Knowledge Base

About Knowledge Base

Choose an entry point

TaskEntry
Store and preview original filesLibrary
Maintain searchable text and metadataKnowledge Base content page
Parse files into knowledgeLibrary parsing configuration
Compare retrieval resultsKnowledge retrieval test
Use knowledge in a processWorkflow knowledge retrieval node

Current knowledge base types

The current creation form is fixed to Pro. Open existing FAQ or Chunk knowledge bases from the list.

Content and indexes

Pro content includes a vector index, a tokenized index, and Metadata. Vector retrieval matches meaning; tokenized retrieval matches keywords. Metadata holds information such as source and category and can be used for filtering. Both indexes may use the same text.

Before you start

Define the content scope, embedding model, metadata schema, and readers and maintainers. The model and schema cannot be freely changed after creation. Validate the design with a small mock sample. Continue with Create and manage knowledge bases.