Initial dashboard deployment

Dockerfile

@@ -0,0 +1,35 @@
+FROM python:3.11-slim
+
+# System deps for GDAL/geopandas
+RUN apt-get update && apt-get install -y --no-install-recommends \
+    libgdal-dev \
+    gdal-bin \
+    libgeos-dev \
+    libproj-dev \
+    && rm -rf /var/lib/apt/lists/*
+
+WORKDIR /app
+
+# Install Python deps
+COPY requirements.txt .
+RUN pip install --no-cache-dir -r requirements.txt
+
+# Copy application files
+COPY data_pipeline.py .
+COPY dashboard_helpers.py .
+COPY app.py .
+
+# Pre-download databases and bake the data
+RUN python -c "import leafmap; \
+    leafmap.download_file( \
+        'https://opengeos.org/data/duckdb/nyc_data.db.zip', \
+        unzip=True, \
+        overwrite=True \
+    )"
+RUN python data_pipeline.py
+
+# Expose the port HF Spaces expects
+EXPOSE 7860
+
+# Launch Solara
+CMD ["solara", "run", "app.py", "--host=0.0.0.0", "--port=7860"]

README.md

@@ -1,11 +1,33 @@
 ---
-title: HuggingFaceNYCTaxi
-emoji: 🏆
+title: NYC Taxi Mobility Equity Dashboard
 colorFrom: blue
-colorTo: gray
+colorTo: green
 sdk: docker
 pinned: false
 license: mit
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+# NYC Taxi Mobility Equity Dashboard
+
+An interactive Solara dashboard analyzing whether **FHV** and **Yellow Taxi** 
+services are equitably distributed across NYC neighborhoods relative to 
+demographic baselines.
+
+## Features
+
+- Pickup & drop-off choropleth maps for FHV and Yellow Taxi
+- Diverging demographic maps showing deviation from citywide baseline
+- Top-10 stats tables split by service and direction
+- Month selector to explore Jan-Mar 2025 data
+
+## Data Sources
+
+- NYC Taxi Zone geometry: data.source.coop
+- NYC Census Blocks: opengeos.org
+- Trip Records: NYC TLC (Jan-Mar 2025)
+
+## Architecture
+
+- `data_pipeline.py`: one-time data processing, writes to `processed_dashboard.db`
+- `dashboard_helpers.py`: reusable map and stats functions
+- `app.py`: Solara UI components and layout

app.py

@@ -0,0 +1,86 @@
+import solara
+import duckdb
+from dashboard_helpers import build_trip_map, build_demo_map, build_stats
+
+# Connect to pre-baked database in read-only mode
+con = duckdb.connect('processed_dashboard.db', read_only=True)
+con.install_extension('spatial')
+con.load_extension('spatial')
+
+# Read baselines
+baseline_df = con.sql("SELECT * FROM city_baselines").df()
+baseline_white = float(baseline_df['baseline_white_pct'].iloc[0]) / 100.0
+baseline_black = float(baseline_df['baseline_black_pct'].iloc[0]) / 100.0
+
+# Reactive state
+selected_month = solara.reactive('Feb2025')
+
+
+@solara.component
+def TaxiMap(service, metric, cmap, title):
+    month = selected_month.value
+    m = build_trip_map(con, service, month, metric, cmap, title)
+    with solara.Card(title=f'{title} ({month})'):
+        solara.display(m)
+
+
+@solara.component
+def DemographicMap(column, baseline_val, title):
+    m = build_demo_map(con, column, baseline_val, title)
+    with solara.Card(title=title):
+        solara.display(m)
+
+
+@solara.component
+def ServiceStatsCard(service, cmap_pu, cmap_do):
+    month = selected_month.value
+    pu_head, pu_table = build_stats(con, service, month, 'pu', cmap_pu, 'Pickups')
+    do_head, do_table = build_stats(con, service, month, 'do', cmap_do, 'Drop-offs')
+    with solara.Card():
+        with solara.Row():
+            with solara.Column():
+                solara.Markdown(pu_head, style={'text-align': 'center'})
+                solara.display(pu_table)
+            with solara.Column():
+                solara.Markdown(do_head, style={'text-align': 'center'})
+                solara.display(do_table)
+
+
+@solara.component
+def Page():
+    solara.Title('NYC Taxi Mobility Equity Dashboard')
+
+    with solara.Column():
+        with solara.Column(align='center'):
+            solara.Markdown('# NYC Taxi Mobility Equity Dashboard',
+                            style={'text-align': 'center'})
+            solara.Markdown(
+                '<div style="text-align:center;">This interactive dashboard analyzes whether '
+                '<b>FHV</b> and <b>Yellow Taxi</b> services are over- or under-represented '
+                'in areas with different demographic compositions relative to the NYC baseline.</div>'
+            )
+            solara.Select(
+                label='Select Month',
+                value=selected_month,
+                values=['Jan2025', 'Feb2025', 'Mar2025']
+            )
+
+        # Trip maps grid
+        with solara.GridFixed(columns=2):
+            TaxiMap('FHV', 'pu', 'Blues', 'FHV Pickups')
+            TaxiMap('Yellow', 'pu', 'Blues', 'Yellow Pickups')
+            TaxiMap('FHV', 'do', 'Greens', 'FHV Drop-offs')
+            TaxiMap('Yellow', 'do', 'Greens', 'Yellow Drop-offs')
+
+        # Demographic baseline maps
+        with solara.GridFixed(columns=2):
+            DemographicMap('white_pct', baseline_white * 100,
+                           f'White Pop. Deviation from Baseline ({baseline_white*100:.1f}%)')
+            DemographicMap('black_pct', baseline_black * 100,
+                           f'Black Pop. Deviation from Baseline ({baseline_black*100:.1f}%)')
+
+        # Stats tables
+        with solara.Column(align='center'):
+            with solara.GridFixed(columns=2):
+                ServiceStatsCard('FHV', 'Blues', 'Greens')
+                ServiceStatsCard('Yellow', 'YlOrBr', 'YlOrBr')

dashboard_helpers.py

@@ -0,0 +1,61 @@
+import geopandas as gpd
+import leafmap
+
+def build_trip_map(con, service, month, metric, cmap, legend_prefix):
+    if metric == 'pu':
+        table, col = 'trip_counts_pu', 'trips_pu'
+    else:
+        table, col = 'trip_counts_do', 'trips_do'
+    df = con.sql(f"""
+        SELECT
+            tz.zone,
+            t.{col} AS trips,
+            ST_AsText(ST_Transform(tz.geometry, 'EPSG:26918', 'OGC:CRS84')) AS geometry
+        FROM {table} AS t
+        JOIN taxi_zones_utm AS tz ON t.LocationID = tz.LocationID
+        WHERE t.service = '{service}' AND t.month = '{month}'
+    """).df()
+    gdf = gpd.GeoDataFrame(df, geometry=gpd.GeoSeries.from_wkt(df['geometry']), crs='EPSG:4326')
+    m = leafmap.Map(center=[40.7, -73.9], zoom=10, draw_control=False)
+    m.layout.height = '400px'
+    m.add_basemap('CartoDB.DarkMatter')
+    m.add_data(gdf, column='trips', cmap=cmap,
+               legend_title=f'{legend_prefix} ({month})')
+    return m
+
+
+def build_demo_map(con, column, baseline_val, legend_title):
+    df = con.sql(f"""
+        SELECT
+            zd.TaxiZone AS zone,
+            ROUND(zd.{column} - {baseline_val}, 1) AS deviation,
+            ST_AsText(ST_Transform(tz.geometry, 'EPSG:26918', 'OGC:CRS84')) AS geometry
+        FROM zone_demographics AS zd
+        JOIN taxi_zones_utm AS tz ON zd.LocationID = tz.LocationID
+        WHERE zd.TotalPop > 0
+    """).df()
+    gdf = gpd.GeoDataFrame(df, geometry=gpd.GeoSeries.from_wkt(df['geometry']), crs='EPSG:4326')
+    m = leafmap.Map(center=[40.7, -73.9], zoom=10, draw_control=False)
+    m.layout.height = '400px'
+    m.add_basemap('CartoDB.DarkMatter')
+    m.add_data(gdf, column='deviation', cmap='RdYlBu', legend_title=legend_title)
+    return m
+
+
+def build_stats(con, service, month, metric, cmap, label):
+    if metric == 'pu':
+        table, col, alias = 'trip_counts_pu', 'trips_pu', 'Pickups'
+    else:
+        table, col, alias = 'trip_counts_do', 'trips_do', 'Dropoffs'
+    df = con.sql(f"""
+        SELECT
+            tz.zone AS Neighborhood,
+            t.{col} AS {alias}
+        FROM {table} t
+        JOIN taxi_zones_utm tz ON t.LocationID = tz.LocationID
+        WHERE t.service = '{service}' AND t.month = '{month}'
+        ORDER BY t.{col} DESC LIMIT 10
+    """).df()
+    header = f'### {service} Top 10 {label} ({month})'
+    styled = df.style.background_gradient(cmap=cmap, subset=[alias]).hide(axis='index')
+    return header, styled

data_pipeline.py

@@ -0,0 +1,146 @@
+import os
+import pathlib
+import duckdb
+import pandas as pd
+import leafmap
+import requests
+
+# DuckDB connection & extensions
+con = duckdb.connect('processed_dashboard.db')
+con.install_extension('httpfs')
+con.load_extension('httpfs')
+con.install_extension('spatial')
+con.load_extension('spatial')
+
+# Taxi-zone geometry
+TAXI_ZONES_URL = 'https://data.source.coop/cholmes/nyc-taxi-zones/taxi_zones.parquet'
+con.sql(f"CREATE OR REPLACE VIEW taxi_zones AS SELECT * FROM '{TAXI_ZONES_URL}'")
+
+# NYC census blocks
+DB_PATH = pathlib.Path('nyc_data.db')
+if not DB_PATH.exists():
+    leafmap.download_file(
+        'https://opengeos.org/data/duckdb/nyc_data.db.zip',
+        unzip=True,
+        overwrite=True,
+    )
+
+con.execute("ATTACH 'nyc_data.db' AS nyc_data (READ_ONLY)")
+
+# Reproject taxi zones to UTM
+con.sql("""
+    CREATE OR REPLACE TABLE taxi_zones_utm AS
+    SELECT * EXCLUDE (geometry),
+        ST_Transform(geometry, 'EPSG:2263', 'EPSG:26918') AS geometry
+    FROM taxi_zones
+""")
+
+# Zone demographics via spatial join
+con.sql("""
+    CREATE OR REPLACE TABLE zone_demographics AS
+    SELECT
+        tz.LocationID,
+        tz.zone AS TaxiZone,
+        tz.borough AS Borough,
+        SUM(cb.popn_total) AS TotalPop,
+        SUM(cb.popn_white) AS WhitePop,
+        SUM(cb.popn_black) AS BlackPop,
+        100.0 * SUM(cb.popn_white) / SUM(cb.popn_total) AS white_pct,
+        100.0 * SUM(cb.popn_black) / SUM(cb.popn_total) AS black_pct
+    FROM nyc_data.nyc_census_blocks AS cb
+    JOIN taxi_zones_utm AS tz ON ST_Intersects(tz.geometry, cb.geom)
+    GROUP BY tz.LocationID, tz.zone, tz.borough
+""")
+
+# Citywide baselines
+baseline_df = con.sql("""
+    SELECT
+        ROUND(100.0 * SUM(popn_white) / SUM(popn_total), 2) AS baseline_white_pct,
+        ROUND(100.0 * SUM(popn_black) / SUM(popn_total), 2) AS baseline_black_pct
+    FROM nyc_data.nyc_census_blocks
+""").df()
+
+baseline_white = float(baseline_df['baseline_white_pct'].iloc[0]) / 100.0
+baseline_black = float(baseline_df['baseline_black_pct'].iloc[0]) / 100.0
+
+# Save baselines into the db for app.py to read later
+con.sql(f"""
+    CREATE OR REPLACE TABLE city_baselines AS
+    SELECT
+        {baseline_white * 100} AS baseline_white_pct,
+        {baseline_black * 100} AS baseline_black_pct
+""")
+
+# Trip data
+pu_field = {'FHV': 'PUlocationID', 'Yellow': 'PULocationID'}
+do_field = {'FHV': 'DOlocationID', 'Yellow': 'DOLocationID'}
+trip_urls = {
+    'FHV_Jan2025': 'https://d37ci6vzurychx.cloudfront.net/trip-data/fhv_tripdata_2025-01.parquet',
+    'FHV_Feb2025': 'https://d37ci6vzurychx.cloudfront.net/trip-data/fhv_tripdata_2025-02.parquet',
+    'FHV_Mar2025': 'https://d37ci6vzurychx.cloudfront.net/trip-data/fhv_tripdata_2025-03.parquet',
+    'Yellow_Jan2025': 'https://d37ci6vzurychx.cloudfront.net/trip-data/yellow_tripdata_2025-01.parquet',
+    'Yellow_Feb2025': 'https://d37ci6vzurychx.cloudfront.net/trip-data/yellow_tripdata_2025-02.parquet',
+    'Yellow_Mar2025': 'https://d37ci6vzurychx.cloudfront.net/trip-data/yellow_tripdata_2025-03.parquet',
+}
+
+con.execute("CREATE OR REPLACE TABLE trip_counts_pu (service VARCHAR, month VARCHAR, LocationID INTEGER, trips_pu BIGINT)")
+con.execute("CREATE OR REPLACE TABLE trip_counts_do (service VARCHAR, month VARCHAR, LocationID INTEGER, trips_do BIGINT)")
+
+DATA_DIR = 'trip_data'
+os.makedirs(DATA_DIR, exist_ok=True)
+
+for key, url in trip_urls.items():
+    local_path = os.path.join(DATA_DIR, f'{key}.parquet')
+    if not os.path.exists(local_path):
+        print(f'Downloading {key}...')
+        r = requests.get(url, timeout=120)
+        r.raise_for_status()
+        with open(local_path, 'wb') as f:
+            f.write(r.content)
+    else:
+        print(f'Already exists: {local_path}')
+    service, month = key.split('_')
+    pu, do = pu_field[service], do_field[service]
+    con.sql(f"""
+        INSERT INTO trip_counts_pu
+        SELECT '{service}', '{month}', CAST({pu} AS INTEGER), COUNT(*)
+        FROM '{local_path}'
+        WHERE {pu} IS NOT NULL AND CAST({pu} AS INTEGER) NOT IN (0, 264, 265)
+        GROUP BY {pu}
+    """)
+    con.sql(f"""
+        INSERT INTO trip_counts_do
+        SELECT '{service}', '{month}', CAST({do} AS INTEGER), COUNT(*)
+        FROM '{local_path}'
+        WHERE {do} IS NOT NULL AND CAST({do} AS INTEGER) NOT IN (0, 264, 265)
+        GROUP BY {do}
+    """)
+
+# Representative ratios
+rr_pu_df = con.sql(f"""
+    SELECT
+        tp.service, tp.month,
+        SUM(tp.trips_pu * zd.white_pct) * 1.0 / SUM(tp.trips_pu) / {baseline_white*100} AS RR_white_PU,
+        SUM(tp.trips_pu * zd.black_pct) * 1.0 / SUM(tp.trips_pu) / {baseline_black*100} AS RR_black_PU
+    FROM trip_counts_pu AS tp
+    JOIN zone_demographics AS zd ON tp.LocationID = zd.LocationID
+    WHERE zd.TotalPop > 0
+    GROUP BY tp.service, tp.month
+""").df()
+
+rr_do_df = con.sql(f"""
+    SELECT
+        td.service, td.month,
+        SUM(td.trips_do * zd.white_pct) * 1.0 / SUM(td.trips_do) / {baseline_white*100} AS RR_white_DO,
+        SUM(td.trips_do * zd.black_pct) * 1.0 / SUM(td.trips_do) / {baseline_black*100} AS RR_black_DO
+    FROM trip_counts_do AS td
+    JOIN zone_demographics AS zd ON td.LocationID = zd.LocationID
+    WHERE zd.TotalPop > 0
+    GROUP BY td.service, td.month
+""").df()
+
+rr_combined = pd.merge(rr_pu_df, rr_do_df, on=['service', 'month'], how='outer')
+con.sql("CREATE OR REPLACE TABLE rr_combined AS SELECT * FROM rr_combined")
+
+print('Pipeline complete. processed_dashboard.db is ready.')
+con.close()