Knowledge Base
Manage document collections for RAG (Retrieval-Augmented Generation). Upload documents, then connect them to AI Agent nodes so the AI can search and reference your content at runtime.
Reach it via Upload in the top navigation bar, or go directly to /knowledge-base.
How It Works
Documents you upload are chunked into pieces, converted to vector embeddings, and stored in a searchable index. When an AI Agent node uses a Knowledge Base Tool, the tool searches the index at runtime and returns the most relevant chunks as context for the AI to answer from.
The AI does not memorize your documents — it searches them fresh on each execution.
Collections
The Knowledge Base page lists your Collections — named groups of documents. Each collection card shows:
- Collection name
- Number of documents
- Last updated time
- Visibility badge: Private (only you), Organization (everyone in your org), or Shared (specific users or groups)
Use the Search box at the top to filter collections by name.
Creating a Collection
Click New Collection. A dialog opens with:
| Field | Required | Description |
|---|---|---|
| Collection Name | Yes | A descriptive name (e.g., "Support Documentation", "HR Policies") |
| Visibility | No | Private (default) or Organization |
Click Create. The empty collection opens immediately, ready for documents.
You can change sharing settings later from the collection menu.
Adding Documents
Open a collection. The Documents tab shows all documents in that collection, with action buttons at the top.
There are four ways to add content:
Upload a File
Click Upload (↑). Select one or more files from your computer.
Supported file types: .pdf, .docx, .doc, .txt, .md
After uploading, each document goes through processing stages:
| Status | Meaning |
|---|---|
| Pending | Queued for processing |
| Scraping | Text is being extracted |
| Vectorizing | Embeddings are being generated |
| Completed | Ready for search |
| Failed | Processing failed (error message shown) |
The page polls automatically every 3 seconds and updates status badges until all documents finish processing.
Paste Text
Click Text (document icon). A dialog opens with:
- Document Name (required)
- Content text area (required) — paste any text content
The text is automatically chunked and embedded.
Scrape a URL
Click URL (link icon). A dialog opens with:
- URL (required) — the page to scrape (e.g.,
https://docs.example.com/guide) - Document Name (optional) — defaults to the URL if left empty
The system fetches and indexes the page content.
Import from Storage
Click Storage (folder icon) to import from a connected cloud storage provider (e.g., Google Drive). A file browser opens so you can navigate and select files from your connected drive.
Document Actions
Each document card shows its name, source (Drive, S3, or local), upload date, and status badge.
| Action | When available |
|---|---|
| Edit (pencil) | When status is Completed — rename the document |
| Delete (trash) | Any time — removes the document and its embeddings |
Click Refresh (↺) to reload the document list from the server.
Click Sync to re-synchronize documents that came from cloud storage.
Query & Search Tab
Click the Query & Search tab (magnifying glass icon) to test your collection's search before using it in a workflow.
| Field | Description |
|---|---|
| Search Query | Type your question or search phrase; press Enter to search |
| Results | Number of chunks to return (1–20, default: 5) |
Results appear below with:
- Source — the document the chunk came from
- Relevance — High (green), Medium (orange), or Low (red) based on similarity score
- Content snippet — the first 300 characters of the matching chunk
- Additional metadata from the document
Use this to verify that your documents are indexed correctly and that relevant content is returned for the queries your AI Agent will make.
Collection Sharing
Click Share (share icon) on any collection you created. The sharing dialog lets you:
- Share with entire organization — all users get read access
- Share with specific users — search by name or email and add individuals
- Share with groups — select from available user groups
As the creator, you always retain full access regardless of sharing settings.
Using a Collection in Workflows
- Note the collection name (or ID shown in the URL when viewing the collection)
- In the Workflow Designer, add a Knowledge Base Tool node from the Components tab (AI category)
- In Step Settings, select your collection
- Connect the Knowledge Base Tool to an AI Agent node using a tool connection
- The AI Agent will automatically use the collection to search for relevant information when answering prompts
Editing a Collection
Click the Edit (pencil) icon on any collection you own. You can update the collection name. Visibility is set at creation and managed separately via the Share dialog.
Deleting a Collection
Click Delete (trash icon) on a collection. A confirmation dialog appears before deletion. Deleting a collection removes all its documents and embeddings permanently.
Note: If workflows reference a deleted collection via a Knowledge Base Tool, those nodes will fail at runtime.