Description

Akeru Search: Gaianet Hackathon Project

TL;DR

Akeru Search is an AI-powered search engine combining a sleek, minimalist frontend with a robust backend. It features a Next.js-based UI with dynamic summaries and a TypeScript backend integrating multiple AI models (including gemma and Llama 3.1) for advanced search and summarization. The project showcases the potential of decentralized AI agents on the gaianet platform.

Project Overview

Akeru Search is a modern, AI-powered search engine with a minimalist interface, combining a sleek frontend with a powerful backend for advanced search and summarization capabilities.

Frontend (Akeru Search)

- Repository:

- Key Features:

- Minimalist, centered search bar with "Living Knowledge" title

- Quick search suggestions via clickable pills

- Grid layout for search results with hover preview

- Dynamic, gradually revealing summary for context

- Dark theme with elegant color schemes

- Tech Stack:

- Next.js 14

- React

- Tailwind CSS

- Framer Motion for animations

Backend (Agents API)

- Repository:

- Key Features:

- Multiple AI model integrations (GPT-3.5, Llama 3.1, Mistral AI, Gemma)

- Web search capabilities via Brave Search API

- Modular design for easy AI model integration

- Streaming and non-streaming response support

- API key authentication

- Tech Stack:

- Runtime: Bun

- Framework: Express.js

- Language: TypeScript

- AI Integration: LangChain

- Documentation: Swagger UI

- Containerization: Docker

This project showcases a powerful combination of modern frontend design and advanced AI-powered backend capabilities, making it an impressive entry for the gaianet decentralized agents platform hackathon.

Progress During Hackathon

We first explored multiple nodes and their capacities. We spent time reverse engineering the CLI to understand the full gaia system of full nodes. Then we worked with the public nodes to prototype. We worked with both brave search API and google api but we found brave had better results and was a more decentralized privacy focused approach. Then we moved to creating the RAG system in Langchain. We tested streaming results back to a sleek UI. We included in the UI link preview, sources and animations to make the best possible UI.

Fundraising Status

No funds yet. Looking to build a full agent swarm from this demo and create a agents protocol on top of gaia

Akeru Search

Videos