Open /upload from the sidebar and either click the dropzone or drag files (or a folder!) onto it. Each file is hashed in your browser, uploaded directly to object storage, and registered as a Kodori document. Duplicates dedupe automatically — if the same bytes already live in the tenant, no second copy is created.
Folder uploads preserve the directory structure as upload metadata and offer to create a Kodori Collection from the top-level folder name when the batch finishes. This is the fastest way to bring a project from a shared drive into a single Collection.
Once a file lands, content extraction starts in the background:
- text-shaped formats (.txt, .md, .csv, .json, .yaml, .xml, .svg) decode immediately - Office files (.docx, .xlsx, .pptx) extract via pure-JS adapters - PDFs and images extract via Anthropic's vision API - legacy .ai files are sniffed for an embedded PDF and routed accordingly
You can watch extraction progress on the document detail page; it auto-refreshes every 15 seconds while the tab is open.