add-image-vision

Processing WhatsApp image attachments requires downloading the media, resizing it with sharp, and delivering it as a base64 multimodal content block. This pipeline enables agents to decode and describe user-sent images within secure group containers.

Adding multimodal vision to container-based agents requires syncing changes across the agent-runner and container-runner. You merge the skill branch, rebuild the container, and deliver base64 content blocks so the agent can reason over visual data.

Yes, the image processing pipeline requires media preparation before encoding. Images are downloaded from WhatsApp, resized using the sharp library, and stored in the workspace before being encoded into base64 content blocks for agent delivery.

Agents ignore visual context from WhatsApp photos when multimodal content blocks are not configured. Without downloading and resizing images via sharp to deliver base64 data, the agent lacks the visual cues needed to interpret user-sent media.

Yes, you can use sharp to resize WhatsApp images before agent delivery. The pipeline downloads image attachments, applies sharp resizing, stores the processed media in the workspace, and encodes it as base64 for the agent to describe.