A Knowledge Base enables your AI assistant to chat with your own content — from documents to entire knowledge libraries. Instead of manually searching, ask in plain English and get answers from your integrated sources.
Creating a Knowledge Base
- Click Knowledge in the sidebar.
- Click Create Knowledge Base.
- Choose a type:
Local RAG
Client-side embedding on your machine.
- Select Local type.
- Enter a Name for the knowledge base.
- Choose an Embedding Engine (e.g., OpenAI) and Model (e.g., text-embedding-3-large).
- Click Create.
StationOne RAG (Enterprise)
Server-side RAG managed by your organization.
- Select StationOne type.
- Select a Partition from the dropdown (loaded from your organization’s configuration).
- Click Create.
Ragie.ai (Third-Party)
Integration with Ragie’s RAG-as-a-Service.
- Select Ragie type.
- Enter your Ragie API Key.
- Enter a Name and optional Partition.
- Click Create.
Adding Knowledge
Local Files
- Open a Knowledge Base.
- Click + Add File to select individual files (PDF, DOCX, XLSX, PPTX, TXT, MD, CSV, JSON).
- Click + Add Folder to add all files from a directory recursively.
- Files are processed and indexed automatically. Status shows “In Progress” → “Done”.
Web Content
- Click + Add URL to index a single web page.
- Click + Add Website to crawl an entire sitemap.
- Enter the URL and confirm. Pages are fetched and indexed.
Notes
Create custom text entries with titles and descriptions for quick reference material.
Configuring a Knowledge Base
Click the gear icon (⚙) in the top-right of the Knowledge Base view. Settings include:
- Max Document Size: Maximum size per document (default: 16 million characters).
- Chunk Size: How documents are split for indexing (in characters).
- Chunk Overlap: Overlap between chunks for better context continuity.
- Relevance Cutoff: Minimum relevance score (0–1) for search results.
- Search Result Count: Number of results returned per query.
- Exclude Patterns: Glob patterns to skip certain files or folders (e.g.,
node_modules/**,*.tmp).
Click Save to apply, Reset to restore defaults.
Scanning for Changes
To detect new or modified documents:
- Click the sync icon (folder with arrows) in the Knowledge Base header.
- If the knowledge base has web sources (URLs or sitemaps), a dialog asks:
- Continue: Re-scan both files AND web sources.
- Files Only: Only check local file changes.
- Cancel: Abort the scan.
- A spinner shows while scanning. New or updated documents are re-indexed.
Searching a Knowledge Base
- Click the search icon in the Knowledge Base header.
- Enter a search query in the dialog.
- Results show matched content with relevance scores.
- Each result includes the source title, content excerpt, and score.
Using a Knowledge Base in Chat
- In the chat prompt area, click the + button.
- Select Knowledge from the menu.
- A list of available knowledge bases appears with checkboxes for multi-select.
- Select one or more knowledge bases.
- Selected KBs appear as badges in the prompt area.
- Type your question and send — the AI searches the selected knowledge bases and uses the results as context.
Using Knowledge Bases in Agents and Operators
In Agents
Each workflow step can have its own knowledge bases. In the Agent Editor → Workflow step → click the Knowledge button in the step footer → select knowledge bases.
In Operators
When creating or editing an operator, attach knowledge bases in the operator settings. All conversations with that operator will automatically use the attached knowledge bases.
In Skills
Skills can have knowledge bases attached directly. When the skill is used in a conversation, its knowledge bases are automatically included.
Multiple Document Repository Selection
StationOne supports selecting multiple knowledge bases simultaneously. In the prompt area or agent step, the Knowledge menu supports multi-select with checkboxes. You can combine results from different knowledge bases in a single query.