MemeMeow-Studio/MemeMeow

What it solves

MemeMeow is a natural language meme retrieval tool that eliminates the need to remember specific filenames or manual tags. It allows users to find the perfect meme by describing the scene or mood they want to convey, such as "expressing helplessness" or "morning commuter."

How it works

The system uses a combination of text and visual embeddings to make memes searchable:

• Semantic Search: Uses embedding models to enable Q&A style retrieval based on natural language descriptions.
• Automated Context Understanding: An integrated OpenCode Agent analyzes uploaded images to generate structured descriptions and context without manual tagging.
• Visual Similarity: Employs the DINOv2 vision model to perform nearest-neighbor searches for visually similar images.
• Hybrid Uploads: Users can choose "Direct Upload" (using filenames and visual vectors) or "Full Analysis" (using an AI agent to determine the image's origin, usage, and meaning).

Who it’s for

Individuals who manage large collections of memes and want a fast, intuitive way to search through them using natural language or visual similarity rather than manual folder organization.

Highlights

• AI-Powered Tagging: Automatically generates context and descriptions for images via an AI agent.
• Multi-Modal Retrieval: Supports both text-based semantic search and image-based visual similarity search.
• Web-Based Management: Includes a Vue 3 interface for browsing, filtering, and organizing memes into collections.
• Containerized Deployment: Easily deployable via Docker Compose with a PostgreSQL/pgvector backend.