blevesearch/bleve

What is bleve?

bleve is a search library written in the Go programming language. It lets developers add full-text search to their applications by indexing data (organizing it for fast retrieval) and then querying that index (searching it).

It can index any Go data structure or JSON data, and it ships with sensible defaults that you can customize if needed.

Core capabilities

Field types it can index

bleve understands many kinds of data fields:

text (words/sentences)
number (numeric values)
datetime (dates and times)
boolean (true/false)
geopoint and geoshape (locations and areas on a map)
IP (internet addresses)
vector (mathematical representations used for similarity search)

Query types it supports

You can search using:

Basic matches: term, phrase, match, match_phrase, prefix, regexp, wildcard, fuzzy
Ranges: term range, numeric range, date range, boolean field
Compound queries: combine conditions with conjuncts (all must match), disjuncts (any can match), and boolean logic (must / should / must_not)
Query string syntax: a simple text-based way to write queries
Geo spatial search: find things near a location or within a shape
Vector search: approximate k-nearest neighbors (find similar items by vector)
Synonym search: match words with the same meaning
Hierarchical nested search: search structured, nested data

Scoring and ranking

bleve ranks results using two well-known scoring models: tf-idf and BM25. It also supports hybrid search — combining exact keyword matches with semantic (meaning-based) matches using two fusion methods: RRF (Reciprocal Rank Fusion) and RSF (Relative Score Fusion).

Other features

Result pagination: page through large result sets
Query time boosting: give extra weight to certain terms or fields during a search
Match highlighting: show which parts of a document matched, with fragments of text
Aggregations/faceting: group results by terms, numeric ranges, or date ranges (useful for filters and summaries)

How it works (code examples)

Indexing — store a document:

message := struct {
    Id   string
    From string
    Body string
}{
    Id:   "example",
    From: "xyz@couchbase.com",
    Body: "bleve indexing is easy",
}

mapping := bleve.NewIndexMapping()
index, err := bleve.New("example.bleve", mapping)
if err != nil {
    panic(err)
}
index.Index(message.Id, message)

Querying — search the index:

index, _ := bleve.Open("example.bleve")
query := bleve.NewQueryStringQuery("bleve")
searchRequest := bleve.NewSearchRequest(query)
searchResult, _ := index.Search(searchRequest)

Command-line tool

bleve also ships a CLI (command-line interface) you can install with go install github.com/blevesearch/bleve/v2/cmd/bleve@latest. It lets you interact with a bleve index from the terminal. Commands include:

create — make a new index
index — add files to the index
query — search the index
count — count documents
dump — show index contents
check — verify index integrity
fields — list fields
mapping — show the mapping used
dictionary — print the term dictionary for a field
bulk — load many JSON documents at once
registry — list compiled components
scorch — interact with a scorch index (the underlying storage engine)

Text analysis

bleve includes general-purpose analyzers (customizable) plus pre-built text analyzers for many languages: Arabic, Bulgarian, Catalan, Chinese-Japanese-Korean, Kurdish, Danish, German, Greek, English, Spanish (Castilian and Galician), Basque, Persian, Finnish, French, Gaelic, Hindi, Croatian, Hungarian, Armenian, Indonesian, Italian, Dutch, Norwegian, Polish, Portuguese, Romanian, Russian, Swedish, and Turkish.

There's also a Text Analysis Wizard at bleveanalysis.couchbase.com to help you experiment with analyzers.

Getting help and license

Issues/discussion: GitHub issues and a Google group
License: Apache License Version 2.0

In short: bleve is a flexible, feature-rich search library for Go developers, with a CLI for quick tasks and broad language support for text analysis.