Add functions for travel time calculation

artificial-intelligence/T2_2025/Mental Health and Smart City Chatbot/Housing chatbot/.env

@@ -0,0 +1,2 @@
+GOOGLE_API_KEY="YOUR_GEMINI_API_KEY"
+GTFS_API_KEY="YOUR_GTFS_API_KEY"

artificial-intelligence/T2_2025/Mental Health and Smart City Chatbot/Housing chatbot/README.md

@@ -0,0 +1,36 @@
+# HOUSING CHATBOT
+
+## Overview
+
+The housing chatbot aims to recommend the most suitable properties for the users based on their preferences. Besides basic information such as property type (House, Apartment), rental fee, number of bedrooms/bathrooms, etc, it can handle queries that include distance and travelling time from the properties to other places (such as Deakin University), in both driving and public transportation mode, which listing sites such as [realestate.com.au](realestate.com.au) currently do not support. In trimester 2 - 2025, our team has built the chatbot that could recommend properties based on a static crawled listing, it can handle multi-turn conversations where the user query is too vague (e.g. "I want to find a room") so that it could understand more about user's preferences. Based on the preferences, including basic information and distance/time to other places, the chatbot uses a simple ranking model with linear regression to rank the properties. The chatbot is built with LangGraph, with Gemini Flash 2.0 as the core LLM. The UI is build with Chainlit.
+
+**Flowchat**:
+
+![Flow chart of the chatbot](./Housing%20chatbot%20flowchart.png)
+
+## Code structure
+- `graph.py`: Core components of the chatbot compiled with LangGraph.
+- `json_schema.py`: Pre-defined housing schema to help the LLM extract housing entities from user's query.
+- `prompts.py`: System prompt for the LLM to extract housing entities.
+- `utils.py`: Core functions for the components, including data processing, house ranking, and distance calculation.
+- `runner.py`: Simple runner for the chatbot in the local terminal for testing.
+- `app.py`: UI for the chatbot using Chainlit.
+
+## How to run the code
+
+### 1. Setup the environment
+- `Python >=3.11`.
+- `pip install langgraph langchain-google-genai dotenv chainlit geopy scipy openrouteservice pandas folium geopandas`.
+- Add your Gemini API key and GTFS API key in the `.env` file.
+
+### 2. Run the code on terminal (for testing purposes)
+`python runner.py`
+
+### 3. Run on UI
+`chainlit run app.py -w`
+
+## Future development
+Future plans for the chatbot might include: 
+1. Live listing data through API or crawling pipelines; 
+2. Full distance/travelling time for public transportation, including trams and buses (Right now, it only handles walking, driving, and travelling by trains);
+3. Enable for buying use case (it's only built for rental right now).

artificial-intelligence/T2_2025/Mental Health and Smart City Chatbot/Housing chatbot/app.py

@@ -0,0 +1,74 @@
+from typing import cast
+import chainlit as cl
+
+from langchain.schema.runnable.config import RunnableConfig
+from langchain_core.messages import AIMessage, HumanMessage
+
+from langgraph.graph.state import CompiledStateGraph
+from langgraph.types import Command
+
+from graph import create_graph
+
+
+@cl.on_chat_start
+async def on_chat_start():
+    cl.user_session.set("messages", [])
+    agent = await get_chat_agent()
+    cl.user_session.set("agent", agent)
+
+@cl.on_message
+async def on_message(message: cl.Message):
+    agent = cast(CompiledStateGraph, cl.user_session.get("agent"))
+    config = RunnableConfig(
+        configurable={"thread_id": cl.context.session.id}, 
+        callbacks=[cl.LangchainCallbackHandler()],
+    )
+
+    messages = cl.user_session.get("messages")
+    messages.append(HumanMessage(content=message.content))
+    cl.user_session.set("messages", messages)
+
+    interrupt = None
+    response = cl.Message(content="")
+
+    stream = agent.astream(
+        {"messages": messages},
+        config=config,
+        stream_mode=['messages', 'updates'],
+    )
+
+    while stream:
+        async for stream_mode, pack in stream:
+            if stream_mode == 'messages':
+                msg, metadata = pack
+                if (
+                    msg.content
+                    and not isinstance(msg, HumanMessage)
+                    and metadata["langgraph_node"] in ["show_data_output", "end_query"]
+                ):
+                    await response.stream_token(msg.content)
+                stream = None
+
+            else:
+                if '__interrupt__' in pack:
+                    interrupt = pack['__interrupt__'][0]
+                    res = await cl.AskUserMessage(content=str(interrupt.value)).send()
+
+                    cmd = Command(resume=res["output"])
+
+                    stream = agent.astream(
+                        cmd,
+                        config=config,
+                        stream_mode=['messages', 'updates'],
+                    )
+                else:
+                    stream = None
+
+    messages.append(AIMessage(content=response.content))
+    cl.user_session.set("messages", messages)
+
+    await response.send()
+
+async def get_chat_agent() -> CompiledStateGraph:
+    graph = create_graph()
+    return graph