Palet Docs

An agentic search engine built on a federated protocol. Powered by Solana.

This docs repository contains the official documentation for Palet's Solana AI Hackathon submission. The submission GitHub repo can be found here.

You can try Palet for yourself by following this link. Search queries at the moment are limited to call functionality as Palet is still in early development. Keep in mind that every query you submit will generate real results. Meaning that dozens of live phone calls will be made to real businesses.

Introduction

Palet is a new kind of search engine that gathers information from both online and offline sources to give contextual answers. Unlike traditional search engines that depend on web-based rankings (relying on 'blind' signals like backlinks, keywords, etc.), Palet works like a person doing a search: it tries different keywords, digs through multiple results, and even contacts businesses to confirm details if it needs to. For example, if you want the part number for a 1993 Toyota JZ147 rear logo, Palet will keep looking until it finds it. If you want to know which ice cream shops nearby have strawberry vegan ice cream, Palet will search maps and call shops to check. Regular search engines can’t do all that on their own—but Palet can.

Jump right in.

Getting Started

Get started with a quick overview.

Architecture

Learn more about Open Context, the protocol Palet is built on.

Additional Works

Learn about how we used a model ensemble to hide secrets.

At the moment, Palet is limited to a very early version of Agentic Search. We built it in 10 days. However, we are working on making it better and shipping Deep and Local Search features by early February 2025. GitHub repo can be found here.