Headband Documentation

What Headband is, core concepts, and how to get your first index searching in minutes.

End-to-end paths for Next.js, a plain JS page, Node and Python backends, and Drizzle sync, each with a working example app.

API key types, the bearer header, and what admin vs. search keys can do.

Create, list, fetch, and delete indexes across Meilisearch, Typesense, Elasticsearch and Visor.

Read and update per-index search settings and see what each engine supports.

Add, list, fetch, delete, and bulk-import documents into an index.

Search endpoint parameters, filters, facets, sorting, and highlighting.

Run multiple search queries against one or more indexes in a single request.

Poll asynchronous write tasks by UID until they succeed or fail.

Retrieval-augmented search: the AI search endpoint and its response format.

Combine keyword and vector search in a single hybrid query.

Set up the Headband MCP server and see the tools and resources it exposes.

Serve an llms.txt file describing your indexes for LLM crawlers.

Framework recipes, and sync connector scripts for Postgres, MySQL, Shopify, Stripe and BataDB.

Split traffic between two engines or two settings: hosted assignment, tagged searches, outcome events and results.

Install the React SDK, use the provider, hooks, and components, and see a full example.

Feature-by-feature compatibility matrix and sizing guidance across engines.

The same data in Meilisearch, Typesense, Elasticsearch and Visor side by side: equivalent settings and what does not carry over.

ZVN's own engine: instant and retrieval search modes on one index, where it runs, and what it supports.

Private Meilisearch, Typesense, Elasticsearch and Visor servers: create, connect, keys, cold starts and limits.

Keep a search index in step with your Postgres tables: sources, setup SQL, definitions, pause, rebuild and delete. Early access.

Install the headband command, log in with a device code, and manage indexes, engines, Sync and schema from your terminal.

The OpenAPI 3.1 spec for every /v1 endpoint, and what stable means: additive changes, deprecations and notice.

Error response shape, status codes and request ids returned by the API.