Choose Your Path

Most users start from a practical situation, not from an API name.

By situation

Situation	Start here
I want one working search first	Installation → Quickstart
I start from raw text	Text Encoders → Usage Patterns
I already have token vectors	Quickstart → Python API
I already have a vector database	Storage + Search → Vector Databases
I need repeated-query speed	Usage Patterns
I need one runtime for many same-process callers	Hosted Engine Python

By API shape

Situation	First API to look at
raw text in, search results out	kayak.open_text_retriever(...)
precomputed vectors in, exact search out	kayak.query(...), kayak.documents(...).pack(), kayak.search(...)
database keeps persistence	kayak.open_store(...) plus store.load_index(...)
repeated queries on one fixed slice	store.load_index(...) once, then kayak.search_batch(...)
same-snapshot serving in one process	prepare_exact_search_runtime(...)

Shortest reading order

1. Installation — verify the exact backend is available before anything else
2. Quickstart — run one working example before tuning adapters, layouts, or plans
3. Usage Patterns — pick the right long-term API shape

Common confusions

If you are thinking...	Open...
"How do I make it use Mojo?"	Installation
"Can I pass a Hugging Face model?"	Text Encoders
"Can I keep Postgres, Qdrant, LanceDB, Weaviate, or Chroma?"	Storage + Search
"What is the fastest repeated-query path?"	Usage Patterns
"I need runnable artifacts, not prose."	Examples