Album AI is an experimental project that allows users to interact with their photo albums using natural language. It targets photography enthusiasts and researchers who need to manage and query large image collections, offering a novel way to discover and utilize image metadata through AI.

How It Works

The system leverages large language models (LLMs) like GPT-4o-mini and Haiku for visual analysis, automatically identifying and generating metadata for images. It then employs Retrieval Augmented Generation (RAG) to enable conversational querying of the album's content. Images are stored in a PgSQL database, with metadata vectorized using OpenAI's Embedding API for efficient search and retrieval.

Quick Start & Requirements

• Install by cloning the repository and running the provided build.sh script.
• Requires OpenAI API key and optionally an Anthropic API key.
• Local setup involves modifying a .env.prod file with API keys and hostnames.
• Access the demo at http://localhost:8080.
• Official Demo: album.gcui.ai

Highlighted Details

• Automatically discovers and identifies metadata from images.
• Provides both a traditional image search API and a RAG-based chat API.
• Supports one-click Docker deployment on platforms like Render.
• Uses PgSQL for image storage and OpenAI Embeddings for vectorization.

Maintenance & Community

• Project is actively updated, encouraging stars and contributions.
• Community interaction and feedback are managed via GitHub Issues.
• Discord server available for discussion.

Licensing & Compatibility

• Licensed under the Apache 2.0 License.
• Commercial use requires direct contact with the authors.

Limitations & Caveats

The project is described as experimental, and server deployment guidance is noted as needing improvement. Commercial use is explicitly stated as requiring author contact, suggesting potential licensing nuances beyond the Apache 2.0 declaration.