Image reading on every chat model · Site pages back in search

Image understanding in chat

• You can now attach an image and ask about it on any chat model, including ones without built-in image support — Kollab automatically routes the image to a dedicated vision model and answers from what is actually in the picture.
• The vision engine behind that hand-off moved to DeepSeek V4 Flash Vision, so image questions resolve on a faster, more reliable route and no longer fail when the previous backing model is unavailable.

Site & discoverability — https://kollab.im

• Fixed documentation, comparison, use-case, tool, legal and updates pages that search engines were unable to index, so they can be found again from search.
• Fixed shared-link preview pages leaking into search results, so a shared task no longer competes with the real product pages.

Command line tool

• Fixed the standalone tool definitions in the kollab CLI so their field limits match what the service actually accepts, preventing requests from being rejected after they were sent.