Each product has a library of sources. EDGR extracts their text, splits it into passages and indexes them, so the AI agent can search the library and quote the relevant part when a customer asks something. Open a product and the Knowledges tab holds the library.

​
Add a source

1

Choose Document or Text

Document uploads a file. Text takes a title and a block of text you paste in — useful for a policy or a set of answers that has no file.
2

Upload or paste

Drop files onto the box or click to browse. Accepted: .pdf, .docx, .md, .txt, .png, .jpg, .jpeg, .webp, .gif. Images are read by a vision model and stored as their description.
3

Describe it

The description is context for the model, so write it for a reader who has not seen the file: “Pricing sheet for the enterprise tier, 2026 Q2” beats “pricing”.
4

Wait for indexing

The source lands as Pending, moves to Processing while the text is extracted and indexed, and reaches Active when the agent can use it.

​
The source table

Columns are name, type, character count, status and the date added. The header above it counts total sources, active ones, and anything processing or failed.
StatusMeaning
PendingUploaded, waiting to be processed
ProcessingBeing extracted and indexed
ActiveIndexed. The agent can quote it
FailedProcessing failed. Open the row for the reason

​
Keeping the library useful

A short document about one thing retrieves precisely. A 200-page handbook returns whichever passage looked closest, which is often the wrong one.
The agent quotes whatever is indexed. Stale pricing in the library becomes stale pricing in a reply to a customer. Upload the new version and remove the old one.
The agent paraphrases, and a paraphrased contractual term is a commitment you did not make. Keep the library to marketing and sales material.