Adding documents to the Knowledge Base

Documents are the fastest way to give Serpio facts that are not on your website: internal positioning, a product spec, a pricing sheet, a research report you have the rights to use. This guide covers the Upload documents tab of the Add knowledge modal, what happens after you press upload, and when the document starts informing your writing.

Supported formats and limits

Each file can be up to 20 MiB. The modal rejects an empty file, a file over the limit, or an unsupported extension before upload starts, with a message such as Choose a PDF, DOCX, Markdown or TXT file. Very long documents can exceed Serpio's per-document passage limit; if that happens, nothing is indexed and the error asks you to split the file.

Upload documents

1. 1

   Open the modal

   On the Knowledge base page, press Add knowledge. The Upload documents tab is selected by default.

2. 2

   Choose files

   Click Choose files or drop them here, or drag files onto it. You can add several at once. Each file appears as a row marked Ready to upload. Press Remove on any row you added by mistake.

3. 3

   Set usage

   Below the file list, choose Use this as and Permission for writing, tick any Content focuses, and set Effective from or Review by if the document is time-bound. The same settings apply to every file in the batch.

4. 4

   Press Upload documents

   Files upload one at a time. Each row moves through Uploading with a percentage, then Upload received; queueing processing…, then Queued for processing. Follow progress in your knowledge base. A row that fails shows the reason in red and can be retried from the same modal.

Press Close once every row reads queued. The modal does not need to stay open; processing continues on the server.

What happens during processing

A status strip appears above the source list for each queued document and refreshes every few seconds. It walks through four messages.

1. Knowledge queued for processing… The file is stored privately and a worker has not picked it up yet.
2. Extracting source content… Serpio is reading the file and turning it into plain text.
3. Indexing knowledge… It becomes available when indexing succeeds. The text is split into passages and each passage is embedded and stored.
4. Processing complete. Your knowledge base is up to date. The document is live.

In the list, a brand new document shows Processing until its first version is ready. When you replace a document that already has a good version, the row shows Ready · updating instead, and the old version keeps serving until the new one activates. If processing fails, the strip shows the error and the row reads Needs attention for a first upload or Ready · refresh failed for a replacement. Serpio retries interrupted jobs up to three times on its own; after that, open the source and press Reprocess original.

Scanned PDFs and files with no text

If a PDF fails with the image-only message, the file has no extractable text. Serpio does not perform OCR. Get the text out another way and upload that instead.

1. 1

   Export from the original

   If you have the source document in Word, Google Docs, or a design tool, export it again as a PDF with real text, or as DOCX or Markdown.

2. 2

   Run OCR outside Serpio

   Most PDF tools can recognise text in a scan and save a searchable PDF or a text file. Check the result for errors before uploading; the writer will trust what it reads.

3. 3

   Upload the text version

   Use Add knowledge for a new source, or Replace document in the failed source's drawer to keep the same entry.

The same rule applies to any file that yields only a few dozen characters of text. The error says the source did not contain enough readable text.

Filling in usage settings

The settings under the file list are explained in full in Usage settings and permissions. The short version: Use this as decides whether the file describes your brand (Brand knowledge) or backs up claims (Reference evidence). Permission for writing decides whether facts from it may appear in article copy (Facts allowed in articles) or only shape understanding (Background context only). Documents have no Check URL for changes setting; as the form notes, documents change only when replaced.

When a document becomes available

A document is available the moment its version reaches Ready. That bumps the brand's knowledge revision, and every article started from then on can retrieve its passages. Idea collections and articles already written remain unchanged. Processing a document does not spend article credits.

From the drawer you can download the stored file with Original (the link is valid for five minutes), upload a newer file with Replace document, or stop it informing writing with Pause use. See Versions, statuses, and source actions.