Knowledge

The Knowledge block connects your workflow to a knowledge base, enabling retrieval-augmented generation (RAG) by performing vector similarity searches, uploading new chunks, and creating documents -- all at runtime.

Overview

Search: Run a semantic similarity query against your knowledge base and retrieve the most relevant chunks

Upload Chunk: Append a new text chunk to an existing document so it becomes searchable in future queries

Create Document: Add a brand-new document (with optional tags) to the knowledge base

Operations

Performs a vector similarity search and returns the top matching chunks.

Field	Required	Description
Knowledge Base	Yes	The knowledge base to search
Search Query	No	Natural-language query. Optional when using tag filters alone.
Number of Results	No	How many chunks to return (default: 10)
Tag Filters	No	Narrow results to documents that match specific tags

The search query is optional -- you can filter purely by tags to retrieve all chunks under a certain category.

Appends a text chunk to an existing document inside the knowledge base. The chunk is automatically embedded and becomes available for future searches.

Field	Required	Description
Knowledge Base	Yes	Target knowledge base
Document	Yes	The document to add the chunk to
Chunk Content	Yes	The text content to upload

Creates a new document in the knowledge base. The content is chunked, embedded, and indexed.

Field	Required	Description
Knowledge Base	Yes	Target knowledge base
Document Name	Yes	Name for the new document
Document Content	Yes	Full text content
Document Tags	No	Tags for categorization and filtering

Configuration

Operation

Select...

Knowledge Base*

Select knowledge base

Search Query

Enter your search query (optional when using tag filters)

Number of Results

Enter number of results (default: 10)

Tag Filters

Add tag filters

Document*

Select document

Chunk Content*

Enter the chunk content to upload

Document Name*

Enter document name

Document Content*

Enter the document content

Document Tags

Outputs

Output	Type	Description
`results`	json	Array of matching chunks (search) or confirmation data (write operations)
`query`	string	The query that was used
`totalResults`	number	Count of results returned

Best Practices

Use tag filters to scope searches -- tags let you partition a knowledge base by topic, customer, or category without creating separate bases.
Keep chunks focused -- shorter, topically coherent chunks produce better similarity matches than long, multi-topic text.
Pair with an Agent block -- pipe the search results into an Agent's context to build a RAG pipeline.

The search query field is optional. When you only need documents that match certain tags, leave the query empty and use tag filters instead.