MemeMeow is a natural language-based tool for retrieving specific meme images, primarily focused on the "Zhang Weixiang" meme. It allows users to search for memes using descriptive queries, making it easier to find relevant images without knowing exact filenames or tags. The project targets users who frequently use memes in communication and offers a web interface, API, and even an iOS shortcut for convenient access.

How It Works

The system utilizes embedding models to enable question-and-answer style retrieval, automatically responding to queries with appropriate memes. It supports extensibility by integrating with Vision-Language Models (VLMs) for efficient image tagging and resource package creation. Users can also contribute their own meme resource packages to the community.

Quick Start & Requirements

• Install: Clone the repository and install dependencies using pip install -r requirements.txt.
• Run: Start the application with python -m streamlit run app.py.
• Prerequisites: Python 3.11+. Optional: Silicon Flow API Key or OpenAI API Key for VLM tagging.
• Notes: Initial local model runs require downloading model files.
• Links: Online Demo, Feedback

Highlighted Details

• AI-generated meme responses are distinct from personal opinions.
• Supports VLM for automatic image tagging and filename generation.
• Allows creation and sharing of meme resource packages.
• Provides a public API at https://api.zvv.quest/search.

Maintenance & Community

The project is community-driven, with contributions welcomed via Pull Requests and Issues. Links to related applications and community resources are available.

Licensing & Compatibility

The project is licensed under the MIT License, permitting commercial use and linking with closed-source projects.

Limitations & Caveats

Uploading new images and then exporting resource packages requires regenerating the cache. Migrating from older versions involves re-labeling and exporting/importing resource packages.