lucisearch 0.2.0

An embeddable, in-process search engine written in Rust

lucisearch

An embeddable, in-process search engine written in Rust with Python bindings.

Features

• Full-text search with BM25 scoring, phrase queries, fuzzy matching
• Structured queries — term, range, bool, geo_distance, geo_shape, nested
• Aggregations — terms, stats, histogram, date_histogram, cardinality, percentiles, and 12 more
• Sort by field with keyword, numeric, boolean columns
• Pagination — from/size and cursor-based search_after
• Document CRUD — get, delete, update by _id, delete by query
• Hybrid search — dense vector kNN + text via RRF fusion
• 2-8x faster than Elasticsearch on every supported feature
• No server, no cluster, no HTTP — just pip install and search

Quick Start

import luci

# Create an index
index = luci.Index.create("my_index.luci", {
    "properties": {
        "title": {"type": "text"},
        "body": {"type": "text"},
        "tag": {"type": "keyword"},
        "price": {"type": "float"},
    }
})

# Add documents (bulk for multiple, add for single)
index.bulk([
    {"title": "Hello World", "body": "Getting started", "tag": "intro", "price": 9.99},
    {"title": "Search Engine", "body": "Full-text search", "tag": "tech", "price": 29.99},
])

# Search (ES-compatible query DSL)
results = index.search({
    "query": {"match": {"title": "hello"}},
    "size": 10
})

for hit in results["hits"]:
    print(hit["_score"], hit["_source"]["title"])

# Sort by field
results = index.search({
    "query": {"match_all": {}},
    "sort": [{"price": "asc"}],
    "size": 5
})

# Aggregations
results = index.search({
    "query": {"match_all": {}},
    "aggs": {"by_tag": {"terms": {"field": "tag"}}},
    "size": 0
})

# CRUD operations
doc = index.get("my_doc_id")
index.update("my_doc_id", {"price": 19.99})
index.delete("my_doc_id")
index.delete_by_query({"term": {"tag": "draft"}})

Performance

Benchmarked on 100,000 documents (Apple M5 Max):

vs Elasticsearch

Query	Luci	ES	Speedup
sort_numeric	1074μs	2149μs	2.0x
fields_only	281μs	1225μs	4.4x
explain	404μs	1368μs	3.4x
from_0	369μs	1046μs	2.8x
rescore	362μs	1994μs	5.5x
source_disabled	275μs	897μs	3.3x
collapse	3332μs	5308μs	1.6x

vs Tantivy

Query	Luci	Tantivy	Speedup
match_phrase	190μs	409μs	2.1x
match_all	272μs	488μs	1.8x
from_1000	273μs	519μs	1.9x
no_source	68μs	140μs	2.0x
bool_must	280μs	362μs	1.3x

License

MIT