Docs

Everything you need to start chunking.

Quickstart, concepts, API reference, and integration guides — written by the engineers who built CHUNKZA.

Get started

Install the CLI, ingest your first corpus, and export chunks to a vector store in under five minutes.

  • Quickstart
  • Install the CLI
  • Your first pipeline
  • Conceptual overview

Chunking strategies

Understand and configure every splitting strategy CHUNKZA ships.

  • Layout-aware segmentation
  • Semantic boundary detection
  • Parent-child chunking
  • Overlap & windows
  • Schema-aware tables

Diagnostic panel

Inspect boundaries, embeddings, and retrieval replay.

  • Boundary preview
  • Embedding distribution map
  • Strategy diff
  • Retrieval replay

Integrations

Export to your stack and pipe chunks into your RAG runtime.

  • Pinecone
  • Weaviate
  • Qdrant
  • pgvector
  • Chroma
  • Milvus
  • LangChain
  • LlamaIndex

API reference

Programmatic access to the full CHUNKZA pipeline.

  • Ingestion API
  • Chunking API
  • Export API
  • Webhooks
  • Authentication

Guides

End-to-end playbooks for common scenarios.

  • Tuning recall on enterprise KBs
  • Parent-child for copilots
  • Research paper chunking
  • Support article retrieval
Versioned policies

Chunking as code. Reproducible by design.

Every pipeline is a versioned YAML policy — diffable, reviewable, and replayable. No more "why did retrieval change between releases?"

  • Policy files live in your repo, reviewable in PRs
  • Versioned and reproducible across environments
  • Per-section overrides for special document types
  • Webhooks fire on every successful export
chunkza.yml
# chunkza.yml — a versioned chunking policy
version: 1.4
source:
  type: s3
  bucket: acme-kb
  prefix: docs/
strategy:
  layout_aware: true
  semantic_boundary:
    enabled: true
    model: chunkza-boundary-v2
  parent_child:
    enabled: true
    parent_max_tokens: 1024
    child_max_tokens: 256
  overlap_tokens: 48
metadata:
  inject: [section_path, source_uri, page, acl_roles]
export:
  target: pinecone
  index: acme-kb-v3
  embeddings: text-embedding-3-large

Ready to read the docs in depth?

Request access and we'll provision your workspace plus a walkthrough with a solutions engineer.