Really happy to see how seriously Rayon is taking AI, especially with the new AI panel for renders, image editing and turning assets into CAD blocks. It already feels like a powerful co-pilot inside the drawing environment.
I think the next big step is making the AI context aware within the canvas itself. Right now, the screenshot approach works for smaller plans, but on larger layouts (1,000 to 2,000 sqm) a full view capture loses detail and requires manual effort each time.
Feature suggestion: allow users to select objects or regions directly on the canvas and pass that selection to the AI as context, without needing a manual screenshot. Behind the scenes, Rayon could generate a high resolution vector or raster export of the selected area and send that to the model automatically. This would preserve line quality, labels and fine details, improving render accuracy.
Miro is a good reference here. Their AI is embedded into the canvas and treats the board as live context. You can select elements, reference them in a prompt, and the AI responds based on what is actually on the board. It feels seamless because there is no switching to a separate interface or capturing images manually.
I would love to see Rayon move toward this kind of integration: select a floor plan or region, prompt the AI, and have it work directly from the underlying data or a high fidelity export, all handled in the background. That would make the AI feel truly native to the design workflow rather than an extra step.
