# Hybrid Search with Manticore Search

Combine full-text and vector retrieval in Manticore Search for more relevant results.

## What is Hybrid Search?

Hybrid search combines full-text and vector search in a single query. It helps you handle both exact keywords and semantic meaning at the same time, so results can match identifiers, product names, and error codes while still understanding natural-language intent.

## When to use Hybrid Search?

- Improving search relevance for natural-language queries
- Building RAG pipelines that need stronger retrieval
- Combining exact keyword matching with semantic similarity
- Searching product catalogs with names, SKUs, and descriptions
- Searching support knowledge bases with error codes and symptoms
- Handling queries that mix identifiers with descriptive text
- Improving content discovery in large document collections
- Ranking search results using both lexical precision and semantic recall
- Searching multilingual or synonym-rich content
- Implementing AI-powered search without losing exact-match behavior


## Why Manticore Search is good for Hybrid Search

- Manticore Search supports hybrid search natively, combining `MATCH()` and `KNN()` in one query.
- It works through both SQL and JSON interfaces, making integration straightforward.
- You can keep exact-match precision for identifiers while adding semantic understanding for natural language.
- Multiple KNN subqueries can be used in the same hybrid query when needed.
- Manticore lets you build hybrid retrieval without adding a separate search engine.


## Pros

- Combines exact matching with semantic understanding
- Improves relevance for real-world queries that mix keywords and intent
- Supports both SQL and JSON interfaces
- Works with vector fields and auto-embeddings
- Can use multiple KNN subqueries in one hybrid query
- Reduces the need for custom query orchestration


## Cons

- Requires vector data or auto-embeddings in addition to full-text fields
- May need tuning to balance lexical and semantic signals
- Consumes more resources than plain full-text search alone


## How to get started

### Install Manticore Search

- Visit the official Manticore Search website: https://manticoresearch.com/
- Follow the [installation instructions](/install/) for your operating system
- Alternatively, use Docker: `docker pull manticoresearch/manticore`


### Prepare your data

- Create a table with text fields for full-text search
- Add vector fields or configure auto-embeddings for semantic retrieval
- Index your documents so both text and vector signals are available


### Run your first hybrid query

- Use `MATCH()` together with `KNN()` in SQL, or the equivalent JSON request
- Test queries that combine exact terms with natural-language intent
- Review the fused results and confirm they improve relevance


### Tune your hybrid search

- Adjust your full-text query and vector query inputs
- Apply attribute filters to keep results within the right category or tenant
- Experiment with multiple KNN subqueries if your data has several semantic dimensions


### Enjoy Hybrid Search

- Use Manticore Search to deliver more relevant search results
- Feel free to [create an issue](https://github.com/manticoresoftware/manticoresearch/) if you encounter any problems
- Consider our [professional services](/services/) for advanced implementations


## Resources

- [Hybrid Search documentation](https://manual.manticoresearch.com/Searching/Hybrid_search)
- [Hybrid Search blog post](/blog/hybrid-search/)
- [Vector Search documentation](https://manual.manticoresearch.com/Searching/KNN)
- [KNN Prefiltering blog post](/blog/knn-prefiltering/)
- [Auto-embeddings blog post](/blog/auto-embeddings/)


## Learn more about other use cases

Do not stop here when learning when you need **AI Database** and how **Manticore Search** can help you. There are many other use cases that you can explore.


## Install Manticore Search for Hybrid Search

Try Manticore Search for implementing Hybrid Search in your applications today!

[Install Now](/install/)