feat: updated streamlit app to use structured filters

app.py

@@ -13,6 +13,7 @@ from qdrant_client import QdrantClient, models
 from streamlit_folium import st_folium
 
 from utils.embedding_utils import ColPaliEmbeddingGenerator
+from utils.llm_utils import extract_filters_from_query
 
 load_dotenv()
 
@@ -123,6 +124,14 @@ st.markdown(
 
 with st.sidebar:
     st.header("Filters")
+    use_auto_filter = st.toggle(
+        "🤖 Enable Auto-Filter",
+        value=False,
+        help="Use AI to automatically extract filters from your query.",
+    )
+
+    if use_auto_filter:
+        st.info("Filters will be extracted from your query automatically.")
 
     min_area = st.slider(
         "Minimum Burn Area (Hectares)",
@@ -131,13 +140,18 @@ with st.sidebar:
         value=0,
         step=10,
         help="Filter results to show only burn scars larger than this value.",
+        disabled=use_auto_filter,
     )
 
     st.subheader("Date Range")
-    min_date_input = st.date_input("Start Date", value=date(2018, 1, 1))
-    max_date_input = st.date_input("End Date", value=date(2021, 12, 31))
+    min_date_input = st.date_input(
+        "Start Date", value=date(2018, 1, 1), disabled=use_auto_filter
+    )
+    max_date_input = st.date_input(
+        "End Date", value=date(2021, 12, 31), disabled=use_auto_filter
+    )
     st.subheader("Spatial Filter")
-    use_spatial = st.checkbox("Filter by Location")
+    use_spatial = st.checkbox("Filter by Location", disabled=use_auto_filter)
 
     if "lat_input_widget" not in st.session_state:
         st.session_state.lat_input_widget = 37.0
@@ -168,14 +182,20 @@ with st.sidebar:
         col_lat, col_lon = st.columns(2)
         with col_lat:
             lat_input = st.number_input(
-                "Latitude", format="%.4f", key="lat_input_widget"
+                "Latitude",
+                format="%.4f",
+                key="lat_input_widget",
+                disabled=use_auto_filter,
             )
         with col_lon:
             lon_input = st.number_input(
-                "Longitude", format="%.4f", key="lon_input_widget"
+                "Longitude",
+                format="%.4f",
+                key="lon_input_widget",
+                disabled=use_auto_filter,
             )
 
-        radius_km = st.slider("Radius (km)", 10, 5000, 230)
+        radius_km = st.slider("Radius (km)", 10, 5000, 230, disabled=use_auto_filter)
 
 user_query = st.text_input(
     "Enter your query:", "Help me find the burn scars that have more than 100 hectares"
@@ -190,46 +210,115 @@ if st.button("Search"):
 
             filter_conditions = []
 
-            if min_area > 0:
-                filter_conditions.append(
-                    models.FieldCondition(
-                        key="burn_area",
-                        range=models.Range(gt=min_area),
+            if use_auto_filter:
+                extracted = extract_filters_from_query(user_query)
+                st.success(f"🤖 Extracted Filters: {extracted}")
+
+                # 1. Burn Area
+                if extracted.min_burn_area:
+                    filter_conditions.append(
+                        models.FieldCondition(
+                            key="burn_area",
+                            range=models.Range(gt=extracted.min_burn_area),
+                        )
                     )
-                )
-
-            start_dt = datetime.combine(min_date_input, datetime.min.time())
-            end_dt = datetime.combine(max_date_input, datetime.max.time())
 
-            filter_conditions.append(
-                models.FieldCondition(
-                    key="acquisition_date",
-                    range=models.DatetimeRange(
-                        gte=start_dt.isoformat(), lte=end_dt.isoformat()
-                    ),
-                )
-            )
+                # 2. Date Range
+                if extracted.date_range:
+                    try:
+                        ex_start = extracted.date_range.start
+                        ex_end = extracted.date_range.end
+
+                        # Defaults if only one is provided
+                        if not ex_start:
+                            ex_start = "2018-01-01"
+                        if not ex_end:
+                            ex_end = "2021-12-31"
+
+                        filter_conditions.append(
+                            models.FieldCondition(
+                                key="acquisition_date",
+                                range=models.DatetimeRange(
+                                    gte=f"{ex_start}T00:00:00", lte=f"{ex_end}T23:59:59"
+                                ),
+                            )
+                        )
+                    except Exception as e:
+                        st.warning(f"Could not parse extracted dates: {e}")
+
+                # 3. Spatial
+                if extracted.location:
+                    lat, lon = get_coordinates(extracted.location)
+                    if lat is not None:
+                        st.info(
+                            f"📍 Geocoded '{extracted.location}' to ({lat:.4f}, {lon:.4f})"
+                        )
+
+                        ex_radius = (
+                            extracted.radius_km if extracted.radius_km else 100.0
+                        )
+
+                        lat_delta = ex_radius / 111.0
+                        lon_delta = ex_radius / (111.0 * np.cos(np.radians(lat)))
+
+                        filter_conditions.append(
+                            models.FieldCondition(
+                                key="centroid_lat",
+                                range=models.Range(
+                                    gte=lat - lat_delta, lte=lat + lat_delta
+                                ),
+                            )
+                        )
+                        filter_conditions.append(
+                            models.FieldCondition(
+                                key="centroid_lon",
+                                range=models.Range(
+                                    gte=lon - lon_delta, lte=lon + lon_delta
+                                ),
+                            )
+                        )
+            else:
+                # Manual Filters
+                if min_area > 0:
+                    filter_conditions.append(
+                        models.FieldCondition(
+                            key="burn_area",
+                            range=models.Range(gt=min_area),
+                        )
+                    )
 
-            if use_spatial:
-                lat_delta = radius_km / 111.0
-                lon_delta = radius_km / (111.0 * np.cos(np.radians(lat_input)))
+                start_dt = datetime.combine(min_date_input, datetime.min.time())
+                end_dt = datetime.combine(max_date_input, datetime.max.time())
 
                 filter_conditions.append(
                     models.FieldCondition(
-                        key="centroid_lat",
-                        range=models.Range(
-                            gte=lat_input - lat_delta, lte=lat_input + lat_delta
+                        key="acquisition_date",
+                        range=models.DatetimeRange(
+                            gte=start_dt.isoformat(), lte=end_dt.isoformat()
                         ),
                     )
                 )
-                filter_conditions.append(
-                    models.FieldCondition(
-                        key="centroid_lon",
-                        range=models.Range(
-                            gte=lon_input - lon_delta, lte=lon_input + lon_delta
-                        ),
+
+                if use_spatial:
+                    lat_delta = radius_km / 111.0
+                    lon_delta = radius_km / (111.0 * np.cos(np.radians(lat_input)))
+
+                    filter_conditions.append(
+                        models.FieldCondition(
+                            key="centroid_lat",
+                            range=models.Range(
+                                gte=lat_input - lat_delta, lte=lat_input + lat_delta
+                            ),
+                        )
+                    )
+                    filter_conditions.append(
+                        models.FieldCondition(
+                            key="centroid_lon",
+                            range=models.Range(
+                                gte=lon_input - lon_delta, lte=lon_input + lon_delta
+                            ),
+                        )
                     )
-                )
 
             query_filter = None
             if filter_conditions: