reviving app

app/app.py

@@ -16,21 +16,20 @@ A data visualization tool built for the Trust for Public Land
 '''
 
 pmtiles = get_pmtiles_url() # generate PMTiles url
-
 with st.sidebar:
     leafmap_choice = st.selectbox("Leafmap module", ['maplibregl','foliumap'])
 
 if leafmap_choice == "maplibregl":
     leafmap = importlib.import_module("leafmap.maplibregl")
-    m = leafmap.Map(style="positron")
+    m = leafmap.Map(style="positron", use_message_queue=True)
+    
 else:
     leafmap = importlib.import_module("leafmap.foliumap")
     m = leafmap.Map(center=[35, -100], zoom=5, layers_control=True, fullscreen_control=True)
 
-basemaps = leafmap.basemaps.keys()
 
 with st.sidebar:
-    b = st.selectbox("Basemap", basemaps)
+    b = st.selectbox("Basemap", basemaps, index= 3)
     m.add_basemap(b)
     st.divider()
 
@@ -52,6 +51,7 @@ with st.sidebar:
         county_choice = 'All'
     st.divider()
 
+legend, position, bg_color, fontsize, shape_type, controls = get_legend(paint, leafmap_choice)
 # get all the ids that correspond to the filter
 gdf = filter_data(tpl_table, state_choice, county_choice, year_range)
 gdf_landvote = filter_data(landvote_table, state_choice, county_choice, year_range)
@@ -101,45 +101,36 @@ prompt = ChatPromptTemplate.from_messages([
 structured_llm = llm.with_structured_output(SQLResponse)
 few_shot_structured_llm = prompt | structured_llm
 
-
-def run_sql(query,paint):
+@st.cache_data(show_spinner = False)
+def run_sql(query, llm_choice):
     """
     Filter data based on an LLM-generated SQL query and return matching IDs.
-
-    Args:
-        query (str): The natural language query to filter the data.
-        color_choice (str): The column used for plotting.
+    Args: query (str): The natural language query to filter the data.
     """
     output = few_shot_structured_llm.invoke(query)
     sql_query = output.sql_query
     explanation =output.explanation
     if not sql_query: # if the chatbot can't generate a SQL query.
-        st.success(explanation)
-        return pd.DataFrame({'fid' : []})
+        return pd.DataFrame({'fid' : []}),'', explanation
     result = con.sql(sql_query).distinct().execute()
-    if result.empty :
+    if result.empty:
         explanation = "This query did not return any results. Please try again with a different query."
-        st.warning(explanation, icon="⚠️")
-        st.caption("SQL Query:")
-        st.code(sql_query,language = "sql") 
         if 'geom' in result.columns:
-            return result.drop('geom',axis = 1)
+            return result.drop('geom',axis = 1), sql_query, explanation 
         else: 
-            return result
-    elif ("fid" and "geom" not in result.columns): 
-        st.write(result)  # if we aren't mapping, just print out the data  
+            return result, sql_query, explanation 
+    return result, sql_query, explanation
+
     
-    with st.popover("Explanation"):
-        st.write(explanation)
-        st.caption("SQL Query:")
-        st.code(sql_query,language = "sql")        
-    return result
 
 with chatbot_container:
     with llm_left_col:
         example_query = "👋 Input query here"
         prompt = st.chat_input(example_query, key="chain", max_chars=300)
-
+    _,log_query_col, _ = st.columns([.001, 5,1], vertical_alignment = "top")
+    with log_query_col:
+        log_queries = st.checkbox("Save query", value = True, help = "Saving your queries helps improve this tool and guide conservation efforts. Your data is stored in a private location. For more details, see 'Why save your queries?' at the bottom of this page.")
+        
 # new container for output so it doesn't mess with the alignment of llm options 
 with st.container():
     if prompt: 
@@ -147,19 +138,60 @@ with st.container():
         try:
             with st.chat_message("assistant"):
                 with st.spinner("Invoking query..."):
-
-                    out = run_sql(prompt,paint)
-                    if ("fid" in out.columns) and (not out.empty):
-                        ids = out['fid'].unique().tolist()
-                        cols = out.columns.tolist()
-                        bounds = out.total_bounds.tolist()
-                        style = tpl_style(ids, paint, pmtiles)
+                    llm_output, sql_query, llm_explanation = run_sql(prompt, llm_choice)
+                    minio_logger(log_queries, prompt, sql_query, llm_explanation, llm_choice, 'query_log.csv', "shared-tpl")
+                    # no sql query generated by chatbot
+                    if sql_query == '':
+                        st.success(llm_explanation)
+                        not_mapping = True
+                    else:    
+                        # sql query generated but no results returned
+                        if llm_output.empty:
+                            st.warning(llm_explanation, icon="⚠️")
+                            st.caption("SQL Query:")
+                            st.code(sql_query, language="sql")
+                            st.stop()
+                            
+                        # output without mapping columns (id, geom)
+                        elif "fid" not in llm_output.columns and "geom" not in llm_output.columns:
+                            st.write(llm_output)
+                            not_mapping = True
+    
+                        # print results 
+                        with st.popover("Explanation"):
+                            st.write(llm_explanation)
+                            if sql_query != '':
+                                st.caption("SQL Query:")
+                                st.code(sql_query,language = "sql") 
+                                
+                    # extract ids, columns, bounds if present
+                    if "fid" in llm_output.columns and not llm_output.empty:
+                        ids = list(set(llm_output['fid'].tolist()))
+                        llm_cols = extract_columns(sql_query)
+                        bounds = llm_output.total_bounds.tolist()
                     else:
-                        ids = []
+                        ids, llm_cols = [], []
+                        not_mapping = True
+
         except Exception as e:
-            error_message = f"ERROR: An unexpected error has occured with the following query:\n\n*{prompt}*\n\n which raised the following error:\n\n{type(e)}: {e}\n"
-            st.warning("Please try again with a different query", icon="⚠️")
-            st.write(error_message)
+            tb_str = traceback.format_exc()  # full multiline traceback string
+            if isinstance(e, openai.BadRequestError):
+                st.error(error_messages["bad_request"](llm_choice, e, tb_str), icon="🚨")
+                
+            elif isinstance(e, openai.RateLimitError):
+                st.error(error_messages["bad_request"](llm_choice, e, tb_str), icon="🚨")
+            
+            elif isinstance(e, openai.APIStatusError):
+                st.error(error_messages["bad_request"](llm_choice, e, tb_str), icon="🚨")
+            
+            elif isinstance(e, openai.InternalServerError):
+                st.error(error_messages["internal_server_error"](llm_choice, e, tb_str), icon="🚨")
+            
+            elif isinstance(e, openai.NotFoundError):
+                st.error(error_messages["internal_server_error"](llm_choice, e, tb_str), icon="🚨")
+            else:
+                prompt = prompt.replace('\n', '')
+                st.error(error_messages["unexpected_llm_error"](prompt, e, tb_str))
             st.stop()
 ##### end of chatbot code 
 
@@ -175,19 +207,60 @@ if 'style' not in locals():
 
 # add pmtiles to map (using user-specified module)
 if leafmap_choice == "maplibregl":
-    m.add_pmtiles(pmtiles, style=style, tooltip=True, fit_bounds=True)
     m.fit_bounds(bounds) 
+    m.add_pmtiles(pmtiles, style=style, 
+                  name="Conservation Almanac",
+                  attribution = "Trust for Public Land (2025)", tooltip=True, 
+                  template = tooltip_template)
+    m.fit_bounds(bounds)
+    if style_choice == "Acquisition Cost":
+        colors, vmin, vmax, orientation, position, label, height, width = get_colorbar(gdf,paint)    
+        m.add_colorbar(palette = colors, vmin=vmin, vmax=vmax, 
+                       orientation = orientation, position = position, transparent = True,
+                       height = height, width = width)
+       
+    else:
+        m.add_legend(title = '', legend_dict = legend, fontsize = fontsize, 
+                 bg_color = bg_color, position = position, 
+                 shape_type = shape_type)
 else: 
-    m.add_pmtiles(pmtiles, style = style, tooltip = True, zoom_to_layer= False)
+    m.add_pmtiles(pmtiles, style=style, 
+                  name="Conservation Almanac by Trust for Public Land",
+                  tooltip=False, zoom_to_layer=True)
+    # add custom tooltip to pmtiles layer
+    for layer in m._children.values():
+        if isinstance(layer, leafmap.PMTilesLayer):
+            pmtiles_layer = layer
+            break
+            
+    pmtiles_layer.add_child(CustomTooltip())
+    style = {'background-color': 'rgba(255, 255, 255, 1)'}
+    m.add_legend(title = '', legend_dict = legend, style = style, 
+                 position = position, shape_type = shape_type, draggable = False)
     m.zoom_to_bounds(bounds)   
+ 
 
 m.to_streamlit()
-
+        
 with st.expander("🔍 View/download data"): # adding data table  
-    if 'out' not in locals():
-        st.dataframe(gdf.drop('geom').head(100).execute(), use_container_width = True)  
+    if 'llm_output' not in locals():
+        # cols = ['tpl_id','municipality','date','geom']
+        group_cols = ['sponsor','program','year']
+        gdf_grouped = (gdf.head(100).execute().groupby(group_cols)
+            .agg({col: ('sum' if col in ['acres','amount'] else 'first') 
+                  for col in gdf.columns if col not in group_cols})).reset_index()
+        cols = ['fid', 'state', 'county','site',  'acres', 'year', 'amount', 'owner', 'owner_type', 
+                'manager', 'manager_type', 'purchase_type', 'easement', 'easement_type', 'access_type',
+                'purpose_type', 'duration_type', 'data_provider', 'data_source', 'source_date', 
+                'data_aggregator', 'comments', 'program_id', 'program', 'sponsor_id', 'sponsor', 'sponsor_type']
+        
+        st.dataframe(gdf_grouped[cols], use_container_width = True)  
     else:
-        st.dataframe(out.drop('geom',axis = 1), use_container_width = True)
+        if ('geom' in llm_output.columns) and (not llm_output.empty):
+            llm_output = llm_output.drop('geom',axis = 1)
+        st.dataframe(llm_output, use_container_width = True)
+
+
 
 public_dollars, private_dollars, total_dollars = tpl_summary(gdf)
 

app/utils.py

@@ -3,6 +3,8 @@ from ibis import _
 from variables import *
 import altair as alt
 import re
+from leafmap.foliumap import PMTilesMapLibreTooltip
+from branca.element import Template
 
 def get_pmtiles_url():
     return client.get_presigned_url(
@@ -13,8 +15,6 @@ def get_pmtiles_url():
     )
     
 def get_counties(state_selection):
-    tpl_table.head().execute()
-
     if state_selection != 'All':
         counties = tpl_table.filter(_.state == state_selection).select('county').distinct().order_by('county').execute()
         counties = ['All'] + counties['county'].tolist()
@@ -120,15 +120,45 @@ def tpl_style(ids, paint, pmtiles):
     }
     return style
 
-def get_legend(paint):
+def get_colorbar(gdf, paint):
+    """
+    Extracts color hex codes and value range (vmin, vmax) from paint
+    to make a color bar. Used for mapping continuous data. 
+    """
+    # numbers = [x for x in paint if isinstance(x, (int, float))]
+    vmin = gdf.amount.min().execute()
+    vmax = gdf.amount.max().execute()
+    # min(numbers), max(numbers),
+    colors = [x for x in paint if isinstance(x, str) and x.startswith('#')]
+    orientation = 'vertical'
+    position = 'bottom-right'
+    label = "Acquisition Cost"
+    height = 3
+    width = .2
+    return colors, vmin, vmax, orientation, position, label, height, width
+    
+
+def get_legend(paint, leafmap_backend, df = None, column = None):
     """
     Generates a legend dictionary with color mapping and formatting adjustments.
     """
-    legend = {cat: color for cat, color in paint['stops']}
-    position, fontsize, bg_color = 'bottom-left', 15, 'white'
-    bg_color = 'rgba(255, 255, 255, 0.6)'
-    fontsize = 12
-    return legend, position, bg_color, fontsize
+    if 'stops' in paint:
+        legend = {cat: color for cat, color in paint['stops']}
+    else:
+        legend = {}
+    if df is not None:
+        if ~df.empty:
+            categories = df[column].to_list() #if we filter out categories, don't show them on the legend 
+            legend = {cat: color for cat, color in legend.items() if str(cat) in categories}
+    position, fontsize, bg_color = 'bottomright', 15, 'white'
+    controls={'navigation': 'bottom-left', 
+              'fullscreen':'bottom-left'}
+    shape_type = 'circle'
+
+    if leafmap_backend == 'maplibregl':
+        position = 'bottom-right'
+    return legend, position, bg_color, fontsize, shape_type, controls 
+
 
 @st.cache_data
 def tpl_summary(_df):
@@ -178,4 +208,76 @@ def chart_time(timeseries, column, paint):
         y = alt.Y('amount:Q'),
             color=alt.Color(column,scale= alt.Scale(domain=domain, range=range_))
     ).properties(height=350)
-    return plt
+    return plt
+    
+class CustomTooltip(PMTilesMapLibreTooltip):
+    _template = Template("""
+    {% macro script(this, kwargs) -%}
+    var maplibre = {{ this._parent.get_name() }}.getMaplibreMap();
+    const popup = new maplibregl.Popup({ closeButton: false, closeOnClick: false });
+
+    maplibre.on('mousemove', function(e) {
+        const features = maplibre.queryRenderedFeatures(e.point);
+        const filtered = features.filter(f => f.properties && f.properties.fid);
+
+        if (filtered.length) {
+            const props = filtered[0].properties;
+            const html = `
+                <div><strong>fid:</strong> ${props.fid || 'N/A'}</div>
+                <div><strong>Site:</strong> ${props.site || 'N/A'}</div>
+                <div><strong>Sponsor:</strong> ${props.sponsor || 'N/A'}</div>
+                <div><strong>Program:</strong> ${props.program || 'N/A'}</div>
+                <div><strong>State:</strong> ${props.state || 'N/A'}</div>
+                <div><strong>County:</strong> ${props.county || 'N/A'}</div>
+                <div><strong>Year:</strong> ${props.year || 'N/A'}</div>
+                <div><strong>Manager:</strong> ${props.manager || 'N/A'}</div>
+                <div>
+                    <strong>Amount:</strong> ${
+                        props.amount
+                            ? `$${parseFloat(props.amount).toLocaleString('en-US', { minimumFractionDigits: 2, maximumFractionDigits: 2 })}`
+                            : 'N/A'
+                    }
+                </div>
+                <div>
+                    <strong>Acres:</strong> ${
+                        props.acres
+                            ? parseFloat(props.acres).toLocaleString('en-US', { minimumFractionDigits: 2, maximumFractionDigits: 2 })
+                            : 'N/A'
+                    }
+                </div>
+            `;
+            popup.setLngLat(e.lngLat).setHTML(html).addTo(maplibre);
+            if (popup._container) {
+                popup._container.style.zIndex = 9999;
+            }
+        } else {
+            popup.remove();
+        }
+    });
+    {% endmacro %}
+    """)
+    
+minio_key = os.getenv("MINIO_KEY")
+if minio_key is None:
+    minio_key = st.secrets["MINIO_KEY"]
+
+minio_secret = os.getenv("MINIO_SECRET")
+if minio_secret is None:
+    minio_secret = st.secrets["MINIO_SECRET"]
+
+def minio_logger(consent, query, sql_query, llm_explanation, llm_choice, filename="query_log.csv", bucket="shared-tpl",
+                 key=minio_key, secret=minio_secret,
+                 endpoint="minio.carlboettiger.info"):
+    mc = minio.Minio(endpoint, key, secret)
+    mc.fget_object(bucket, filename, filename)
+    log = pd.read_csv(filename)
+    timestamp = datetime.datetime.now(datetime.timezone.utc).strftime("%Y-%m-%d %H:%M:%S")
+    if consent:
+        df = pd.DataFrame({"timestamp": [timestamp], "user_query": [query], "llm_sql": [sql_query], "llm_explanation": [llm_explanation], "llm_choice":[llm_choice]})
+
+    # if user opted out, do not store query
+    else:  
+        df = pd.DataFrame({"timestamp": [timestamp], "user_query": ['USER OPTED OUT'], "llm_sql": [''], "llm_explanation": [''], "llm_choice":['']})
+    
+    pd.concat([log,df]).to_csv(filename, index=False, header=True)
+    mc.fput_object(bucket, filename, filename, content_type="text/csv")

app/variables.py

@@ -60,6 +60,7 @@ style_options = {
     "Acquisition Cost":  
             ["interpolate",
                 ['exponential', 1], 
+                # ['linear'], 
                 ["get", "amount"],
                     0,	"#fde725",
                     36000,	"#b4de2c",
@@ -122,6 +123,106 @@ style_choice_columns = {'Manager Type': style_options['Manager Type']['property'
                 'Measure Cost': 'conservation_funds_approved',
              }
 
+basemaps = ['CartoDB.DarkMatter', 'CartoDB.DarkMatterNoLabels', 
+            'CartoDB.DarkMatterOnlyLabels','CartoDB.Positron', 
+            'CartoDB.PositronNoLabels', 'CartoDB.PositronOnlyLabels',
+            'CartoDB.Voyager', 'CartoDB.VoyagerLabelsUnder', 'CartoDB.VoyagerNoLabels',
+            'CartoDB.VoyagerOnlyLabels', 'CyclOSM', 'Esri.NatGeoWorldMap',
+            'Esri.WorldGrayCanvas', 'Esri.WorldPhysical', 'Esri.WorldShadedRelief',
+            'Esri.WorldStreetMap', 'Gaode.Normal', 'Gaode.Satellite',
+            'NASAGIBS.ASTER_GDEM_Greyscale_Shaded_Relief', 'NASAGIBS.BlueMarble', 
+            'NASAGIBS.ModisTerraBands367CR','NASAGIBS.ModisTerraTrueColorCR',
+            'NLCD 2021 CONUS Land Cover', 'OPNVKarte',
+            'OpenStreetMap', 'OpenTopoMap', 'SafeCast',
+            'TopPlusOpen.Color', 'TopPlusOpen.Grey', 'UN.ClearMap',
+            'USGS Hydrography', 'USGS.USImagery', 'USGS.USImageryTopo',
+            'USGS.USTopo']
+
+# legend_labels = {
+#     "Acquisition Cost":  [],
+#     "Manager Type":  
+#                 ['Federal', 'State', 'Local','District','Unknown','Joint','Tribal','Private', 'NGO'],
+#     "Access": 
+#             ['Open Access', 'Closed','Unknown','Restricted'],
+#     "Purpose": 
+#     ['Forestry','Historical','Unknown','Other','Farming','Recreation','Environment','Scenic','RAN'],
+# }
+
+
+#maplibregl tooltip 
+tooltip_cols = ['fid','state','site','sponsor','program','county','year','manager',
+                'amount','acres']
+tooltip_template = "<br>".join([f"{col}: {{{{ {col} }}}}" for col in tooltip_cols])
+
+
+error_messages = {
+    "bad_request": lambda llm, e, tb_str: f"""
+**Error – LLM Unavailable** 
+
+*The LLM you selected `{llm}` is no longer available. Please select a different model.*
+
+**Error Details:**
+`{type(e)}: {e}`
+
+""",
+
+    "internal_server_error": lambda llm, e, tb_str: f"""
+**Error – LLM Temporarily Unavailable**
+
+The LLM you selected `{llm}` is currently down due to maintenance or provider outages. It may remain offline for several hours.
+
+**Please select a different model or try again later.**
+
+**Error Details:**
+`{type(e)}: {e}`
+
+""",
+
+    "unexpected_llm_error": lambda prompt, e, tb_str: f"""
+🐞 **BUG: Unexpected Error in Application**
+
+An error occurred while processing your query:
+
+> "{prompt}"
+
+**Error Details:**
+`{type(e)}: {e}`
+
+Traceback:
+
+```{tb_str}```
+---
+
+🚨 **Help Us Improve!**
+
+Please help us fix this issue by reporting it on GitHub:
+[📄 Report this issue](https://github.com/boettiger-lab/CBN-taskforce/issues)
+
+Include the query you ran and any other relevant details. Thanks!
+""",
+
+    "unexpected_error": lambda e, tb_str: f"""
+🐞 **BUG: Unexpected Error in Application**
+
+
+**Error Details:**
+`{type(e)}: {e}`
+
+Traceback:
+
+```{tb_str}```
+
+---
+
+🚨 **Help Us Improve!**
+
+Please help us fix this issue by reporting it on GitHub:
+[📄 Report this issue](https://github.com/boettiger-lab/CBN-taskforce/issues)
+
+Include the steps you took to get this message and any other details that might help us debug. Thanks!
+"""
+}
+
 from langchain_openai import ChatOpenAI
 import streamlit as st
 from langchain_openai.chat_models.base import BaseChatOpenAI
@@ -143,9 +244,11 @@ llm_options = {
     "deepseek-r1t2-chimera": ChatOpenAI(model = "tngtech/deepseek-r1t2-chimera:free", api_key=openrouter_api, base_url = "https://openrouter.ai/api/v1",  temperature=0),
     "kimi-dev-72b": ChatOpenAI(model = "moonshotai/kimi-dev-72b:free", api_key=openrouter_api, base_url = "https://openrouter.ai/api/v1",  temperature=0),
     "hunyuan-a13b-instruct": ChatOpenAI(model = "tencent/hunyuan-a13b-instruct:free", api_key=openrouter_api, base_url = "https://openrouter.ai/api/v1",  temperature=0),
-    "olmo": ChatOpenAI(model = "olmo", api_key=api_key, base_url = "http://ellm.nrp-nautilus.io/",  temperature=0),
-    "llama3": ChatOpenAI(model = "llama3", api_key=api_key, base_url = "http://ellm.nrp-nautilus.io/",  temperature=0),
-    "qwen3": ChatOpenAI(model = "qwen3", api_key=api_key, base_url = "http://ellm.nrp-nautilus.io/",  temperature=0),
-    "gemma3": ChatOpenAI(model = "gemma3", api_key=api_key, base_url = "http://ellm.nrp-nautilus.io/",  temperature=0),
+    # "deepseek-chat-v3-0324": ChatOpenAI(model = "deepseek/deepseek-chat-v3-0324:free", api_key=openrouter_api, base_url = "https://openrouter.ai/api/v1",  temperature=0),
+    "olmo": ChatOpenAI(model = "olmo", api_key=api_key, base_url = "https://llm.nrp-nautilus.io/",  temperature=0),
+    "llama3": ChatOpenAI(model = "llama3", api_key=api_key, base_url = "https://llm.nrp-nautilus.io/",  temperature=0),
+    # "deepseek-r1": BaseChatOpenAI(model = "deepseek-r1", api_key=api_key, base_url = "https://llm.nrp-nautilus.io/",  temperature=0),
+    "qwen3": ChatOpenAI(model = "qwen3", api_key=api_key, base_url = "https://llm.nrp-nautilus.io/",  temperature=0),
+    "gemma3": ChatOpenAI(model = "gemma3", api_key=api_key, base_url = "https://llm.nrp-nautilus.io/",  temperature=0),
 
-}
+}

requirements.txt

@@ -1,5 +1,5 @@
 git+https://github.com/boettiger-lab/cng-python
-leafmap[maplibre]==0.38.12
+leafmap==0.53.3
 ibis-framework[duckdb]==10.3.1
 lonboard==0.10.4
 altair==5.3.0
@@ -14,5 +14,5 @@ langchain==0.2.17
 langchain-community==0.2.19
 langchain-core==0.2.43
 langchain-openai==0.1.25
-streamlit==1.45.1
+streamlit==1.50.0
 minio==7.2.15