What's new
- Knowledge uploads. Drop in PDFs, Word documents (.docx) and reference photos. The AI extracts the text (or runs OCR on images), writes a short summary, and uses it whenever you ask questions by voice — so answers come from *your* materials, not the generic internet.
- Organize by folder, assign to projects. Group documents into folders (e.g. "Trade standards", "Vendor handbooks") or pin them to a specific project. The AI knows which job each document belongs to and weighs that when answering.
Improvements
- Live progress per file. Each upload shows its own progress bar while the AI processes it, with a retry button on any file that fails — you can see at a glance which files are done and which still need attention.
- Bulk actions. Multi-select on the knowledge list lets you retry or delete dozens of files at once instead of clicking through each one.