Skip to main content

Upload documents

End user & admin

Add your own files to a collection so agents can search them. Ask handles PDFs, Office documents, and scanned images, including Arabic, without extra setup.

Prerequisites​

  • An Ask account on your organization's instance.
  • A collection you can write to. If you do not have one yet, see Knowledge bases, explained.
  • The files you want to upload, on your device or in a supported connector. For connector-based imports, see Connectors.

Supported formats​

Ingestion runs through the Knowledge Management Service (KMS). At time of writing, KMS supports:

  • Documents: PDF, DOC, DOCX, PPTX.
  • Spreadsheets & data: XLS, XLSM, XLSX, CSV, TSV, JSON, XML.
  • Text, Markdown & web: TXT, MD / Markdown, HTML, HTM.
  • Images: JPG, JPEG, PNG, processed with OCR.
  • Audio: MP3, WAV, M4A, OGG, WEBM, FLAC.
  • Arabic OCR: supported for scanned PDFs and image files. Quality is improving as Squad 1's new multilingual OCR rolls out.

More file types, and nested folders pulled from connectors, are on the roadmap. See Connectors for pulling content from Notion, SharePoint, and Google Drive.

Upload from your device​

  1. Open Knowledge from the left sidebar.

  2. Pick the collection you want to add files to, or click New collection to start a fresh one.

  3. Drag files from your desktop into the Drag and drop files zone, or click + Add files and choose Upload files (or Upload folder to bring in a whole folder from your computer).

  4. Wait for ingestion to finish. Files are parsed, chunked, embedded as needed. Larger files, and especially large PDFs, take longer.

  5. Once a file shows as ready, it is searchable from any agent the collection is attached to.

Ingestion statuses​

As a file moves through ingestion, its status updates in the collection's file table. The happy path, in sequence:

  • Uploading: the file is transferring to Ask. If this is taking a long time, it is almost always a function of your network, so check your Wi-Fi connection. On a VPN it will take longer.
  • Queued: uploaded and waiting for processing.
  • Processing: this is where parsing, chunking, and embedding happen.
  • Ready: done; the file shows a checkmark and is searchable by attached agents.

If something goes wrong, a file lands in one of three terminal states: Upload Failed (the transfer didn't complete, so re-upload), Processing Failed (ingestion choked, so try re-uploading or split the file), or Cancelled. You can filter the file table by any status.

Don't navigate away while uploading

Wait until a file finishes Uploading before you leave the page. If you navigate away mid-upload, the upload is aborted. Once the file reaches Processing, it is safe to navigate away, and it continues in the background.

OCR & Arabic​

Scanned PDFs and image files go through OCR automatically, so you do not need to flag them. Ask is actively rolling out a new multilingual OCR pipeline tuned for Arabic. If a previously poor extraction improves on its own, that is why.

tip

Born-digital PDFs (exported from Word, InDesign, etc.) extract cleaner than scans. If you have a choice, upload the digital version.

Size and quantity limits​

The per-file limit is 200 MB, and the drop zone states it as Max Size: 200 MB. (Limits can differ on sovereign-cloud or on-prem deployments; the drop zone always shows your instance's value.) If a file is rejected, try splitting it, since large reference manuals search better as several scoped PDFs than as one giant file anyway.

Heads up

Treat uploads like any other corporate data. A document you upload to a Shared collection is visible to everyone in your organization who can see that collection, including via any agent it is attached to.

Tips & gotchas​

Name files for humans

The file name shows up in citations. HR-policy-2026-v3.pdf is much friendlier than doc_4382_final_FINAL.pdf.

note

Ingestion is not instant. Expect anything from a few seconds for a small DOCX to several minutes for a long, image-heavy PDF. You can keep working while it runs.

Knowledge baseConnectorsPull content from Notion, SharePoint, Google Drive, and live databases.Knowledge baseOrganize & manageRename, pin, and clean up collections from the Knowledge sidebar.Knowledge baseAttach a knowledge base to an agentWire a freshly uploaded collection into an agent so it can use it.