Answers from your own documents, cited and on-prem.
Ask a question and get an answer grounded in the files your teams already use, from documents and spreadsheets to decks and web pages, with a citation back to the source. Embeddings run on your own hardware, and every retrieval respects the permissions of the person asking.
Grounded in your own documents, and private by default.
Load what teams already use
PDF, DOCX, XLSX, PPTX, HTML, Markdown, TXT and CSV, at up to 1 GB per file and 100 files per request.
Nothing leaves your perimeter
Embeddings run on your own hardware. No document, and no fragment of one, is sent to a third party.
Trace every answer to its source
Each answer cites the documents it drew from, so your team can check the work.
People see only what they're cleared to see
Three-tier scoping enforces permissions on every retrieval, so a user gets back only what they are cleared to see.
Read scanned documents too
Scanned PDFs become searchable text, read by a sovereign vision model when one is registered.
One source for chat, agents and workflows
The same knowledge grounds your chat, your agents and your workflows. You index a document once and use it everywhere.
Ask your own documents, then check the citation behind every answer.
See grounded, cited answers come back from your own knowledge, with embeddings on your hardware and permissions enforced on every retrieval.