> For the complete documentation index, see [llms.txt](https://docs.uptiq.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.uptiq.ai/console/knowledge.md).

# Knowledge

Build queryable knowledge from your documents, then attach it to agents and apps.

Knowledge turns your documents into something an agent can search. You describe what the knowledge should hold, point it at source files, and Qore processes them into a table of contents, tags, and embeddings that an agent can retrieve from during a conversation.

Open it at **Console → Knowledge**. It works inside the project you have selected.

## At a glance

| Task                               | Start with                               | Result                                       |
| ---------------------------------- | ---------------------------------------- | -------------------------------------------- |
| Build knowledge from a description | The **Create New Knowledge** prompt      | A knowledge base, set up for you             |
| Group source files                 | **Knowledge Datasets**                   | A named dataset that tracks ingestion        |
| Inspect what was extracted         | A knowledge base → **Content**, **Tags** | The table of contents and the tags per chunk |
| Use what you built                 | **Try Knowledge**, or attach it          | Grounded answers, in the chat or in an agent |

Two things are easy to confuse:

**A dataset** is a named grouping of source files. It tracks ingestion and nothing more — it does not interpret what the files say. See [Knowledge Datasets](/console/knowledge/knowledge-datasets.md).

**A knowledge base** is what Qore builds *from* those files: a table of contents, tags, and embeddings an agent can query. That is what this page is about.

## Create knowledge

Knowledge is created by description rather than by form. The landing page asks **"Describe what this knowledge should hold…"**, and Qore sets up the knowledge base from your answer — the page's own subtitle is *"Describe what it should know and I will set it up."*

Starter suggestions sit under the prompt — **Loan Policy KB**, **Member Support**, **Product & Rates**, **Training & Onboarding**. Select one to begin from that shape instead of a blank prompt.

Use the **+** button to attach files as you describe it, or the **Datasets** selector to build from datasets you already have.

{% hint style="info" %}
You do not have to have a dataset first. Describing the knowledge and attaching files in one step is the normal path; datasets matter when you want to reuse the same source files across several knowledge bases, or when you are crawling a website.
{% endhint %}

## Find your knowledge

**MY KNOWLEDGE** lists what this project holds. Each row shows the name, its description, and when it was last updated.

* **Search knowledge…** filters by name and description. It does not search inside the documents — for that, build the knowledge base and query it.
* The list and grid icons switch layout.
* **▷** opens the knowledge base and lets you query it.
* The **⋮** menu holds the per-item actions, including delete.

Knowledge built by crawling a website is named for its source, in the form `Web — en.wikipedia.org — 9c1ad5f5`.

**All Datasets** in the top right opens [Knowledge Datasets](/console/knowledge/knowledge-datasets.md). **Legacy** opens the retiring options — see [Migrate legacy knowledge](#migrate-legacy-knowledge).

## Work on one knowledge base

Opening a knowledge base gives you three panes.

**The chat, on the left.** It opens with *"Hi! I can help you build or refine this knowledge. What would you like to do?"* Use it to add sources or change what the knowledge covers, without leaving the page.

**The sources, in the middle.** **Documents** lists the files in the knowledge base; **Datasets** lists the datasets it draws from.

**What was extracted, on the right.** Three tabs — **Content**, **Tags** and **Preprocessed Files** — show what Qore made of each document. These are covered on their own page:

{% content-ref url="/pages/tXEHJV59aB5WZpmY6arb" %}
[Document processing](/console/knowledge/document-processing.md)
{% endcontent-ref %}

Along the top: a **dataset count** (for example, `1 dataset`), **+ Attach** to add another source, and **▷ Try Knowledge** to query it.

### Try Knowledge

**Try Knowledge** asks a question against the knowledge base and returns an answer grounded in its documents, with citations identifying which ones supported it. Use it to check that a knowledge base answers what you expect before attaching it to an agent.

## When processing fails

Building a knowledge base runs a pipeline: ingestion, table-of-contents generation, tagging, then embeddings. A failure names the stage it stopped at.

The most common one is provider throttling during table-of-contents generation:

> **The provider rate limited the request.**\
> **Failed stage: TOC\_GENERATION**

**Retry** re-runs only the failed stage. **Rebuild** runs every stage from ingestion onward. Try Retry first — most of these failures are transient — and rebuild only when retrying does not clear it.

{% hint style="warning" %}
**Wait before retrying.** An immediate retry usually meets the same rate limit, and every attempt consumes provider quota. Repeated failures across knowledge bases usually mean a tenant-wide throttle rather than a problem with your files — check with your administrator before continuing.
{% endhint %}

A file must reach **Completed** before it contributes. A knowledge base built while files are still `in-progress`, `queued`, or `failed` will be missing their content.

## Migrate legacy knowledge

{% hint style="danger" %}
**Support for the legacy options ends on 30 September 2026**, as stated in the product. Migrate before that date. What happens to content left in a legacy option after it is unavailable has not been confirmed — if you have files you cannot move in time, raise it with your platform administrator now rather than after the date.
{% endhint %}

**Legacy** in the top right holds the options being retired. There is no automatic migration — you recreate the knowledge base from the same source files and stop using the old option.

| Legacy option      | Replacement                                                    | How to migrate                                  |
| ------------------ | -------------------------------------------------------------- | ----------------------------------------------- |
| `Legacy Dataset`   | [Knowledge Datasets](/console/knowledge/knowledge-datasets.md) | Recreate the dataset with the same files.       |
| `RAG Data Store`   | [Knowledge Datasets](/console/knowledge/knowledge-datasets.md) | Recreate the dataset with the same files.       |
| `Legacy Knowledge` | Knowledge                                                      | Rebuild the knowledge base from the same files. |

{% hint style="danger" %}
**Deleting a knowledge base cannot be undone here.** It removes the generated structure — the table of contents, tags, and embeddings. The source files stay in their dataset, so you can rebuild, but the knowledge base itself does not come back.
{% endhint %}

## Notes and limitations

* **You cannot link directly to one knowledge base.** The URL does not change as you open one, so share the Knowledge link and say which to open.
* **Dataset and file deletion are restricted.** Ask your administrator to remove a dataset or an individual file.
* **Search is metadata only.** Name and description, not document contents.

### Related

<table data-view="cards"><thead><tr><th>Title</th><th>Description</th><th data-card-target data-type="content-ref">Target</th></tr></thead><tbody><tr><td><strong>Knowledge Datasets</strong></td><td>Group source files, add them from uploads, drives, or a website crawl.</td><td><a href="/console/knowledge/knowledge-datasets.md">Knowledge Datasets</a></td></tr><tr><td><strong>Document processing</strong></td><td>The table of contents, tags, and preprocessed files Qore extracts.</td><td><a href="/console/knowledge/document-processing.md">Document processing</a></td></tr><tr><td><strong>Attaching Knowledge to an Agent</strong></td><td>Give an agent access to a knowledge base.</td><td><a href="/console/agent-builder/build/attaching-knowledge-to-an-agent.md">Attaching Knowledge to an Agent</a></td></tr><tr><td><strong>Knowledge &#x26; Context</strong></td><td>The concept behind retrieval, and when to reach for it.</td><td><a href="/core-concepts/knowledge-and-context.md">Knowledge &amp; Context</a></td></tr></tbody></table>


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter:

```
GET https://docs.uptiq.ai/console/knowledge.md?ask=<question>&goal=<endgoal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.