Add tool to read Philly crash data

Add RAG to pull in Philly public crash data with Pandas

app.py

@@ -5,8 +5,8 @@ import getpass
 from smolagents import GradioUI, CodeAgent, HfApiModel
 
 # Import our custom tools from their modules
-from tools import DuckDuckGoSearchTool, WeatherInfoTool, HubStatsTool
-from retriever import load_guest_dataset
+from tools import DuckDuckGoSearchTool, WeatherInfoTool, HubStatsTool, CrashInfoRetrieverTool
+from retriever import load_guest_dataset, load_crash_data
 
 #Set up in HuggingFace Space Secrets
 #os.environ["HF_TOKEN"] = getpass.getpass("Enter HF token:")
@@ -31,9 +31,11 @@ hub_stats_tool = HubStatsTool()
 # Load the guest dataset and initialize the guest info tool
 guest_info_tool = load_guest_dataset()
 
+crash_info_tool = load_crash_data()
+
 # Create Alfred with all the tools
 alfred = CodeAgent(
-    tools=[guest_info_tool, weather_info_tool, hub_stats_tool, search_tool], 
+    tools=[guest_info_tool, weather_info_tool, hub_stats_tool, search_tool, crash_info_tool], 
     model=model,
     add_base_tools=True,  # Add any additional base tools
     planning_interval=3   # Enable planning every 3 steps

retriever.py

@@ -2,7 +2,8 @@ from smolagents import Tool
 from langchain_community.retrievers import BM25Retriever
 from langchain.docstore.document import Document
 import datasets
-
+import pandas as pd
+import requests
 
 class GuestInfoRetrieverTool(Tool):
     name = "guest_info_retriever"
@@ -50,4 +51,27 @@ 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"
+    response = requests.get(url)
+    with open("philly_crashes.csv", "wb") as f:
+        f.write(response.content)
+
+    df = pd.read_csv("philly_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')}"
+        ])
+        docs.append(Document(page_content=content, metadata={"id": row.get("id", "unknown")}))
+    
+    return docs
 

tools.py

@@ -10,7 +10,6 @@ from huggingface_hub import list_models
 
 class GuestInfoRetrieverTool(Tool):
     name = "guest_info_retriever"
-    priority = 10 # Higher priority, prefer personal this dataset knowledge and augment with internet search
     description = "Retrieves detailed information about gala guests based on their name or relation."
     inputs = {
         "query": {
@@ -32,7 +31,7 @@ class GuestInfoRetrieverTool(Tool):
             return "No matching guest information found."
 
 # Initialize the tool
-# guest_info_tool = GuestInfoRetrieverTool(docs) # no, not here, in app.py
+# guest_info_tool = GuestInfoRetrieverTool(docs)
 
 
 class WeatherInfoTool(Tool):
@@ -81,3 +80,24 @@ class HubStatsTool(Tool):
         except Exception as e:
             return f"Error fetching models for {author}: {str(e)}"
 
+
+class CrashInfoRetrieverTool(Tool):
+    name = "crash_info_tool"
+    description = "Retrieves traffic crash details in Philadelphia by keyword (e.g. location, date, weather)."
+    inputs = {
+        "query": {
+            "type": "string",
+            "description": "Any keyword like a date, street name, or crash type."
+        }
+    }
+    output_type = "string"
+
+    def __init__(self, docs):
+        self.retriever = BM25Retriever.from_documents(docs)
+
+    def forward(self, query: str):
+        results = self.retriever.get_relevant_documents(query)
+        if results:
+            return "\n\n".join([doc.page_content for doc in results[:3]])
+        else:
+            return "No crash data found matching your query."