> ## Documentation Index
> Fetch the complete documentation index at: https://docs.unify.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Knowledge

> The durable facts and rules your assistant relies on

Alongside the playbooks and skills sits the third library: **Knowledge** —
in the Console's words, *"durable facts and rules your teammate relies
on."* Where [Guidance](/learning/guidance-and-functions) captures *how* to
do things, Knowledge captures *what is true*: the reference material your
assistant consults so its answers rest on your reality, not general
world-knowledge.

## What lives here

* **Facts about your business** — products and pricing, org structure,
  who's responsible for what, key dates and numbers.
* **Policies and rules** — the refund policy, the escalation thresholds,
  compliance constraints the assistant must respect.
* **Digested documents** — the durable takeaways from material you've
  [shared](/learning/teaching#hand-it-the-reading): process docs, specs,
  contracts, style guides.
* **Reference data it gathered itself** — facts it looked up and confirmed
  while working, kept so it doesn't have to re-research them.

## How it fills up

Like the other libraries, Knowledge grows through natural interaction
rather than data entry:

* **Send it the reading.** Documents, links, and files you'd give a new
  hire get digested — the durable facts land here.
* **Tell it things.** "Our enterprise tier starts at 50 seats" or "Priya
  owns vendor relationships now" becomes a knowledge entry it will use
  from then on.
* **It learns on the job.** Facts established during work — confirmed
  details, resolved ambiguities — get written down so the next job starts
  from what's known.

When your assistant then answers "what's our refund window?" or drafts a
reply that cites the right policy, this library is where that certainty
comes from — searched at working time alongside guidance and skills.

## The Knowledge pane

In the Console's Brain section, **Knowledge** uses the same split document
library as Guidance: entries listed on the left (title, tags, scope badge,
dates), the full entry rendered as a readable document on the right, with a
divider you can drag. Search, filter by **scope or tag**, and use **Copy**
to take an entry's text elsewhere.

The scope badge matters here just as it does for playbooks: an entry is
either **personal** to this assistant or shared with a
[team](/teams/shared-context) — which is how a whole unit keeps its
reference facts in sync instead of each assistant maintaining its own
version of the truth.

## Keeping it accurate

Knowledge is only as useful as it is current, and maintenance is
conversational:

* **Ask what it knows** — "what do you know about our pricing?" and it
  answers from the library, so you can spot gaps or stale entries.
* **Correct it** — "the enterprise tier moved to 40 seats last month"
  updates the entry.
* **Retire it** — "forget the old travel policy, it's been replaced"
  removes what no longer holds.

## Knowledge, Guidance, or Functions?

The three libraries divide cleanly, and knowing which is which makes your
teaching land in the right place:

| Library       | Holds             | Example                                                      |
| ------------- | ----------------- | ------------------------------------------------------------ |
| **Knowledge** | What is *true*    | "Invoices over £10k need director sign-off"                  |
| **Guidance**  | How to *act*      | "When processing invoices, check the PO number first, then…" |
| **Functions** | What it can *run* | The invoice-validation workflow itself, as reusable code     |

One fact, one playbook, one skill — often born from the same piece of work,
each filed where it belongs.