Add documents

A knowledge base gets its documents from its document dataset. Anything you add to that dataset, from Brain or from Studio, reaches the knowledge base once its text is extracted.

How documents reach a knowledge base

Every document goes through the same steps:

1. Upload. The file is added to the document dataset of the knowledge base.
2. Text extraction. Annota Studio extracts the text: OCR for PDFs and images, transcription for audio, conversion for office and text files. See Documents and OCR.
3. Indexing. The knowledge base indexes the extracted text and writes a summary of the document. If Approved documents only is on, a document is indexed only after it is approved in review.

When you edit a document’s text in Studio, the knowledge base indexes the changed document again. When you delete a document from the dataset, or it is no longer approved while Approved documents only is on, it leaves the knowledge base.

Upload from the knowledge base

1. In Brain, open Knowledge and select a knowledge base.
2. On the Documents tab, drag files onto Drag & drop files here, or select the area to browse.
3. Check the list of files, then select the upload button, for example Upload 3 files.

The upload area is shown only to people who can edit the knowledge base.

Note

Uploaded documents are not approved yet. If the knowledge base has Approved documents only on, approve them in Studio before they are indexed. See Review assets.

Add documents from Studio

You can also upload to the dataset in Studio and manage it there: select Open dataset on the knowledge base page. See Upload images and documents. Documents you add in Studio show up in the knowledge base automatically.

Supported formats

Kind	Formats	Notes
Scanned and digital documents	PDF, PNG, JPEG	PDFs up to 1,000 pages. Text extraction uses credits.
Office and text files	DOCX, ODT, RTF, PPTX, XLSX, CSV, HTML, EPUB, XML, Markdown, TXT	Converted for free. Up to 1 MB of text per document.
Audio recordings	MP3, M4A, WAV	Transcribed into timestamped pages with an overview. Billed by the minute.

Every file can be up to 100 MB. To be indexed, a document’s extracted text can be up to 1 MB in total and up to 64 KB per page.

Document states

The Documents tab lists every document of the dataset with its Length and Status:

Status	Meaning
Queued, Processing, Paused	Text extraction has not finished yet.
Indexing	The document is waiting to be indexed or is being indexed. A document indexed again keeps serving its previous content meanwhile.
Indexed	Brain can search and cite the document. Select its name to read it.
Failed	Indexing failed. The reason is shown under the name.
Excluded	The document is left out, with the reason under its name.

A document is excluded when its text could not be fully extracted, when only approved documents are indexed and it is not approved, or when it is unreadable or beyond the limits above.

Fix failed and excluded documents

1. If the status card says Needs attention, select Retry. Brain indexes the failed documents again and shows Indexing restarted.
2. If a document is excluded because its text could not be fully extracted, retry its text extraction in Studio. See Documents and OCR.
3. If a document is excluded because it is not approved, approve it in Studio, or turn off Approved documents only in the knowledge base Settings.
4. If a document is beyond the limits, split it into smaller files and upload them again.

If your organization runs out of credits, the status card says so and waiting documents are indexed after a top up. Organization owners and admins can top up in Settings > Billing.

Next steps

• Ask questions
• Knowledge bases
• Documents and OCR
• Credit usage