add-image-vision

You can enable image vision for NanoClaw agents by downloading WhatsApp images, resizing them via sharp, base64 encoding the data, and passing it as multimodal content blocks directly to Claude.

The best way to send base64-encoded images to Claude is by resizing the source image with sharp, converting the output to a base64 string, and structuring it as a multimodal content block for Claude's vision API.

Yes, you need the sharp library to resize images before sending them to Claude, as it handles the image processing step required to create the base64-encoded multimodal content blocks for Claude.

Yes, you can use NanoClaw agents to understand images sent in WhatsApp group channels by processing the attachments into base64-encoded multimodal content blocks for Claude vision processing.

Your NanoClaw image processing pipeline requires code changes in src/image.ts to properly handle the download, sharp resizing, and base64 encoding logic needed to deliver multimodal content blocks to Claude.