Desearch-ai/subnet-22

Subnet 22

What it solves

AI applications need relevant, current web content to answer queries, but fetching and processing pages on the fly is slow and expensive. Desearch builds a persistent, maintained index ahead of time, so searches are fast and reliable.

How it works

On Bittensor Subnet 22, miners crawl assigned web pages, extract their text, and convert it into vectors using their own GPUs. Validators re-check a sample of both the crawling and embedding work to ensure quality. Accepted work becomes part of a shared index that the Desearch team operates, so queries never wait for miner work. Miners earn rewards based on how much of their work is accepted.

Who it’s for

Developers building AI agents and research tools that need quick, up-to-date web search. Also for anyone wanting to participate as a miner (crawling and embedding) or validator (checking others’ work) to earn rewards on Bittensor.

Highlights

  • Decentralized competition on Bittensor Subnet 22 for web indexing
  • Pre-built index avoids slow, on-the-fly retrieval
  • Miners handle crawling and embedding; validators ensure quality
  • Public API with Python, JavaScript, and MCP server SDKs

Related

  • Project
    iann0036/iam-dataset

    This project provides a consolidated set of structured JSON data for AWS, Azure, and Google Cloud identity and access management (IAM), including SDK call mappings, service authorization references, role definitions, permissions, and more. It is used by multiple security and access-management tools to avoid duplicate scraping and standardization efforts.

  • Project
    dgraph-io/dgraph-js

    dgraph-js is the official JavaScript client for Dgraph, a distributed graph database. It provides a gRPC-based API for schema alteration, transactions, mutations, queries, and upserts, with support for ACL, TLS, and multi-tenancy.

  • Project
    ko-ichi-h/khcoder

    KH Coder is a software tool for quantitative content analysis and text mining used for computational linguistics across multiple languages.

  • Project
    slive777/OpenAver

    OpenAver is a local, desktop-based metadata scraper and media manager for JAV collections that organizes files, fetches metadata, and provides a visual browsing interface with AI-ready API capabilities.