AI Image Embedding Processor & Search Agent using n8n and OpenAI

Unlock Advanced Image Search & Analysis with this AI Agent

This AI Agent transforms how you interact with your image assets. It intelligently processes images by first fetching them (e.g., from Google Drive), then performing a dual analysis: extracting technical color channel information and leveraging OpenAI's powerful vision models to generate rich, semantic keywords describing the image's content, mood, and objects. This comprehensive data is then structured into a document, converted into a vector embedding using OpenAI, and stored in an in-memory vector database via Langchain nodes. The result? Your images become easily searchable not just by filename, but by their actual visual and contextual meaning, enabling sophisticated semantic search capabilities.

Key Features & Benefits:

• Automated Image Ingestion: Fetches images from sources like Google Drive or any other n8n-supported binary source.
• Dual-Mode Image Analysis: Extracts both technical color data (e.g., channel statistics, background color) and AI-generated semantic keywords.
• AI-Powered Keyword Generation: Utilizes OpenAI's vision models (e.g., GPT-4V) to comprehensively describe image content, subjects, mood, and even artistic techniques.
• Efficient Embedding Creation: Generates vector embeddings from the combined image data using OpenAI's embedding models.
• In-Memory Vector Storage: Leverages Langchain's in-memory vector store for quick setup and fast semantic search, perfect for prototyping, testing, or smaller datasets.
• Semantic Image Search: Enables users to search for images using natural language queries that match content and context, going far beyond simple metadata tags.
• Customizable Processing: Includes steps for image resizing (optimizing for AI model performance) and flexible data structuring for metadata.
• Foundation for Advanced Applications: Easily extendable for image-to-image search or integration with persistent, scalable vector databases.

• An n8n instance (Cloud or self-hosted).
• OpenAI API Key with access to vision-capable models (e.g., gpt-4-vision-preview or newer for analysis, and an embedding model like text-embedding-ada-002).
• Google Drive credentials configured in n8n if sourcing images from Google Drive (or credentials for your chosen image source).

1. Download the n8n workflow JSON file.
2. Import the workflow into your n8n instance.
3. Configure the 'Google Drive' node: set up your Google Drive credentials and specify the image File ID to download. Alternatively, replace this node with any other node that provides an image in binary format.
4. Configure OpenAI Credentials: In the 'Get Image Keywords' (OpenAI Analyze Image) node, select or add your OpenAI API credentials. Ensure the chosen model supports image analysis.
5. Configure OpenAI Embedding Credentials: In both 'Embeddings OpenAI' nodes (one for storing, one for searching), select or add your OpenAI API credentials. These will use an embedding model.
6. (Optional) Adjust Image Processing: The 'Resize Image' node is set to 512x512, optimal for some OpenAI models. Modify as needed.
7. (Optional) Customize Metadata: The 'Document for Embedding' node sets metadata like source, format, backgroundColor. Tailor this to your needs.
8. (Optional) Vector Store Key: The 'In-Memory Vector Store' nodes use image_embeddings as the memoryKey. Change this if you need to manage multiple image indexes.
9. Test Search: Modify the prompt in the 'Search for Image' node (e.g., 'student having fun') to test the search functionality.
10. Activate the workflow. Run it with an image to populate the vector store, then test the search functionality.