Skip to main content
Knowledgebase is organisational search. You ask in your own words. You get cited passages from documents the team uploaded or connected: policies, design decisions, wiki pages, meeting notes. If nothing matches, it says so. It does not invent a paragraph. It lives under Knowledgebase in the sidebar. Drafted answers and follow-up questions live in AI Chat.
Knowledgebase is available to Keldyn admins who also hold the Read organisational context (knowledge.read) permission. Google Drive can be connected by anyone who can manage integrations; Confluence, Notion, and Otter stay Keldyn-admin-only. If you do not see Knowledgebase, ask an admin.

​
What it’s for

Use Knowledgebase when you need the organisation’s own wording:
  • What does our access-control policy actually say?
  • Where did we record the decision to host customer data in the EU?
  • What did the last security review conclude?
Keldyn searches one index, then shows the passages it used and where they came from. If nothing matches, it says so instead of inventing a result. Knowledge connectors are not evidence integrations. Connecting Drive, Confluence, Notion, or Otter does not feed automated control evidence. Those tools keep the index current so people and agents can find what the organisation has already written down.

​
How it works

The first time you open Knowledgebase you see five stages. They are the actual path from a source file to a cited passage.
1

Ingest

Files from the Org Library, use cases, Drive, Confluence, Notion, Otter, and notes you type in land in one team index. Unchanged files are skipped using a content hash, so a quiet source does not get rewritten every sync.
2

Categorize

Each document gets a browse folder, a type, and an authority score. Approved Library wording ranks highest. Text is split on headings so a question about one section can retrieve that section, not the whole file.
3

Plan query

You ask in your own words. Simple questions search by keyword. Trickier ones use a planner. Names are matched to records that already exist in Keldyn: they are never guessed.
4

Retrieve and rank

Keyword and meaning search run together. Passages are scored by relevance, authority, and freshness, then narrowed to the best few.
5

Result

You get cited passages and a link back to the source. If nothing matches, Knowledgebase says so instead of inventing a result. For a drafted answer from those passages, open AI Chat.
Knowledgebase is a query builder: one question, filters on the page, passages back. AI Chat drafts an answer, keeps follow-ups, and can also read live GRC records (controls, tasks, use cases). Use Knowledgebase when you want to inspect the sources; use AI Chat when you want prose. The same search is available to signed-in users through the Keldyn MCP server.
What is never fetched, who can read an indexed file, and how to take something back out are on Knowledgebase privacy.

​
What feeds the index

Org Library

Documents you upload under Library. Approved Library wording ranks highest when two sources say different things.

Use case documents

Files attached to a use case during assessment or later in the workspace.

Connected sources

Folders, spaces, pages, and meetings you connect under Integrations → Knowledge.

Manual entries

Notes you type into the index browser when nothing else covers the gap.
Some questions never touch the document index. The org profile, org chart, team roster, use case register, architecture summaries, audit rollups, and third-party register are read live. The answer is the current record, not whatever the last sync stored.

​
Connect a knowledge source

Google Drive

Index a folder or shared drive. Docs, PDFs, and text files become searchable.

Confluence

Index selected Confluence Cloud spaces and their supported attachments.

Notion

Index the Notion pages you share with Keldyn during consent.

Otter.ai

Index meeting transcripts from chosen Otter channels.
1

Open Integrations

In the Keldyn web app, go to Integrations. Filter to Knowledge, or open a connector from the Knowledgebase index toolbar.
2

Choose what to expose

Pick the folder, spaces, pages, or channels that should be searchable. That selection is the access boundary. Keldyn does not inherit per-page or per-file permissions from the source.
3

Connect and wait for the first sync

Complete consent (or paste an API key for Otter). The first sync backfills everything in scope. After that, each connector refreshes on its own schedule. Use Sync now if you need an immediate pull.
4

Search Knowledgebase

Open Knowledgebase, ask a question, and follow a citation back to the file. Show in index opens that file’s folder in the browser.
Once a document is indexed, any Keldyn admin on the team who can open Knowledgebase can read it. That includes a Confluence page that was restricted to a few people, or a meeting that was shared with a small group. Only connect sources you are willing to make visible to those admins.

​
Search and browse

Knowledgebase is a query builder. One question, then the filters on the page: source, document type, authority, date range, how many passages. Follow-up chat is a different screen.
For a drafted answer from those passages, or to ask a follow-up, open AI Chat. Knowledgebase itself returns citations only.
The index browser (Knowledgebase → open a folder, or Show in index from a result) lists files the way a drive does: one folder level at a time, with breadcrumbs in the URL. From the browser you can:
  • Open a file to read its sections as collapsible passages. They start collapsed; expand one to load that passage. The section list scrolls inside the dialog.
  • New folder / New file to add a manual entry next to the documents that do not cover a gap.
  • Remove from index to stop Keldyn citing a file or folder. This does not delete the item from Drive, Confluence, Notion, Otter, or the Org Library. A later sync will not put a removed item back.

​
How Keldyn ranks what it finds

When two passages both match, Keldyn prefers the source the organisation has already signed off: A transcript is what someone said. A signed policy is what the organisation decided. When both match, the policy wins. A disagreement in the notes does not outrank it.

​
Keep the index current

Connected sources refresh on a schedule. Drive and Confluence check for changes every few minutes; Notion about every ten minutes; Otter about every three hours. Unchanged files are skipped. Deleted, archived, or unshared items are unpublished so they stop appearing in search, while an existing citation still resolves. Disconnecting a connector unpublishes that connector’s documents. It does not delete them from the source system. Reconnecting restores the index after the next sync.

​
Privacy, in one paragraph

Keldyn indexes what you connect, and nothing until you choose a folder, space, or channel. It never asks Drive, Confluence, Notion, or Otter for write access, and it never fetches Otter audio. Indexed text is readable by Keldyn admins on your team who hold Read organisational context. A Confluence page restricted to four people, or a meeting shared with three, does not stay that narrow after it is indexed. Read the privacy page before you connect anything personal or restricted.

​
Next steps

Knowledgebase privacy

What is collected, who can read it, and how to take it back out.

Google Drive

Choose a folder and connect a Google account.

Confluence

Select spaces and connect Confluence Cloud.

Notion

Share pages with Keldyn during Notion consent.

Otter.ai

Paste an Enterprise API key and choose channels.