Switch philly dataset URL

retriever.py

@@ -52,24 +52,24 @@ def load_guest_dataset():
     return GuestInfoRetrieverTool(docs)
 
 def load_crash_data():
-    # Public Philly crash data CSV (small demo file)
-    url = "https://phl.carto.com/api/v2/sql?q=SELECT+*+FROM+vision_zero_crash+LIMIT+5000&format=csv"
+    # Public Philly crash fatality (csv)
+    url = "https://phl.carto.com/api/v2/sql?filename=fatal_crashes&format=csv&skipfields=cartodb_id,the_geom,the_geom_webmercator&q=SELECT%20*,%20ST_Y(the_geom)%20AS%20lat,%20ST_X(the_geom)%20AS%20lng%20FROM%20fatal_crashes"
     response = requests.get(url)
-    with open("philly_crashes.csv", "wb") as f:
+    with open("fatal_crashes.csv", "wb") as f:
         f.write(response.content)
 
-    df = pd.read_csv("philly_crashes.csv")
+    df = pd.read_csv("fatal_crashes.csv")
 
     # Convert to LangChain documents
     docs = []
     for _, row in df.iterrows():
         content = "\n".join([
-            f"Date: {row.get('crash_date', 'N/A')}",
-            f"Location: {row.get('location', 'N/A')}",
-            f"Injury Count: {row.get('injury_count', 'N/A')}",
-            f"Fatalities: {row.get('fatal_count', 'N/A')}",
-            f"Crash Type: {row.get('crash_type', 'N/A')}",
-            f"Weather: {row.get('weather', 'N/A')}"
+            f"Year: {row.get('year', 'N/A')}",
+            f"Street: {row.get('primary_st', 'N/A')}",
+            f"Age: {row.get('age', 'N/A')}",
+            f"Arrest: {row.get('arrest_yes', 'N/A')}",
+            f"Crash Type: {row.get('veh2', 'N/A')}",
+            f"Outcome: {row.get('investigat', 'N/A')}"
         ])
         docs.append(Document(page_content=content, metadata={"id": row.get("id", "unknown")}))