Algolia
Hosted search-as-a-service platform delivering real-time, full-text search for apps and websites
SeekStorm
Sub-millisecond full-text search library with real-time indexing
SeekStorm delivers sub-millisecond lexical search, true real-time indexing, and advanced features such as faceted, geo, and sorted results, all powered by Rust and SIMD acceleration for single-machine scalability.
Overview
High‑Performance Keyword Search
SeekStorm is a Rust‑based full‑text search library that achieves sub‑millisecond query latency through SIMD‑accelerated ranking and multithreaded processing. It supports true real‑time incremental indexing, meaning every document becomes searchable the moment it is ingested, without costly segment merges or commit delays.
Rich Feature Set & Easy Deployment
Beyond basic BM25 ranking, SeekStorm offers faceted counting and filtering, geo‑proximity queries, multi‑field sorting, and support for 18 languages with stemming and custom tokenizers (including Chinese segmentation). The library can be embedded directly in applications or run as a multi‑tenant RESTful server, packaged as a Docker image and compatible with Windows, Linux, and macOS. Its compressed document store (ZStandard) and memory‑mapped indexing enable billion‑scale datasets on a single commodity server, making it suitable for high‑traffic services that need low tail latency and predictable performance.
Highlights
Sub‑millisecond latency with SIMD‑accelerated ranking
True real‑time incremental indexing without merge pauses
Built‑in faceted, geo‑proximity, and multi‑field sorting
Rust library and RESTful multi‑tenant server with Docker support
Pros
- Extremely low tail latency for interactive experiences
- Single‑node scalability to billions of documents
- Language‑agnostic tokenizers and 18‑language stemming
- Apache 2.0 license enables unrestricted commercial use
Considerations
- Memory‑mapped index may require ample RAM for hot data
- Advanced features increase configuration complexity
- Focused on lexical search; does not provide vector semantics
- Multi‑tenant server is early‑stage and may lack enterprise tooling
Managed products teams compare with
When teams consider SeekStorm, these hosted platforms usually appear on the same shortlist.
Looking for a hosted option? These are the services engineering teams benchmark against before choosing open source.
Fit guide
Great for
- High‑traffic web services needing instant search results
- E‑commerce platforms requiring faceted navigation and geo filtering
- Log or telemetry analysis where real‑time indexing is critical
- Embedded applications on commodity servers where cost and energy matter
Not ideal when
- Workflows that rely primarily on semantic vector search
- Environments demanding built‑in clustering or distributed sharding
- Projects requiring deep integration with proprietary hardware accelerators
- Teams without Rust or Docker familiarity for deployment
How teams use it
E‑commerce product catalog search
Customers receive sub‑10 ms filtered and sorted results, boosting conversion rates.
Real‑time log monitoring dashboard
New log entries become searchable instantly, enabling live alerting without latency spikes.
Location‑based service discovery
Geo‑proximity queries return nearby resources within milliseconds, supporting mobile app responsiveness.
Multi‑tenant SaaS knowledge base
Each tenant’s index is isolated and managed via API keys, simplifying onboarding and scaling on a single server.
Tech snapshot
Rust99%
HTML1%
SCSS1%
Dockerfile1%
Frequently asked questions
What programming languages can call SeekStorm?
The core library is written in Rust and provides C‑compatible bindings; higher‑level clients can be built in any language that can call a RESTful API.
How does real‑time indexing differ from near‑real‑time?
Every document is searchable immediately after ingestion, without a commit step or background segment merge.
Can I run SeekStorm on Windows?
Yes, the binary and Docker image are cross‑platform and have been tested on Windows, Linux, and macOS.
What hardware acceleration is used?
SIMD instructions for both x86‑64 and AArch64 are leveraged to speed up tokenization and ranking.
Is there a limit to index size?
The engine is designed for billion‑scale indices; practical limits depend on available storage and memory.
Project at a glance
Active- Stars
- 1,811
- Watchers
- 1,811
- Forks
- 64
LicenseApache-2.0
Repo age1 year old
Last commit2 days ago
Primary languageRust
Last synced yesterday