Compozy Symbol
ALPHA
DocsInstallGetting StartedFeaturesPricing
CoreCLI ReferenceSchema DefinitionAPI Reference

Knowledge Base Schema

JSON Schema for knowledge base definitions

Knowledge Base Definition

Schema Definition
BaseConfig declares a knowledge base and governs how it is ingested and retrieved.
chunking
object

ChunkingConfig tunes how documents are split before embedding.

description
string
embedder
string
id
string
ingest
string
metadata
object

MetadataConfig carries optional descriptive metadata for knowledge bases.

preprocess
object

PreprocessConfig configures preprocessing steps applied to raw content.

retrieval
object

RetrievalConfig manages how stored chunks are queried and injected into prompts.

sources
array

vector_db
string

Example

knowledge_bases:
  - id: quickstart_docs
    embedder: openai_default
    vector_db: filesystem_faststart
    sources:
      - kind: markdown_glob
        glob: "docs/**/*.md"
    chunking:
      strategy: token
      size: 512
      overlap: 64
    retrieval:
      top_k: 5
      min_score: 0.2
      filters:
        tag: "policy"
  • sources supports markdown_glob and url in the MVP. Additional kinds appear when feature flags are enabled.
  • Omit size, overlap, or top_k to inherit defaults from config.knowledge.
  • Use metadata under sources[].metadata to attach tags for filtered retrieval.

Resources

Knowledge Configuration Guide
Step-by-step instructions for defining knowledge bases and bindings.
Knowledge Ingestion
Operational guide for enumerating sources and triggering ingestion.

Vector Database Schema

JSON Schema for vector store configuration

Knowledge Binding Schema

JSON Schema for workflow, task, and agent knowledge bindings