# SeekStorm

> SeekStorm: vector & lexical search - in-process library & multi-tenancy server, in Rust.

- **URL**: https://www.freshcrate.ai/projects/SeekStorm
- **Author**: SeekStorm
- **Category**: Databases
- **Latest version**: `v3.2.1` (2026-06-04)
- **License**: Apache-2.0
- **Source**: https://github.com/SeekStorm/SeekStorm
- **Homepage**: https://seekstorm.com
- **Language**: Rust
- **GitHub**: 1,865 stars, 66 forks
- **Registry**: github (`SeekStorm/SeekStorm`)
- **Tags**: `ai-search`, `bm25`, `dense-retrieval`, `enterprise-search`, `faceting`, `full-text-search`, `geosearch`, `hybrid-search`, `rust`, `vector-database`

## Description

SeekStorm: vector & lexical search - in-process library & multi-tenancy server, in Rust.

## Recent releases

| Version | Date | Urgency | Changes |
| --- | --- | --- | --- |
| `v3.2.1` | 2026-06-04 | High | ### Fixed  - delete_index fixed, both library method and REST API endpoint. - Intermittent indexing after commits of incomplete levels (< 65_536 documents) fixed. - In open_index now the shard_number is restored correctly. Previously it was always set to the number of physical cores, not respecting when it was previously set to a lower number via force_shard_number.  ### Improved  - Incomplete levels with less than 10_000 vectors per shard are now also clustered, resulting in improved qu |
| `v3.2.0` | 2026-05-13 | High | ### Added  - Vector search now with NEON SIMD acceleration for the AArch64 target (Apple Silicon, AWS Graviton). - The REST API info endpoint /api/v1/live now returns information weather SIMD is enabled or not. - Server info card    - SIMD entry: AVX2/Neon/None   - Web server (UI, REST API) entry: now has a clickable URL to the embedded web UI.  ### Fixed  - Vector search performance regression fixed: #[inline(always)] - Compile error for the AArch64 target (since v3.0.0) fixed. Fixes |
| `v3.1.3` | 2026-05-05 | High | ### Added  - For Euclidean, ScalarQuantizationI8, new non-affine quantization methods were added. - For Euclidean, ScalarQuantizationI8, new non-affine similarity functions were added.  ### Improved  - For Euclidean, ScalarQuantizationI8, it is now automatically selected between affine and non-affine quantization (displayed in index info card), depending on the data. Always best recall. |
| `v3.1.0` | 2026-04-29 | High | ### Added  - [TurboQuant](https://en.wikipedia.org/wiki/TurboQuant) (TQ) quantization for vector search added.     TurboQuant reduces the quantization error (rounding errors with different directions might change the ratio between dimensions) with the following benefits:   - requires no training in contrast to Product Quantization (PQ),   - provides better recall compared to Product Quantization (PQ),    - allows lower nprobe with improved query latency in ANN search,    - allows higher v |
| `v3.0.2` | 2026-04-23 | High | ### Added  - New vector search example (external vectors and vector query via REST API, Euclidean with I8 quantization) added to test_api.rest.  ### Fixed  - I8 quantization for Euclidean distance fixed for range below 1.0. |
| `v3.0.1` | 2026-04-22 | High | ### Changed  - VectorHeader.zero_point changed from i32 to i16 - VectorHeader.padding: u16 removed  ### Fixed  - index: removed duplicate use std::sync::LazyLock; |
| `v3.0.0` | 2026-04-19 | High | ### 🔥 SeekStorm v3.0.0 adds vector search and hybrid search:  #### SeekStorm uses **two** separate, **first-class**, **native** index architectures, under one roof.  - **Lexical search**: sharded and leveled inverted index.  - **Vector search**: sharded and leveled IVF index for ANN or exhaustive search. - Shared document store, shared document ID space. - Both first-class engines are integrated at the query planner level. - **Query planner** with QueryModes (Lexical, Vector, Hybrid…) an |
| `v2.3.2` | 2026-03-09 | Low | - stemmer crate renamed |
| `v2.3.1` | 2026-03-09 | Low | - Stemmers updated to Snowball 3.0.0 - Added stemmer language support for   - Armenian   - Basque   - Catalan   - Czech   - DutchPorter   - Esperanto   - Estonian   - Hindi   - Indonesian   - Irish   - Lithuanian   - Lovins   - Nepali   - Persian   - Polish   - Porter   - Serbian   - Sesotho   - Ukrainian   - Yiddish |
| `v2.3.0` | 2026-02-09 | Low | ### Added  - Added `FieldType::Binary` that for storing binary data in base64 format. This field type will not be tokenized and indexed.     For embedding binary data, e.g. images, audio, video, pdf, … in JSON or CSV documents. A self-contained alternative to storing URLs to external resources.     Using the [Data URI scheme](https://en.wikipedia.org/wiki/Data_URI_scheme) in JavaScript you can create an Image object and put the base64 as its src, including the data:image... part like this: |

## Dependency audit

- **Score**: 28/100
- **Total deps**: 48
- **Resolved**: 48
- **Unresolved**: 0
- **License conflicts**: 0
- **Warnings**: 36
- **Scanned**: 2026-05-11

## Citation

- HTML: https://www.freshcrate.ai/projects/SeekStorm
- Markdown: https://www.freshcrate.ai/projects/SeekStorm.md
- Dependencies JSON: https://www.freshcrate.ai/api/projects/SeekStorm/deps

_Generated by freshcrate.ai. Indexes github releases for AI-agent ecosystem packages._