Streamlit Web Application

Overview

The Planwise Streamlit web application provides an interactive, user-friendly interface for our recommendation system. Users can adjust preference sliders, chat with the recommender, view recommended places on a map, and provide feedback through ratings and reviews.

Features

User Authentication

• Login/Signup: Secure user authentication with username and password
• Session Management: Persistent user sessions with OAuth2 token-based authentication
• Profile Management: User preferences and reviews are saved and associated with user accounts

Preference Input Methods

Interactive Sliders

Users can adjust preference sliders (0-5) across 29 categories including:

• Art galleries
• Museums
• Restaurants
• Parks
• Cafes
• And many more

The interface allows users to:

• Select which categories matter to them
• Specify preference strength for each category
• Save preferences for future sessions

Natural Language Chat

Users can express preferences conversationally:

• Type messages like “I love nature and museums but don’t care for nightlife”
• The system extracts preferences from natural language
• Extracted preferences automatically update the sliders

Recommendation Controls

• Location Setting: Users can specify their current location coordinates
• Recommendation Method: Choose between different algorithms:
  • Autoencoder-Based
  • SVD-Based
  • Transfer-Based
  • Embeddings-Based
  • Ensemble (combines all models)
• Routing Profile: Select between “foot-walking” and “driving-car” for route optimization

Recommendation Visualization

List View

Each recommended place shows:

• Name and icon
• Average rating
• User rating count
• Distance from user location
• Recommendation score
• Category tags
• Expandable details section

Map View

• Interactive map showing all recommended places
• Optimized route between locations
• Tooltips with place names
• Color-coded paths

Route Details

• Step-by-step breakdown of the optimized route
• Distance between consecutive stops
• Total travel distance
• Categorized stops for better planning

User Feedback

• Rating System: Users can rate places on a 1-5 scale
• Review Comments: Optional text comments for more detailed feedback
• Edit Functionality: Users can update ratings and reviews they’ve previously submitted
• Persistent Reviews: All feedback is saved to the user’s profile and used to improve future recommendations

Technical Implementation

Application Structure

The Streamlit app is organized into several functional sections:

1. Authentication Flow: Handles login, signup, and session management
2. User Input Collection: Processes preference input through sliders and chat
3. Recommendation Engine Interface: Connects to the underlying models
4. Results Visualization: Displays recommendations and routes
5. Feedback Collection: Manages user ratings and reviews

Key Components

• Session State Management: Uses Streamlit’s session state for persistent user data
• API Integration: Communicates with the FastAPI backend for authentication, preferences, and reviews
• Model Integration: Direct integration with recommendation models for real-time prediction
• Map Rendering: Uses PyDeck for interactive map visualization
• Route Optimization: Integrates with OpenRouteService API for route planning

Usage Guide

Getting Started

1. Create an Account:
   • Navigate to the login tab
   • Click “Sign Up” and enter your credentials
   • Return to login and enter your credentials
2. Set Your Preferences:
   • Use the sliders to indicate your interests
   • Or chat with the recommender to express preferences naturally
3. Generate Recommendations:
   • Set your current location
   • Choose a recommendation method
   • Click “Generate Recommendations”
4. Explore Results:
   • Browse the list of recommended places
   • View them on the interactive map
   • Check the optimized route details
5. Provide Feedback:
   • Rate and review places you’ve visited
   • Edit previous reviews as needed
   • Your feedback improves future recommendations

Running the Application

To run the Streamlit app locally:

cd reco/streamlit
streamlit run app.py

Or with Docker:

docker compose -f docker-compose.dev.yml up

The application will be available at http://localhost:8501.