Adding documents and sources
Everything Kumkuat scores, compares or alerts on is a document in your workspace library. Documents arrive three ways.
Upload
Open Communications in the sidebar and click Add Document. You can upload PDFs, Word, PowerPoint and Excel files, plain text, CSV, Markdown, images and audio, or paste a web page or YouTube link. Add tags on the upload screen if you want to. Large uploads run in the background; you do not have to wait on the page.
Feeds
Two feeds are seeded when you complete the workspace wizard, and can be managed from Comms Sources when that page is enabled for your workspace:
- Press release feed — press releases for your company and its competitors, keyed by ticker.
- Transcript feed — earnings-call transcripts, stored speaker-segmented, for a configurable number of quarters.
Each feed can be paused, given a cadence, or run now. See Comms Sources.
Crawled sources
Knowledge bases hold sources that Kumkuat crawls on a schedule: newsrooms and company sites, RSS feeds, saved web searches, and public social accounts on X, LinkedIn, Bluesky and Truth Social. Knowledge bases can be linked to personas so that what an audience has said recently informs how it reacts. See Knowledge bases.
What happens to a document
- Text is extracted on the platform. For a file that local extraction cannot read, a signed, time-limited link is passed to the crawl provider as a fallback.
- Auto-tags are applied — CEO, executives, company, communication type and any tags you have added. Aliases normalise the many names an executive may go by.
- Signals are extracted against your themes, so Compare Narratives and Alerts have something to work with.
- Embeddings are built so chat and narrative testing can retrieve the most relevant passages.
Documents are stored in the United States, and nothing is published outside the platform automatically. The data flow page lists every field that is retrieved and where it lands.