Community
£100 / month
Get our community tools working for your use case.
- Access to all our community repositories
- Hands-on setup and configuration for your needs
- Basic support and base updates
AI-native data operations at structural depth.
Genefold transforms vector spaces into compact spectral artifacts that power search, drift detection, OOD monitoring, and data valuation across ML and LLM workflows.
In design Platform
The Platform is taking shape.
Your embeddings hold more than search results. We are designing a platform to help you uncover what is missing, understand where retrieval falls short, and turn your data into a search asset you can inspect, compare, and deploy on your own terms. We are going to introduce DAVAs to data engineering workflows: Data Assets Vectorisation Apps.
Live Demo
Try spectral search on real vulnerability data.
~360k
CVE records
<100ms
Search latency
NVD
Data source
The CVE Search Engine indexes the National Vulnerability Database with ArrowSpace spectral retrieval. Search across ~360k CVE records and see how manifold-aware ranking surfaces relevant vulnerabilities that cosine-only baselines miss.
- ~360k CVE records from NVD
- Live tau modulation: spectral, hybrid, or cosine-like
- Compare spectral and baseline results side by side
CVE-2014-0160
Heartbleed — OpenSSL information disclosure
Spectral match
CVE-2021-3449
OpenSSL CA certificate denial of service
Baseline
CVE-2022-4304
OpenSSL BN_mod_sqrt timing side-channel
Spectral match
Genefold-vd
Bespoke Vector Database. Native ArrowSpace search.
Genefold-vd is the only vector database with native ArrowSpace support — bringing spectral retrieval into semantic search. It builds a signal graph from your embeddings, persists the structural artifacts, and answers λτ-indexed queries, so ranking uses the geometry of your data rather than cosine proximity alone. See the retrieval approach live in the CVE Search demo.
- Native spectral retrieval. ArrowSpace λτ-indexed queries with configurable tau — tune ranking from purely spectral to cosine-like.
- Persistent vector and graph storage. Embeddings, adjacency, Laplacian, and λ values in a Lance-backed collection.
- Highly distributed, shutdown-resilient storage. Materialised views scale with the Lance format; the collection persists across process shutdowns and restarts, and crashed builds are recoverable.
- Incremental ingestion. Bootstrap a collection or append new vectors through a single ingest path.
- Dataset maintenance. Append, remove, and compact data; verify stored artifacts and sweep orphaned ones.
- Inspectable search foundations. Metadata and graph artifacts are queryable for audit and diagnostics.
- Application integration. Scriptable CLI with JSON output and a TCP query service. Linux and macOS.
Open source
Built in the open. Production-ready today.
pyarrowspace
arrowspace for Python — graph analysis and vector search. pip install arrowspace for spectral indexing, bounded scores, and manifold-coherent retrieval.
Vector analytics and search using dispersion models. cargo add arrowspace for graph analysis, search, and energy-distribution statistics.
arrowspace_tunerHyperparameter optimisation and benchmark tooling for ArrowSpace spectral pipelines. Tune tau-modulation, graph-wiring parameters, and retrieval profiles.
arro-serverFastAPI server exposing large Zarr v3 datasets and their ArrowSpace-derived metadata over HTTP — dataset browsing, numpy-style slicing, stats, and a search seam for spectral retrieval.
arro-nlp-frontendAPI server for OpenAI-like NLP interactions — the conversational frontend layer of our CVE Search pipeline.
pygenestore
Store your numpy arrays at scale with the Lance format —
millions of rows, with synchronous and async APIs.
A set of skills to use arrowspace and pyarrowspace — prompt-driven workflows for spectral operations.
Model Context Protocol server for ArrowSpace — enabling LLMs and agents to perform spectral search and embedding operations natively.
8 featured public repositories
Building with embeddings?
Try ArrowSpace, open an issue, star the repo, or tell us how spectral search fits your stack. We read every message.
genefold-ai
Our company AI — trained on spectral intelligence principles and embedding workflows to assist with research, tooling, and data operations.
View all repositories at github.com/Genefold
Plans & support
From first deployment to operation at scale.
Choose the engineering support, commercial access, and deployment coverage your team needs — from getting started with our community tools to running custom solutions at scale.
Professional
£1,000 / month
Adapt the technology to your business, with support through deployment.
Everything in Community, plus:
- Customisation of up to two features per month
- Extended support for deployment
- Access to stable versions of our commercially licensed repositories
- Non-exclusive commercial licence for your business
Full-cover
Flagship£5,000 / month
Work with our latest technology.
Everything in Professional, plus:
- Access to all our commercially licensed repositories at their latest stage of development
- Possibilty of IP assignment (Exclusive License) for custom-built products derived from our IP
- Support and deployment of custom-based solutions at scale in remote enclaves
- On-call service
- Service infrastructure for the Genefold-vd vector database to run at scale on its latest version
Every tiers will provide early-stage access to the future platform. Public repositories remain openly accessible — subscriptions add support, customisation, and commercial licensing: Talk to us.
Early access
Stay ahead of the curve.
Leave your email to receive updates, early access invitations, and news directly from the Genefold team.
Engineering
How we build spectral intelligence.
Read our Engineering Blog:
002 — Foundations
Read all engineering notes
Genefold First Principles: Depletion of Uncertainty
Entropy is the irreducible uncertainty of the source. Cross-entropy is the mathematical structure that measures how close the model's distribution is to the truth. Training depletes the gap — the KL divergence.
001 — AlgorithmsDesigning ArrowSpace and graph wiring
How graph wiring in the feature-space Laplacian powers spectral indexing for vector retrieval, and what the experimental evidence shows on CVE and TREC-COVID.
Science & explainers
Core paper
2025
ArrowSpace: Spectral Search for Embeddings and Graph Analysis
Introduces spectral indexing with graph-Laplacian structure and bounded spectral scores for vector search.
Open paper page
Core paper
2026
Epiplexity And Graph Wiring: An Empirical Study for the Design of a Generic Algorithm
Every dataset generates information, every manifold draws a unique surface.
Open paper page
Software
Ready
ArrowSpace: a generic algorithm for data operations
Hundreds of downloads per week: pip install arrowspace
Core paper
2026
From Embedding Geometry to Spectral Search: Energy Dispersion Networks For Vector Retrieval
Lorenzo Moriondo, Ilias Azizi — coupling geometric similarity with spectral information for improved retrieval and adaptive tau-modulation in RAG pipelines.
Open paper
Preprint
2026
Building Latent Spaces out of Sandwiches: Complementary Spectral Factorizations for Graph-Metric Embeddings
Lorenzo Moriondo — embedding feature vectors through complementary slices of a graph Laplacian, with cut-invariant Gram geometry and interpretable spectral energy budgets.
Open paper
Workshop paper
2026
Genefold Data Governance: A Semantic Governance Platform for AI Safety, Drift Monitoring, and Federated Data Asset Management
Lorenzo Moriondo, Tommaso Moriondo — Solid-based federated access control, ArrowSpace-powered semantic retrieval, and drift monitoring operating directly on governed data assets.
Open paper