Skip to main content
Documents are the corpus your assistant answers from. Add content to a knowledge base and AutoSage processes it in the background, turning it into grounded, citable passages. Everything below happens from the dashboard — no code required.

​
What you can add

Files

PDFs (including scanned PDFs and images, read with OCR), audio files (transcribed to text), and structured documents such as Word files and spreadsheets.

A single URL

Point AutoSage at one web page and it scrapes the page to clean text.

A website

Discover the pages on a site and pick which ones to ingest. A crawl covers up to about 25 pages.
Upload is automatic. As soon as you provide a file, the dashboard begins ingesting it — there is no separate “start” step.

​
How ingestion works

Once content arrives, AutoSage processes it in the background so it becomes searchable and citable:
1

Parse

The raw file, page, or recording is read into text — including OCR for scanned PDFs and images, and transcription for audio.
2

Summarize

AutoSage generates a summary that helps the assistant reason about the document as a whole.
3

Chunk

The text is split into passages sized for accurate retrieval.
4

Embed & index

Each passage is embedded and added to the knowledge base’s private vector store, ready to be retrieved and cited.

​
Document status

Every document moves through a clear lifecycle you can watch in the dashboard:
StatusWhat it means
Pending uploadThe document is being received.
QueuedReceived and waiting for background processing to begin.
ProcessingBeing parsed, summarized, chunked, and indexed.
ProcessedReady — the document is now searchable and can be cited in answers.
FailedProcessing could not complete. Reprocess the document or remove it.
Wait until a document reads Processed before expecting answers to include it. Until then, its content is still being made searchable.

​
Reprocess and delete

Run a document through processing again — useful after a failed attempt, or when you want it re-indexed. The document returns to the processing states and becomes Processed when it finishes.
Removing a document takes it out of search immediately and cleans up its stored content and index entries. Deleting is safe and does not affect other documents in the knowledge base.

​
Keeping your quota tidy

Documents count toward your tenant’s max documents and max storage quotas from the moment they are added — including documents that are still pending or that ended in a failed state.
If an upload fails, delete it once you have retried. Clearing out pending and failed documents you no longer need keeps room in your quota for the content you actually want searchable.
When a new document would push the tenant past its max documents or max storage quota, the upload is refused with a clear message. Remove content you no longer need, or raise the tenant’s quota, then add the document again.

​
Next steps

Shape the answers

Set a persona and instructions so answers sound the way you want.

Build an agent

Turn your knowledge base into a deployable, schedulable agent.