README / README.md
traversaal-ai's picture
Update README.md
6320546 verified
metadata
title: README
emoji: 🏆
colorFrom: yellow
colorTo: purple
sdk: static
pinned: false

Hackathon Overview: Challenger

Traversaal.ai and Qdrant are thrilled to host a cutting-edge hackathon that challenges participants across three progressive levels, each designed to push the boundaries of AI-driven hotel search and recommendation systems. With a focus on enhancing user experiences and leveraging advanced models, the hackathon unfolds in three distinct levels.

Data Overview:

Participants will be working with a comprehensive dataset comprising hotel information across five vibrant cities. Each city dataset encompasses 30 hotels, with approximately 40 reviews for each hotel. The dataset includes various attributes, such as hotel names, descriptions, images, price ranges, ratings, reviews, and location details.

Dataset link

Colab to load data

Data Dictionary:

  • hotel_name: Name of the hotel.
  • hotel_description: Description of the hotel.
  • hotel_image: Image of the hotel.
  • price_range: Price range of the hotel.
  • rating_value: Aggregate rating of the hotel.
  • review_count: Number of reviews for the hotel.
  • street_address: Street address of the hotel.
  • locality: Locality of the hotel.
  • country: Country where the hotel is located.

Hackathon Levels:

Level 1: Semantic Hotel Search RAG System

Objective: Build a RAG (Retrieval-Augmented Generation) based system using Qdrant as the Vector Db, that empowers users to input semantic queries about the hotels they are searching for. The system should not only retrieve relevant hotels in the corresponding city but also utilize a decoder model to explain why a particular hotel matches their preferences.

Level 2: Integration with Traversaal.ai's Ares API

Objective: Augment the Level 1 RAG model by integrating Traversaal.ai's Ares API, which performs real-time internet searches. Participants are encouraged to enhance their RAG applications by incorporating relevant details about hotels or locations obtained dynamically through the Ares API. E.g. “food near these hotels”, “things to do in this area” or “articles/blogs about the hotel not available in the dataset”

Participants can utilize this api endpoint by signing up at: api.traversaal.ai and get access to 100 web searches for free per user - no credit card needed

You can see Ares documentation here.

Level 3: Conversational Chatbot with OpenAI Functions

Objective: Develop a conversational style chatbot capable of engaging in multiple conversations with users about their hotel preferences. The chatbot should seamlessly invoke OpenAI functions to generate RAG outputs. Additionally, participants are expected to leverage the Ares API endpoint within the chatbot to provide users with real-time information.

Key Highlights:

  • Innovative Approach: Participants are encouraged to adopt innovative approaches in designing RAG models, integrating external APIs, and developing conversational chatbots.
  • Real-time Augmentation: The incorporation of Traversaal.ai's Ares API introduces a real-time dimension, enabling participants to enhance their applications with up-to-date information from the internet.
  • User-Centric Solutions: The ultimate goal is to create AI-driven solutions that not only retrieve relevant results but also offer insightful explanations and engage users in dynamic conversations.
  • Collaboration Opportunities: The hackathon provides a platform for collaboration, knowledge exchange, and exploration of state-of-the-art technologies in the realm of AI and natural language processing.

Submission Criteria:

  1. Request to join Traversaal Hackathon org by clicking here:
  2. You will submit your code as a Streamlit or Gradio App in this org
  3. Your code should be well documented with your team name and email addresses
  4. For any issues, please email us at: [email protected]

Evaluation:

Code submissions will be based on:

  1. Completion of each level
  2. Creativity applied in solving the problem
  3. Overall model performance based on a secret dataset