This project provides an open-source, self-hostable meme search engine for organizing and retrieving personal meme collections using AI. It targets users who want to leverage AI for content-based meme retrieval, offering local processing for privacy and control.

How It Works

The engine indexes memes using AI-powered image-to-text models to extract descriptions and text from images. These descriptions are then stored as vector embeddings in a PostgreSQL database with the pgvector extension, enabling fast semantic search. Users can also add manual descriptions and tags for enhanced organization and retrieval.

Quick Start & Requirements

• Install/Run: docker compose up
• Prerequisites: Docker, local PostgreSQL instance (if not using Docker Compose's bundled DB).
• Setup: Initial model download may take time.
• Docs: https://github.com/neonwatty/meme-search

Highlighted Details

• Supports multiple image-to-text models (Florence-2, SmolVLM, Moondream2) with varying parameter sizes.
• Features manual description editing and tag-based filtering.
• Utilizes PostgreSQL with pgvector for efficient vector search.
• Allows organizing memes across multiple directories.

Maintenance & Community

The project is under active development, with a changelog available. Feature requests and contributions are welcomed via the GitHub discussions section.

Licensing & Compatibility

The project appears to be under a permissive license, but specific details are not explicitly stated in the README. Compatibility for commercial use or closed-source linking would require verification of the exact license.

Limitations & Caveats

The initial setup requires downloading AI model weights, which can be time-consuming. Customizing meme directory paths involves modifying Docker volumes and registering them within the application's UI.