Skip to content

What is SPT?

SPT (SmartPoints Technology) is a unified platform for vector intelligence, agent orchestration, and distributed systems. It gives developers a single toolkit to build AI-powered applications that store and query vector embeddings, coordinate multi-agent swarms, and communicate over high-performance messaging — all backed by a Rust engine with Node.js bindings.

Key Capabilities

  • Vector Database (spt-core) — HNSW indexing, product quantization, SIMD-accelerated nearest-neighbour search in sub-millisecond time
  • Swarm Orchestration (sptflo) — 60+ specialized AI agents with self-learning hooks, vector memory, fault-tolerant consensus, and MCP integration
  • NATS Messaging (spt-nats) — Sub-millisecond pub/sub, JetStream persistent queues, and KV state for inter-agent communication
  • Hyperbolic Embeddings — Poincare ball model for hierarchical data that Euclidean spaces cannot represent efficiently
  • Post-Quantum Cryptography (spt-crypto) — ML-DSA and ML-KEM algorithms for forward-secure agent communication

Architecture

mermaid
graph TB
    subgraph "Rust Crates"
        core[spt-core<br/>Vector DB + HNSW]
        sona[spt-sona<br/>Neural Architecture]
        nats[spt-nats<br/>NATS Messaging]
        crypto[spt-crypto<br/>Post-Quantum Crypto]
        graph[spt-graph<br/>Hypergraph Engine]
    end

    subgraph "NAPI-RS Bindings"
        napi[Node.js Native Modules]
    end

    subgraph "Orchestration Crates"
        flo[spt-flo<br/>Wire Publisher + ReasoningBank]
        flomcp[spt-flo-mcp<br/>MCP Tools]
    end

    subgraph "CLI"
        cli[spt CLI]
    end

    core --> napi
    sona --> napi
    nats --> napi
    crypto --> napi
    graph --> napi
    napi --> flo
    flo --> flomcp
    flomcp --> cli

Who is SPT For?

  • AI/ML engineers building retrieval-augmented generation (RAG) and multi-agent systems
  • Platform teams orchestrating distributed agent workflows with fault tolerance
  • Developers who want a batteries-included toolkit for vector search, embeddings, and neural architectures

Next Steps

Released under the MIT License.