Skip to main content
cesteralDocs
FAQContact

KB Ingest Node

Ingest records into a knowledge base for vector search

Last updated 8 Oct 2026

The KB Ingest node indexes each record's content into a knowledge base, chunking and embedding it so it becomes searchable via semantic (vector) search in downstream RAG Query nodes or the AI assistant.

When to Use

Use KB Ingest to build or update a knowledge base from automation data — for example, ingesting product descriptions, company profiles, or research findings that the AI assistant should be able to retrieve.

Configuration

FieldTypeDescription
knowledgeBaseIdstringID of the target knowledge base. Required
contentColumnstringRecord column containing the text to embed and index. Required
metadataColumnsstring[]Additional columns to store as document metadata
chunkStrategyrecursive | fixed | markdownHow content is split into chunks (defaults to the knowledge base's configured strategy)
chunkSizenumberMax chunk size (defaults to the knowledge base's setting)
chunkOverlapPctnumberOverlap between chunks as a percentage (defaults to the knowledge base's setting)

Example

Index company research reports:

knowledgeBaseId: kb_company_research
contentColumn: research_summary
metadataColumns: [industry, region, tier, company_id]
chunkStrategy: recursive
chunkSize: 800

Notes

  • Embedding uses the knowledge base's configured embedding provider and credential — the knowledge base must have an embedding credential set.
  • Knowledge bases require a Professional plan; ingest fails with a clear message otherwise.
  • Each input row is passed through with ingest status columns added: _kb_ingest_success, _kb_ingest_chunks, and _kb_ingest_doc_id, plus the standard _success / _error batch columns.
  • Rows are processed in small batches with a short delay between them for rate limiting; rows with empty content are marked failed with EMPTY_CONTENT.
  • Embedding calls are external API operations, so the node runs on the governed external rail.

Related articles

Node Reference Complete reference for all automation node types, organized by categoryRAG Query Node Retrieve from a knowledge base and optionally generate an answer for each recordMemory Write Node Write or update project memory (Tier 1) from automation data

Still stuck? Browse the FAQ or contact us.