Manga.AI — Make NFTs Great Again

Overview

Manga.AI is an AI-powered metadata search and generation engine for Solana Non-fungible Tokens (NFTs). It enables querying NFTs by their metadata traits (e.g. "brown hair or dark eyes") rather than limiting discovery to collection-based navigation.

NFT metadata on Solana is flexible but inconsistent across projects, which makes it difficult to index, compare, or search NFTs by their attributes. Manga addresses this problem by:

• Parsing natural language queries into structured trait filters
• Matching those traits against indexed NFT metadata and
• Returning matching NFTs ranked by accuracy and relevance to the query

In addition to search, Manga also provides an AI-powered metadata generation engine. Given an image, it extracts visual traits and produces Solana-compliant metadata, making it useful for repairing incomplete collections or preparing assets for minting.

Usage

Before cloning the repository, ensure you have the following installed on your machine:

• Node.js (LTS version recommended — v24 or higher)
• Yarn (package manager)
• Ollama — with the following models pulled locally:
  • llama3.2
  • gemma3
  • nomic-embed-text

You may verify your setup with:

node -v 
// v24.2.0 or higher

yarn -v
// 1.22.22 or higher

ollama list
// NAME                       ID              SIZE      MODIFIED
// llama3.2:latest            84d5af170cac    2.0 GB        -
// gemma3:latest              f6ca2ab7fc3e    3.3 GB        -
// nomic-embed-text:latest    0b49f47022a1    274 MB        -

Once you have confirmed that the prerequisites are installed and the correct versions are available, clone the repository and start the development server:

git clone https://github.com/collinsezedike/manga.ai.git
cd manga.ai
yarn install
yarn dev

Roadmap

Manga.AI is an evolving project, and the current implementation represents the foundation of a much larger vision. Upcoming milestones include:

• Real-time on-chain indexing — move beyond static datasets to support live ingestion of newly minted NFTs.
• Cross-chain expansion — extend metadata parsing and search capabilities to Ethereum, Polygon, Bitcoin Runes, and other emerging NFT/token standards.
• Model fine-tuning — improve language and image understanding by training on contemporary NFT collections, memes, and emerging metadata vocabulary.
• Metadata quality scoring — introduce a ranking heuristic to evaluate metadata consistency, completeness, and semantic clarity.
• Public API & SDK — expose search and metadata generation as programmatic endpoints for wallets, marketplaces, and NFT launchpads.
• RWA & on-chain identity support — extend trait-based discovery to real-world assets and identity-bound NFTs, enabling queryable utility beyond PFPs.