Skip to content
DocumentationTry Hanzo
Hybrid Search + RAG

Hanzo Search

Keyword and meaning in the same query

Push your documents once. Ask for a part number and full-text finds the exact string; describe the thing vaguely and vector search finds it anyway. One request runs both and returns one ranked list, so you do not have to decide in advance which kind of question a user will ask. Then point a model at the same index and get answers with the passages that support them attached.

Text
Typos and facets
Vector
Your embeddings
RRF
One merged rank
RAG
Answers with sources

Four ways to ask, one index

The same documents answer all four. Choose per query, not per deployment.

Full text

Typo tolerance, facets and filters. A user who types Kubernets still finds the page, and a user who wants only last quarter's docs can say so.

Vector

Bring your own embedding model — ours or anyone's — and search by meaning. Useful exactly when the user does not know the word your document uses.

Hybrid

Reciprocal Rank Fusion merges the two lists by rank rather than by score, so a keyword hit and a semantic hit compete fairly without a threshold you have to tune.

RAG chat

The retrieved passages go to a model and come back as an answer with links to what it read. Streamed, so the first sentence arrives before the last one is written.

How it works

Content in, ranked results out, and a model over the top if you want one.

Step 01

Index your content

Push documents at build time or over the REST API. Text fields become searchable as they land; embeddings go to the vector store beside them. Adding a vector field later does not force a text reindex.

Step 02

Query it from anywhere

A React hook for search-as-you-type, the REST API for everything else, and an MCP tool so an agent can search the same index a user searches. One index, three callers.

Step 03

Let people ask instead

Turn on RAG chat and a question gets an answer built from the passages that were retrieved, each one linked. If the index has nothing, the answer says so rather than inventing one.

Works with any docs framework

Native support for Hanzo Docs, Fumadocs, Nextra, and Docusaurus. REST API for everything else.

Hanzo DocsFumadocsNextraDocusaurusREST API

Pricing

Pay for hosted search, vector, and AI. Self-host Meilisearch free forever.

Build

$99/mo
  • 5 projects
  • 10k docs per project
  • 2,500 AI chat sessions
  • Unlimited searches
  • React hook + REST API
Get Started

Scale

$1,450/mo
  • 25 projects
  • 100k docs per project
  • 50,000 AI chat sessions
  • Unlimited searches
  • Priority support
Get Started

Enterprise

Custom
  • Unlimited projects
  • Unlimited docs
  • Unlimited AI chat sessions
  • Unlimited searches
  • Dedicated support + SLA
Contact Sales

Meilisearch is open source. Self-host free forever. Pay only for our hosted API, vector search, and AI features.

Open Source Revenue Sharing

Up to 5% of compute goes back to open source

Every deployment is SBOM-verified. Contributors to Meilisearch earn a share of compute revenue — transparent, on-chain, and customizable by the community.

Hanzo Search

Make your content searchable and conversational

Index once, search everywhere.
Full-text, vector, hybrid, and RAG -- all from one API.

Open source

License: Apache-2.0hanzoai

Get Search

AI-powered search