The coupling matrix learns the shape of your data across every domain, then predicts cross-domain behavior no single-purpose system can see. Realise the shape once; serve it forever. One binary, any hardware.
The substrate is structurally decoupled from the advanced-node chip supply chain. The same binary scales from a single node to thousand-node clusters — and adapts as new hardware classes appear.
The substrate runs at any scale — single-node to thousand-node clusters — without re-architecture. Same binary, same APIs, same operational model.
The same substrate adapts to new hardware classes — accelerators, neuromorphic, analog, future quantum backends — via remote refinement as they appear. No re-installation. No new binary.
Structurally decoupled from the advanced-node-chip supply chain. Decade-old commodity hardware is a current production target — not an aspirational floor. Intelligence is not a function of advanced-node-chip availability.
One substrate frame renders natively to every supported output — web, native CLI, VR/AR headsets, holographic displays, and brain-computer interface channels — without per-output retraining, per-output models, or per-output pipelines.
One estimator learns the shape of every entity type, then serves prediction, search, and compression off the same realised shape. The database is the substrate it runs on.
Financial volatility → network congestion, weather → sensor drift. The coupling matrix finds effects no single-domain system can see, and forecasts them live via /v1/query/forecast.
One index answers both "what happened" and "what's similar." No ETL pipeline between your time-series DB and your vector DB.
The same shape estimator that powers prediction compresses your data without giving up a single bit of fidelity.
Time-series databases can't do similarity. Vector databases can't handle time. Everyone stitches 3–5 systems together with brittle ETL. VectorScaleDB replaces all of them.
vsdbNo dependencies. No runtime. No containers required. Place the binary anywhere and run it — the engine auto-detects your hardware and configures itself.
# macOS · Linux · Windows · ARM64 / x86_64 # 1. Download the binary for your platform # 2. Make it executable and run it ./vsdb start
Package-manager distribution — brew, winget, Docker images, and language SDKs
(pip / npm / cargo) — arrives at launch.
See platform support & setup →
The features that only make sense when time-series and vectors live in the same index.
Every entity has a full behavioral history. Query what it did, when it changed, and what it's likely to do next.
Up to 16,500× lossless. Regime detection learns data patterns instead of applying a one-size-fits-all codec.
Detect cascade effects across entity types. Financial volatility predicting network congestion, not keyword co-occurrence.
Nodes discover each other and exchange behavioral templates. Your data stays on your hardware.
IoT, finance, genomics, NPC AI, neuroscience, autonomous vehicles, and more — purpose-built, not generic blobs.
Consent gates, irrevocable consent, no self-propagation, threshold cryptography. Safety as physics, not policy.
Built with Rust. Zero unsafe. 5,000+ tests. Zero warnings.