The winners of the hypermode knowledge graph + AI challenge

Last month we announced the Hypermode Knowledge Graph + AI Challenge to reinforce how important knowledge graphs are to building successful AI applications. Over 500 members of the Hypermode community rose to the challenge and submitted 50 model-native apps using Modus and knowledge graphs. Today, we'll announce the winners and review the top projects from the first Hypermode Knowledge Graph + AI Challenge.

The rules for the challenge were simple: each project must use the open source Modus API framework, at least one AI model, and a knowledge graph. We wanted to be as open-ended as possible to allow the community to use their creativity to show the types of problems that can be solved with knowledge graphs and AI.

The judging panel included developers and data scientists with strong data and AI backgrounds from Hypermode, Neo4j, Disney, and Marriott International. Let's see which projects they selected.

Best Use of Knowledge Graph RAG: TravelMate

Travelmate is a Smart Travel Planning Tool designed to enhance your journey. Whether you're visiting a new city, traveling to another state or exploring a new country, TravelMate helps you discover the best places to visit. The inspiration for TravelMate came from the frustration of sifting through generalized and often irrelevant travel recommendations.

What it does:

• Identifies the top 20 attractions in a destination using the Google Maps API.
• Fetches detailed descriptions of these attractions from Wikipedia.
• Builds a dynamic knowledge graph using Neo4j to map relationships between attractions.
• Queries the graph to retrieve relevant nodes and relationships up to a depth of 3.
• Uses an LLM to analyze user preferences and graph data to generate personalized travel suggestions.

Tech stack:

• Front-end: A React-based website for an intuitive and user-friendly interface.
• Data collection: Google Maps API for location data and Wikipedia API for detailed descriptions.
• Knowledge graph: Neo4j database accessed via the Neo4j driver and Modus framework to build and query the graph.
• AI integration: LLM integration to process user inputs and graph data for personalized recommendations.
• Backend logic: Gemini API through Modus for efficient communication between components.

Best Use of Dgraph: GitExplore

GitExplore allows you to effortlessly explore users, repositories, followers' projects, and starred repos, and dive deeper with intelligent, context-aware chats using knowledge graphs and LLMs.

The judges were impressed with this project's comprehensive and sophisticated use of Dgraph to build a knowledge graph of GitHub's social ecosystem.

What it does:

GitExplore transforms the way developers interact with their GitHub network by providing a powerful connection visualization system. The platform enables users to:

• Quickly discover and analyze mutual connections between friends and other GitHub users
• Explore repositories and starred projects in an intuitive interface
• Understand connection patterns and shared interests within their network
• Navigate through their GitHub social graph with enhanced visibility and context

Tech stack

• Dgraph: Serves as the primary database, storing and managing the social graph data
• Modus: Provides the framework for building and managing the application
• GitIngest: Handles the efficient ingestion of GitHub data
• Vite: Powers the fast and efficient development environment
• React: Delivers a responsive and interactive user interface

Best Agentic App: Flavour Fiesta

Flavour Fiesta uses Neo4j and Modus to facilitate recipe discovery, personalized food recommendations, shopping list generation and meal planning. It also uses an LLM to enable chat with a knowledge graph of the recipe dataset to create meals and recipes on the fly using a text-to-Cypher GraphRAG pattern. By leveraging tool calling in Modus, the Flavour Fiesta agent is able to perform external actions and retrieve dynamic data.

Key Components:

• Text-to-Cypher pattern: This was the core functionality that translates user queries into Cypher commands to interact with the graph database.
• Tool calling: Implemented a system that allows the agent to perform external actions and retrieve dynamic data, enhancing its capabilities.
• Memory integration: Adding memory was essential for enabling the agent to maintain context across multiple turns of a conversation. It was possible to do so by storing conversations in a Neo4j database.

Best Use of Neo4j: Cite Smart AI

The idea for Cite Smart AI stems from the challenges researchers face when managing academic citations and exploring interconnected papers. Finding relevant research, understanding how papers cite each other, and organizing citation data can be time-consuming and overwhelming. This tool aims to simplify that process, providing a personal AI companion to make research smarter, faster, and more insightful. By blending advanced AI with intuitive user interfaces, Cite Smart AI empowers users to focus on discovery and learning.

What it does:

Cite Smart AI helps users manage and explore academic citations with ease. Its key functionalities include:

• Searching citations: Find relevant papers based on titles, keywords, or research topics.
• Connecting papers: Visualize how papers cite or are cited by others using a dynamic graph structure.
• Answering questions: Use AI to provide context-aware answers about academic papers and citations.
• Storing citation graphs: Save, and expand your citation networks for ongoing research projects.

In essence, it's a personal AI companion focused on simplifying research and citation management.

Tech stack

Front-end:

• Next.js: Framework for building a fast, dynamic, and user-friendly interface.
• ShadCN + TailwindCSS: For modern, consistent, and responsive UI components.
• Supabase: Handles user authentication, ensuring secure access to the app.

Backend:

• Modus + Hypermode: Enables server-side functionality with AssemblyScript and integrates a scalable backend for handling data.
• DeepSeek AI: Powers intelligent citation search and retrieval from Semantic Scholar.
• Text Transformer Model: Provides natural language understanding for answering user queries.

Honorable Mention: MindbookLM

What if you had a friend who never forgot? A companion who remembered every story you shared, every thought you expressed, every moment you wanted to preserve? That's MindbookLM - your personal digital brain that ensures no precious memory needs to fade away.

Unlike traditional journals or note-taking apps that simply store information, MindbookLM becomes a living bridge to your past. It's designed for those future moments when you want to reconnect with who you were, what you felt, and how you've grown.

The Runner-Up: Ups & Downs

The game "Ups and Downs" is inspired by the classic game "Snakes and Ladders." "Ups and Downs" let the player have an experience of life, decisions, and circumstances. The game is a life simulation game where players roll a die and climb the stairs of life. Each step represents a life event outside their control. The steps on which player land on, presents a life question. There are some steps that are randomly selected before the game which are the "Snakes/Ladders" questions, meaning the answer to this question may give success or setback to the player in the life simulation. The outcome of these questions depends on the player's background and decisions, determining whether they face setbacks (snakes) or successes (ladders). Upon reaching the 100th step, players receive an analysis of their journey and can chat with an AI modeled bot about the life they've lived.

Grand Prize Winner: Dev Docs

DevDocs is an AI-driven platform that delivers real-time, precise answers from company documentation. Unlike other AI tools that often rely on outdated syntax or generic data, DevDocs stays current. Whether you need a code snippet, syntax explanation, or an API endpoint, DevDocs ensures you get what you need efficiently.

Key features include:

• AI-Powered Documentation Search: Uses Llama 3.1 and MODUS inference to generate contextually accurate responses.
• Custom Company Integration: Developers can add their own companies and create personalized AI chatbots.
• Instant Accessibility: Simple user interface with seamless navigation.
• Real-Time Results: Eliminates the need for manual browsing of extensive documentation.

Conclusion

Congratulations to all the winners and thank you to everyone who participated in the Hypermode Knowledge Graph + AI Challenge and especially to our judging panel for reviewing each project.

To see all submitted projects and to read more about each one, including demo videos and links to the code for each project, check out the project gallery.