
                    st.write("**Annual Coastal Flooding Loss ($/year):**", props.get("CFLD_EALT", "N/A"))
                    st.write("**Annual Coastal Flooding Loss (Percentile):**", props.get("CFLD_EALS", "N/A"))
                    st.write("**Annual Cold Wave Loss ($/year):**", props.get("CWAV_EALT", "N/A"))
                    st.write("**Annual Cold Wave Loss (Percentile):**", props.get("CWAV_EALS", "N/A"))
                    st.write("**Annual Drought Loss ($/year):**", props.get("DRGT_EALT", "N/A"))
                    st.write("**Annual Drought Loss (Percentile):**", props.get("DRGT_EALS", "N/A"))
                    st.write("**Annual Hail Loss ($/year):**", props.get("HAIL_EALT", "N/A"))
                    st.write("**Annual Hail Loss (Percentile):**", props.get("HAIL_EALS", "N/A"))
                    st.write("**Annual Heat Wave Loss ($/year):**", props.get("HWAV_EALT", "N/A"))
                    st.write("**Annual Heat Wave Loss (Percentile):**", props.get("HWAV_EALS", "N/A"))
                    st.write("**Annual Hurricane Loss ($/year):**", props.get("HRCN_EALT", "N/A"))
                    st.write("**Annual Hurricane Loss (Percentile):**", props.get("HRCN_EALS", "N/A"))
                    st.write("**Annual Ice Storm Loss ($/year):**", props.get("ISTM_EALT", "N/A"))
                    st.write("**Annual Ice Storm Loss (Percentile):**", props.get("ISTM_EALS", "N/A"))
                    st.write("**Annual Landslide Loss ($/year):**", props.get("LNDS_EALT", "N/A"))
                    st.write("**Annual Landslide Loss (Percentile):**", props.get("LNDS_EALS", "N/A"))
                    st.write("**Annual River Flooding Loss ($/year):**", props.get("RFLD_EALT", "N/A"))
                    st.write("**Annual River Flooding Loss (Percentile):**", props.get("RFLD_EALS", "N/A"))
                    st.write("**Annual Wind Loss ($/year):**", props.get("SWND_EALT", "N/A"))
                    st.write("**Annual Wind Loss (Percentile):**", props.get("SWND_EALS", "N/A"))
                    st.write("**Annual Tornado Loss ($/year):**", props.get("TRND_EALT", "N/A"))
                    st.write("**Annual Tornado Loss (Percentile):**", props.get("TRND_EALS", "N/A"))
                    st.write("**Annual Winter Weather Loss ($/year):**", props.get("WNTW_EALT", "N/A"))
                    st.write("**Annual Winter Weather Loss (Percentile):**", props.get("WNTW_EALS", "N/A"))

                with st.expander("#### CEJST Data:"):
                    st.write("**Share of properties at risk of flood in 30 years (percentile):**", props.get("Share of properties at risk of flood in 30 years (percentile)", "N/A"))
                    st.write("**Share of properties at risk of flood in 30 years:**", props.get("Share of properties at risk of flood in 30 years", "N/A"))
                    st.write("**Share of properties at risk of fire in 30 years (percentile):**", props.get("Share of properties at risk of fire in 30 years (percentile)", "N/A"))
                    st.write("**Share of properties at risk of fire in 30 years:**", props.get("Share of properties at risk of fire in 30 years", "N/A"))
                    st.write("**Energy burden (percentile):**", props.get("Energy burden (percentile)", "N/A"))
                    st.write("**PM2.5 (percentile):**", props.get("PM2.5 in the air (percentile)", "N/A"))
                    st.write("**PM2.5 (Volume):**", props.get("PM2.5 in the air", "N/A"))
                    st.write("**Impervious surface or cropland:**", props.get("Share of the tract's land area that is covered by impervious surface or cropland as a percent", "N/A"))
                    st.write("**Asthma Prevalence (Percentile):**", props.get("Current asthma among adults aged greater than or equal to 18 years", "N/A"))
                            
                # Build extra context for the QA chain
                extra_context = (
                    f"County: {county_name}\n"
                    f"Population: {population}\n"
                    f"FIPS: {fips}\n"
                    f"Climate Action Plans: {', '.join(plan_list) if plan_list else 'No climate action plans'}\n"
                    f"NRI Future Risk Index (Higher Warming Pathway):\n"
                    f"Coastal Flooding Mid-Century Projected Risk: {props.get('CFLD_MID_HIGHER_PRISKS', 'N/A')}\n"
                    f"Coastal Flooding Late-Century Projected Risk: {props.get('CFLD_LATE_HIGHER_PRISKS', 'N/A')}\n"
                    f"Coastal Flooding Mid-Century Hazard Multiplier: {props.get('CFLD_MID_HIGHER_HM', 'N/A')}\n"
                    f"Coastal Flooding Late-Century Hazard Multiplier: {props.get('CFLD_LATE_HIGHER_HM', 'N/A')}\n"
                    f"Wildfire Mid-Century Projected Risk: {props.get('WFIR_MID_HIGHER_PRISKS', 'N/A')}\n"
                    f"Wildfire Late-Century Projected Risk: {props.get('WFIR_LATE_HIGHER_PRISKS', 'N/A')}\n"
                    f"Wildfire Mid-Century Hazard Multiplier: {props.get('WFIR_MID_HIGHER_HM', 'N/A')}\n"
                    f"Wildfire Late-Century Hazard Multiplier: {props.get('WFIR_LATE_HIGHER_HM', 'N/A')}\n"
                    f"Drought Mid-Century Projected Risk: {props.get('DRGT_MID_HIGHER_PRISKS', 'N/A')}\n"
                    f"Drought Late-Century Projected Risk: {props.get('DRGT_LATE_HIGHER_PRISKS', 'N/A')}\n"
                    f"Drought Mid-Century Hazard Multiplier: {props.get('DRGT_MID_HIGHER_HM', 'N/A')}\n"
                    f"Drought Late-Century Hazard Multiplier: {props.get('DRGT_LATE_HIGHER_HM', 'N/A')}\n"
                    f"FEMA Risk Profile:\n"
                    f"Disaster Risk (Percentile): {props.get('RISK_SCORE', 'N/A')}\n"
                    f"Disaster Risk (Percentile, relative to state): {props.get('RISK_SPCTL', 'N/A')}\n"
                    f"Disaster Loss ($/year): {props.get('EAL_VALT', 'N/A')}\n"
                    f"Social Vulnerability (Percentile): {props.get('SOVI_SCORE', 'N/A')}\n"
                    f"Community Resilience (Percentile): {props.get('RESL_SCORE', 'N/A')}\n"
                    f"Annual Avalanche Loss ($/year): {props.get('AVLN_EALT', 'N/A')}\n"
                    f"Annual Avalanche Loss (Percentile): {props.get('AVLN_EALS', 'N/A')}\n"
                    f"Annual Coastal Flooding Loss ($/year): {props.get('CFLD_EALT', 'N/A')}\n"
                    f"Annual Coastal Flooding Loss (Percentile): {props.get('CFLD_EALS', 'N/A')}\n"
                    f"Annual Cold Wave Loss ($/year): {props.get('CWAV_EALT', 'N/A')}\n"
                    f"Annual Cold Wave Loss (Percentile): {props.get('CWAV_EALS', 'N/A')}\n"
                    f"Annual Drought Loss ($/year): {props.get('DRGT_EALT', 'N/A')}\n"
                    f"Annual Drought Loss (Percentile): {props.get('DRGT_EALS', 'N/A')}\n"
                    f"Annual Hail Loss ($/year): {props.get('HAIL_EALT', 'N/A')}\n"
                    f"Annual Hail Loss (Percentile): {props.get('HAIL_EALS', 'N/A')}\n"
                    f"Annual Heat Wave Loss ($/year): {props.get('HWAV_EALT', 'N/A')}\n"
                    f"Annual Heat Wave Loss (Percentile): {props.get('HWAV_EALS', 'N/A')}\n"
                    f"Annual Hurricane Loss ($/year): {props.get('HRCN_EALT', 'N/A')}\n"
                    f"Annual Hurricane Loss (Percentile): {props.get('HRCN_EALS', 'N/A')}\n"
                    f"Annual Ice Storm Loss ($/year): {props.get('ISTM_EALT', 'N/A')}\n"
                    f"Annual Ice Storm Loss (Percentile): {props.get('ISTM_EALS', 'N/A')}\n"
                    f"Annual Landslide Loss ($/year): {props.get('LNDS_EALT', 'N/A')}\n"
                    f"Annual Landslide Loss (Percentile): {props.get('LNDS_EALS', 'N/A')}\n"
                    f"Annual River Flooding Loss ($/year): {props.get('RFLD_EALT', 'N/A')}\n"
                    f"Annual River Flooding Loss (Percentile): {props.get('RFLD_EALS', 'N/A')}\n"   
                    f"Annual Wind Loss ($/year): {props.get('SWND_EALT', 'N/A')}\n"
                    f"Annual Wind Loss (Percentile): {props.get('SWND_EALS', 'N/A')}\n"
                    f"Annual Tornado Loss ($/year): {props.get('TRND_EALT', 'N/A')}\n"
                    f"Annual Tornado Loss (Percentile): {props.get('TRND_EALS', 'N/A')}\n"
                    f"Annual Winter Weather Loss ($/year): {props.get('WNTW_EALT', 'N/A')}\n"
                    f"Annual Winter Weather Loss (Percentile): {props.get('WNTW_EALS', 'N/A')}\n"
                    f"CEJST Data:\n"
                    f"Share of properties at risk of flood in 30 years (percentile): {props.get('Share of properties at risk of flood in 30 years (percentile)', 'N/A')}\n"
                    f"Share of properties at risk of flood in 30 years: {props.get('Share of properties at risk of flood in 30 years', 'N/A')}\n"
                    f"Share of properties at risk of fire in 30 years (percentile): {props.get('Share of properties at risk of fire in 30 years (percentile)', 'N/A')}\n"
                    f"Share of properties at risk of fire in 30 years: {props.get('Share of properties at risk of fire in 30 years', 'N/A')}\n"
                    f"Energy burden (percentile): {props.get('Energy burden (percentile)', 'N/A')}\n"
                    f"PM2.5 (percentile): {props.get('PM2.5 in the air (percentile)', 'N/A')}\n"
                    f"PM2.5 (Volume): {props.get('PM2.5 in the air', 'N/A')}\n"
                    f"Impervious surface or cropland: {props.get('Share of the tracts land area that is covered by impervious surface or cropland as a percent', 'N/A')}\n"
                    f"Asthma Prevalence (Percentile): {props.get('Current asthma among adults aged greater than or equal to 18 years', 'N/A')}\n"
                )

                user_question = st.text_input("Ask a Question about the selected County:", key="county_question")
                if st.button("Submit County Query", key="county_submit"):
                    if openai_api_key and user_question:
                        # Call the new maps_qa function for counties
                        result = maps_qa(openai_api_key, user_question, extra_context, plan_list, state_abbr, epa_region)
                        st.write(result)
                    else:
                        st.error("Please provide both an API key and a question.")
            else:
                st.info("Click on a county to view details.")
        
        with cols_county[2]:
            legend_html = generate_legend_html(REGION_COLORS)
            st.markdown(legend_html, unsafe_allow_html=True)

import glob

# ------------------------------------------------------------------------------
# TAB 2: CLIMATE PLAN REPORT GENERATOR
# ------------------------------------------------------------------------------
with summary_tab:
    st.markdown("""
    # Climate Plan Report Generator

    This tool enables users to generate structured, high-level reports from local Climate Action Plans. Using a set of predefined analytical questions, the system extracts key insights related to climate adaptation, mitigation, and resilience strategies.

    Users can either:
    - **Upload a new PDF document** to generate a report using AI
    - **Select an existing report** generated from a previous upload

    These reports are designed to support planning, policy development, and comparative research by offering a concise overview of local climate strategies.

    To use this tool:
    1. Enter your **OpenAI API key**
    2. Upload a **PDF climate action plan** or select from existing reports
    3. Click **Generate** to create a structured report

    The more complete and detailed the source document, the more comprehensive the resulting report will be.
    """)


    # Two options: Load from existing reports or upload a new plan
    report_mode = st.radio(
        "Select how you'd like to generate or view a report:",
        ("Load from existing reports", "Upload a new plan"),
        key="report_mode_selector"
    )

    import re

    if report_mode == "Load from existing reports":
        # List available markdown summaries
        summary_files = sorted(glob.glob("CAPS_Summaries/*_Summary.md"))

        if summary_files:
            # Create a display-friendly name and map it to the real filename
            display_name_to_file = {
                re.sub(r"_Summary\.md$", "", os.path.basename(f)).replace("_", " "): f
                for f in summary_files
            }

            selected_display_name = st.selectbox("Select a report to view:", list(display_name_to_file.keys()))

            if selected_display_name:
                file_path = display_name_to_file[selected_display_name]
                try:
                    with open(file_path, "r") as f:
                        summary_md = f.read()
                    with st.expander("Report", expanded=True):
                        st.markdown(summary_md, unsafe_allow_html=False)

                except Exception as e:
                    st.error(f"Could not load the summary: {e}")
        else:
            st.info("No reports found in the CAPS_Summaries folder.")

    elif report_mode == "Upload a new plan":
        uploaded_file = st.file_uploader(
            "Upload a Climate Action Plan in PDF format",
            type="pdf",
            key="upload_file"
        )

        # Set file paths for prompt and questions
        prompt_file_path = "Prompts/summary_tool_system_prompt.md"
        questions_file_path = "Prompts/summary_tool_questions.md"

        if st.button("Generate", key="generate_button"):
            if not openai_api_key:
                st.warning("Please provide your OpenAI API key.")
            elif not uploaded_file:
                st.warning("Please upload a PDF file.")
            else:
                display_placeholder = st.empty()
                with st.spinner("Processing..."):
                    try:
                        # Call the new summary_generation function
                        results = summary_generation(
                            openai_api_key,
                            uploaded_file,
                            questions_file_path,
                            prompt_file_path,
                            display_placeholder
                        )
                        markdown_text = "\n".join(results)

                        # Use the uploaded file's base name for the download file
                        base_name = os.path.splitext(uploaded_file.name)[0]
                        download_file_name = f"{base_name}_Report.md"

                        st.download_button(
                            label="Download Results as Markdown",
                            data=markdown_text,
                            file_name=download_file_name,
                            mime="text/markdown",
                            key="download_button"
                        )
                    except Exception as e:
                        st.error(f"An error occurred: {e}")

# ------------------------------------------------------------------------------
# TAB 3: MULTI-PLAN Q&A
# ------------------------------------------------------------------------------
with multi_plan_qa_tab:
    st.markdown("""
    # Cross-Plan Knowledge Query

    This tool enables users to ask natural language questions across a curated corpus of over 100 local Climate Action and Adaptation Plans from across the United States.

    By leveraging Retrieval-Augmented Generation (RAG), it identifies relevant content from multiple plans and synthesizes a coherent, evidence-based response. This supports comparative research, pattern identification, and discovery of best practices in local climate policy.

    #### Retrieval Methods
    The tool offers two retrieval strategies for generating responses:

    - **Single-Index Retrieval (Efficient Search):**  
    A unified vector store is created from all plans. Relevant content is retrieved globally from this pooled index, offering faster response times and high-relevance results.  
    *Recommended for broad, general questions.*

    - **Per-Document Retrieval (Greedy Search):**  
    Top-matching text segments are retrieved individually from each document before generating a response. This ensures balanced representation across plans but may increase processing time.  
    *Recommended for detailed or comparative queries.*

    #### How to Use:
    1. Enter your **OpenAI API key**
    2. Enter a question (e.g., *"How are communities addressing wildfire risk?"*)
    3. Select a retrieval method
    4. Click **Ask** to generate a synthesized response
    """)

    input_text = st.text_input("Ask a question:", key="multi_plan_input")
    st.markdown("### Search Method")
    search_method = st.radio("Select a search method: ", ["Efficient", "Greedy"])
    if st.button("Ask", key="multi_plan_qa_button"):
        if not openai_api_key:
            st.warning("Please provide your OpenAI API key.")
        elif not input_text:
            st.warning("Please enter a question.")
        else:
            display_placeholder2 = st.empty()
            with st.spinner("Processing..."):
                try:
                    if search_method == "Efficient":
                        # Call multi_plan_qa for the efficient (single vector store) method
                        multi_plan_qa(
                            openai_api_key,
                            input_text,
                            display_placeholder2
                        )
                    elif search_method == "Greedy":
                        # Call multi_plan_qa_multi_vectorstore for the greedy (multiple vector stores) method
                        multi_plan_qa_multi_vectorstore(
                            openai_api_key,
                            input_text,
                            display_placeholder2
                        )
                except Exception as e:
                    st.error(f"An error occurred: {e}")

# ------------------------------------------------------------------------------
# TAB 4: DOCUMENT Q&A TOOL
# ------------------------------------------------------------------------------
with document_qa_tab:
    st.markdown("""
    # Single-Plan Query Assistant

    This tool enables users to interact directly with a single Climate Action or Adaptation Plan using natural language queries. Responses are grounded exclusively in the selected document, allowing for precise and document-specific analysis.

    Users can either:
    - **Select an existing processed plan** from the document library  
    - **Upload a new PDF document**, which is temporarily processed for querying

    The assistant leverages Retrieval-Augmented Generation (RAG), combining semantic search with large language models to extract relevant information from the plan in response to user queries.

    This tool is particularly useful for:
    - Reviewing the contents of individual plans in depth
    - Extracting information on specific themes, strategies, or metrics
    - Supporting local planning processes or academic analysis

    #### How to Use:
    1. Enter your **OpenAI API key**
    2. Select a plan from the library or upload a new document
    3. Ask questions in natural language (e.g., *“What are the key adaptation strategies?”*)
    4. The assistant will return grounded answers based only on the selected plan

    Previous interactions are shown in a conversational thread to support iterative exploration.
    """)

    # Get list of existing vector store documents
    vectorstore_documents = list_vector_store_documents()

    # Option to upload a new plan or select from existing vector stores
    focus_option = st.radio(
        "Choose a focus plan:",
        ("Select from existing vector stores", "Upload a new plan"),
        key="focus_option_qa"
    )

    if focus_option == "Upload a new plan":
        focus_uploaded_file = st.file_uploader(
            "Upload a Climate Action Plan",
            type="pdf",
            key="focus_upload_qa"
        )
        focus_input = focus_uploaded_file if focus_uploaded_file else None
    else:
        selected_focus_plan = st.selectbox(
            "Select a focus plan:",
            vectorstore_documents,
            key="select_focus_plan_qa"
        )
        focus_input = os.path.join(
            "Individual_All_Vectorstores",
            f"{selected_focus_plan.replace(' Summary', '_Summary')}_vectorstore"
        )

    # Display previous conversation messages
    if "chat_history" in st.session_state:
        for message in st.session_state.chat_history:
            role = "assistant" if isinstance(message, AIMessage) else "user"
            st.chat_message(role).markdown(message.content)

    user_input = st.chat_input("Ask a question")
    if user_input:
        if "chat_history" not in st.session_state:
            st.session_state.chat_history = []
        if openai_api_key and focus_input:
            st.session_state.chat_history.append(HumanMessage(content=user_input))
            st.chat_message("user").markdown(user_input)
            with st.spinner("Processing..."):
                # Call the new document_qa function
                answer = document_qa(openai_api_key, focus_input, user_input)
                st.session_state.chat_history.append(AIMessage(content=answer))
                st.chat_message("assistant").markdown(answer)
        else:
            st.warning("Please provide your OpenAI API key and select a focus plan.")

# ------------------------------------------------------------------------------
# TAB 5: PLAN COMPARISON TOOL
# ------------------------------------------------------------------------------
with plan_comparison_tab:
    st.markdown("""
    # Comparative Policy Analysis

    This tool supports the side-by-side comparison of climate action and adaptation plans. By enabling users to ask structured natural language questions across a **focus plan** and one or more **comparison plans**, it provides insight into differences in strategy, scope, priorities, and language.

    Users can:
    - **Select or upload a focus plan** for detailed analysis
    - **Compare it against other selected or uploaded plans**
    - **Ask targeted comparison questions** to surface similarities and divergences

    Two analysis modes are available:
    - **Standard (OpenAI):** Uses Retrieval-Augmented Generation (RAG) to ground responses in the most relevant sections of each plan
    - **Long-Context (Anthropic):** Uses a long-context model to ingest full plan summaries without retrieval, enabling holistic, document-wide comparison

    This tool is well-suited for:
    - Benchmarking policy approaches across jurisdictions
    - Evaluating alignment with regional or national goals
    - Identifying gaps or innovations in specific plans

    #### How to Use:
    1. Enter your **OpenAI API key** (and **Anthropic key**, if using long-context mode)
    2. Select or upload a **focus plan**
    3. Select or upload one or more **comparison plans**
    4. Enter a comparison question (e.g., *“How do these cities address social vulnerability?”*)
    5. Select an analysis method and click **Compare**

    The system will generate a structured comparative response using the selected documents.
    """)


    # Get list of existing vector store documents for plans
    vectorstore_documents = list_vector_store_documents()

    # Option to upload a new plan or select from existing vector stores for focus
    focus_option = st.radio(
        "Choose a focus plan:",
        ("Select from existing vector stores", "Upload a new plan"),
        key="focus_option"
    )
    if focus_option == "Upload a new plan":
        focus_uploaded_file = st.file_uploader(
            "Upload a Climate Action Plan to compare",
            type="pdf",
            key="focus_upload"
        )
        focus_input = focus_uploaded_file if focus_uploaded_file is not None else None
    else:
        selected_focus_plan = st.selectbox(
            "Select a focus plan:",
            vectorstore_documents,
            key="select_focus_plan"
        )
        focus_input = os.path.join(
            "Individual_All_Vectorstores",
            f"{selected_focus_plan.replace(' Summary', '_Summary')}_vectorstore"
        )

    # Option to upload comparison documents or select from existing vector stores
    comparison_option = st.radio(
        "Choose comparison documents:",
        ("Select from existing vector stores", "Upload new documents"),
        key="comparison_option"
    )
    if comparison_option == "Upload new documents":
        comparison_files = st.file_uploader(
            "Upload comparison documents",
            type="pdf",
            accept_multiple_files=True,
            key="comparison_files"
        )
        comparison_inputs = comparison_files
    else:
        selected_comparison_plans = st.multiselect(
            "Select comparison documents:",
            vectorstore_documents,
            key="select_comparison_plans"
        )
        comparison_inputs = [
            os.path.join(
                "Individual_All_Vectorstores",
                f"{doc.replace(' Summary', '_Summary')}_vectorstore"
            ) for doc in selected_comparison_plans
        ]
    
    st.markdown("### Model")
    
    search_method = st.radio("Select an approach: ", ["Standard (OpenAI)", "Long Context Model (Anthropic)"])

    input_text = st.text_input("Ask a comparison question:", key="comparison_input")

    if st.button("Compare", key="compare_button"):
        if not openai_api_key:
            st.warning("Please provide your OpenAI API key.")
        elif not input_text:
            st.warning("Please enter a comparison question.")
        elif not focus_input:
            st.warning("Please provide a focus plan.")
        elif not comparison_inputs:
            st.warning("Please provide comparison documents.")
        else:
            display_placeholder3 = st.empty()
            with st.spinner("Processing..."):
                try:
                    if search_method == "Standard (OpenAI)":
                        # Call the new comparison_qa function (formerly process_one_to_many_query)
                        comparison_qa(
                            openai_api_key,
                            focus_input,
                            comparison_inputs,
                            input_text,
                            display_placeholder3
                        )
                    elif search_method == "Long Context Model (Anthropic)":
                        # For long-context, pass the focus plan and comparison inputs directly
                        comparison_qa_long_context(
                            openai_api_key,
                            anthropic_api_key,
                            input_text,
                            focus_input,
                            comparison_inputs,
                            display_placeholder3
                        )
                except Exception as e:
                    st.error(f"An error occurred: {e}")

# ------------------------------------------------------------------------------
# TAB 6: PLAN INSIGHTS
# ------------------------------------------------------------------------------
with plan_insights_tab:
    st.markdown("""
    # Dataset Overview & Insights

    This table presents structured insights extracted from individual Climate Action and Adaptation Plans included in the corpus. Each row represents a unique plan, with columns corresponding to answers generated in response to a standardized set of analytical questions.

    The dataset provides a high-level view of how different cities and counties address climate risks, adaptation strategies, mitigation efforts, and resilience planning. It enables researchers, planners, and policymakers to explore trends, compare responses, and identify areas of innovation or inconsistency across jurisdictions.

    #### Features:
    - Each row is linked to a single plan (city, state, year, and plan type)
    - Columns include AI-generated answers to predefined thematic questions
    - Filter, sort, and search to explore patterns across plans

    This view supports rapid exploration of the corpus and can serve as a foundation for deeper analysis using the querying, comparison, or mapping tools in this platform.
    """)

    try:
        # Load the CSV file
        df_plans = pd.read_csv("climate_action_plans_dataset.csv")
        df_plans.columns = [
            "City Name", "State Name", "Year", "Plan Type", 
            "Top Threats Identified", "Adaptation Measures", "Mitigation Measures", "Resilience Measures"
        ]

        # Prevent "2025" from being displayed as "2,025"
        df_plans["Year"] = df_plans["Year"].astype(str)

        st.dataframe(df_plans)
    except Exception as e:
        st.error(f"Error loading CSV: {e}")

st.markdown("""
<hr style="margin-top: 4rem; margin-bottom: 1rem;">

<div style='font-size: 0.85rem; color: var(--text-color);'>

**Disclaimer**
              
This platform uses artificial intelligence (AI) to support the exploration and analysis of local climate action and adaptation plans. While the system can surface useful insights, responses may occasionally be incomplete or inaccurate. Users are encouraged to verify outputs independently, particularly when using this tool for planning, decision-making, or academic research.  

Please **do not share any personal, proprietary, sensitive, or protected information** when interacting with this tool. Queries and uploaded content are processed by third-party AI services (such as OpenAI), and may be stored or used to improve their models. This tool is intended solely for public, non-confidential climate policy documents. 

The underlying corpus does not represent a complete set of climate action plans in the United States. Some localities with existing plans may not be included. This tool is intended for exploratory purposes and should not be interpreted as providing legal or scientific advice.

**Acknowledgments**  
This project was developed as a collaboration between **Professor J.B. Ruhl's lab** and the **Vanderbilt Data Science Institute (DSI)**.  
We also acknowledge valuable contributions from **Ethan Thorpe** and **Mariah Caballero**.  
Development was led by **Umang Chaudhry**, *Senior Data Scientist at the DSI*, with support from student collaborators **Harmony Wang**, **Isabella Urquia**, and **Xuanxuan Chen**.  
</div>
""", unsafe_allow_html=True)