Knowledge

Knowledge holds the documents agents can draw on when answering — product catalogs, policies, reports, plans.

Open Knowledge from the Solutions sidebar to view all documents in your enterprise.

The knowledge list

The main page shows a searchable table of every uploaded document. Each row displays:

• Name — the original filename
• Type — the file format (e.g. pdf, docx, txt)
• Characters — how many characters the platform extracted, displayed in compact form (e.g. "12.3K"); this tells you how much usable text the document contains
• Last Updated on — when the document was last modified

Search filters the list by name as you type. Pagination controls sit below the table — configurable items per page (default 10) with a total items count.

Click any row to open its detail page.

Supported file types

Knowledge accepts the following formats, up to 5 MB per file:

Files that exceed the 5 MB limit or use an unsupported format are rejected with an error message at upload time.

Adding documents

Click + Add Knowledge (top right). A drag-and-drop upload zone opens — drop a file or click to browse. The modal shows the selected file's name and size before you confirm.

Click Upload to submit. The platform processes the file, extracts text, and splits it into chunks for retrieval. A success notification confirms when the upload completes, and the new document appears in the list.

Only one file can be uploaded at a time.

The document detail page

Click any row in the knowledge list to open its detail page. The page has two main areas: document information and the chunk viewer.

Document information

A panel displays:

Chunks

Below the document info, the platform lists the chunks the document was split into for retrieval. Each chunk is labeled (e.g. "Chunk-1", "Chunk-2") and shows its character count. Use the search box to filter chunks by content.

Click a chunk to expand it — a View more / View less toggle controls how much text is shown. Additional chunks load automatically as you scroll (infinite scroll).

Chunk parameters

Selecting a chunk reveals a parameters panel with retrieval and embedding details:

• Chunk name and Chunk length
• Avg paragraph length and Paragraphs count
• Embedding time — how long embedding took
• Embedding spend — the cost of the embedding operation

Managing documents

Rename

From the knowledge list's row menu (⋮) or the detail page header, select Rename to change the document's filename.

Download

Select Download from the row menu or detail page to retrieve the original uploaded file.

Delete

Select Delete from the row menu or detail page. Before deletion, the platform runs an impact check to identify agents and applications currently using this document.

The delete confirmation modal shows:

• Agents using this knowledge — a scrollable list with count
• Applications using this knowledge — a scrollable list with count
• A warning: deleting removes all data and may impact agents currently using it; this cannot be undone

You can still proceed, but doing so may break agents that rely on this document. If the impact check fails, a fallback warning lets you proceed at your own risk.

Using knowledge in an app

Uploading a document does not by itself change agent behavior. To make an agent use a knowledge document, attach it through the agent's Knowledge tools:

1. Open the agent configuration — either from the App Workspace's Agents tab or the enterprise Agents page.
2. In the agent's tool configuration, find the Knowledge tools section.
3. Click to attach a knowledge document. Each attachment requires a tool name and tool description (both required) — these help the agent understand when and how to use the document.
4. Once attached, the agent retrieves relevant passages from the document when a user's question relates to that content.

A single document can be attached to multiple agents, and each agent can have multiple knowledge tools.

Permissions

Access to knowledge is governed by role-based permissions:

Full access is granted to Enterprise Admins. View only is the default for Platform Users. When you have view-only access, mutation controls (upload, rename, delete) are hidden from the interface.

Tips

• Prefer text-based PDFs over scans — extraction quality drives answer quality.
• Check the Characters column after uploading; a low count relative to the document's size may indicate poor text extraction (e.g. from a scanned PDF).
• Keep documents current; re-upload when source material changes.
• Use meaningful filenames — they appear in the knowledge list and help you identify documents at a glance.
• Review the impact check before deleting a document to avoid breaking agents mid-conversation.