Knowledge Tools

The Knowledge provider performs semantic vector search across Zelaxy knowledge bases and lets a workflow add content to them by creating new documents or uploading individual chunks to existing documents. Use these tools when an agent needs to retrieve relevant context (RAG) or persist new text into a knowledge base during a run.

Overview

Operations

Configuration

knowledge_search

Calls POST /api/knowledge/search.

knowledge_upload_chunk

Calls POST /api/knowledge/{knowledgeBaseId}/documents/{documentId}/chunks.

knowledge_create_document

Calls POST /api/knowledge/{knowledgeBaseId}/documents. The content is encoded as a base64 data:text/plain URI and uploaded as a .txt file. Validation: the document name must be 1-255 characters and may not contain < > : " / \ | ? *; content must be non-empty and at most 1MB.

Note: none of these parameters are secrets. There is no apiKey, token, or password — the provider relies on internal Zelaxy authentication via the execution context.

Outputs

knowledge_search

• results (array) — Array of search results from the knowledge base. Each item is an object with id (string), content (string), documentId (string), chunkIndex (number), similarity (number), and metadata (object).
• query (string) — The search query that was executed.
• totalResults (number) — Total number of results found.
• cost (object, optional) — Cost information for the search operation.

knowledge_upload_chunk

• data (object) — Information about the uploaded chunk:
  • id (string) — Chunk ID
  • chunkIndex (number) — Index of the chunk within the document
  • content (string) — Content of the chunk
  • contentLength (number) — Length of the content in characters
  • tokenCount (number) — Number of tokens in the chunk
  • enabled (boolean) — Whether the chunk is enabled
  • createdAt (string) — Creation timestamp
  • updatedAt (string) — Last update timestamp
• message (string) — Success or error message describing the operation result.
• documentId (string) — ID of the document the chunk was added to.
• cost (object, optional) — Cost information for the upload operation.

knowledge_create_document

• data (object) — Information about the created document:
  • id (string) — Document ID
  • name (string) — Document name
  • type (string) — Document type (always document)
  • enabled (boolean) — Whether the document is enabled
  • createdAt (string) — Creation timestamp
  • updatedAt (string) — Last update timestamp
• message (string) — Success or error message describing the operation result.
• documentId (string) — ID of the created document.

YAML Example

knowledge_1:
  type: knowledge
  name: "Knowledge"
  inputs:
    operation: "knowledge_search"
    knowledgeBaseId: "{{KNOWLEDGE_BASE_ID}}"
    query: "How do I reset my password?"
    topK: 5
  connections:
    outgoing:
      - target: next-block-id

Tips

• No API key is needed. These tools hit Zelaxy's internal /api/knowledge/... endpoints and are authenticated automatically from the workflow execution context, so the workflow must run inside the workspace that owns the knowledge base.
• Pick the operation via the block's operation dropdown: search, upload_chunk, or create_document. The serializer maps these to the tool IDs knowledge_search, knowledge_upload_chunk, and knowledge_create_document respectively.
• For knowledge_search, the query is optional if you supply tagFilters — you can retrieve chunks purely by tag. topK is clamped to 1-100 and defaults to 10.
• For knowledge_create_document, keep the document name under 255 characters and avoid the filename characters < > : " / \ | ? *; content is capped at 1MB and stored as a .txt document.
• Reference a search result in a downstream block with double braces, e.g. {{knowledge_1.results}} or {{knowledge_1.totalResults}}. Use {{ENV_VAR}} for values stored as environment variables (such as the knowledge base ID).