Home

⚡ Spector — The AI Memory Backbone

Zero-overhead, agent-ready AI search and cognitive memory — embedded in a single JVM.

Spector is a Java-native AI search engine and cognitive memory system that combines SIMD-accelerated vector search, keyword search (BM25), and biologically-inspired memory consolidation into a single embeddable library. No Docker, no external databases, no Python — just a JAR.

Connect AI agents via the built-in MCP server (Claude Desktop, Cursor, custom agents), embed directly in your Spring Boot app, or run standalone. Spector delivers sub-millisecond search at scale with zero garbage collection pressure thanks to Project Panama off-heap memory.

---

🔥 Key Numbers

Metric	Value
🧠 Cognitive Recall	0.13ms p50 at 1M memories
⚡ Vector Search	88µs p50 (10K docs, 128-dim)
🚀 Peak QPS	61,011 concurrent searches
🤖 MCP Tools	13 tools (6 search + 7 cognitive memory)
🗜️ Compression	4×–32× (SVASQ-8 to IVF-PQ)
✅ Test Suite	685+ tests, all passing
📦 Dependencies	Zero (JDK only)

---

🗺️ Choose Your Path

🚀 I want to use Spector

| Page | What you'll learn |
|:-----|:------------------|
| [[Quick Start|Getting-Started--Quickstart]] | Build, run, and search in 5 minutes |
| [[MCP Server Guide|Sdk-Usage--Mcp-Server]] | Connect Claude Desktop, Cursor, or custom agents |
| [[Installation|Getting-Started--Installation]] | Prerequisites and setup options |
| [[Configuration|Configuration--Parameters]] | All parameters with tuning advice |
| [[REST API Reference|Api-Reference--Rest-Endpoints]] | All endpoints with curl examples |
| [[Cognitive Memory|Memory]] | Getting started with AI agent memory |
| [[Cortex Dashboard|Cortex]] | Real-time neural visualization dashboard |

🧠 I want to understand how it works

| Page | What you'll learn |
|:-----|:------------------|
| [[Architecture Overview|Architecture--Overview]] | Module diagram, data flow, threading model |
| [[Core Concepts|Architecture--Core-Concepts]] | HNSW, IVF-PQ, BM25, RRF, SIMD deep-dives |
| [[Memory Architecture|Memory--Architecture]] | How cognitive memory works under the hood |
| [[6-Phase Scoring Pipeline|Memory--Scoring-Pipeline]] | Fused SIMD scoring across memory tiers |
| [[Cortex Dashboard|Cortex]] | Watch your AI's brain think — 12+ live panels |
| [[SVASQ Quantization|Deep-Dives--Svasq-Deep-Dive]] | Our proprietary SIMD-first quantization engine |
| [[Benchmarks|Deep-Dives--Real-Embedding-Benchmarks]] | Empirical sweeps on 4096-dim embeddings |

🤝 I want to contribute

| Page | What you'll learn |
|:-----|:------------------|
| [[Contributing Guide|Operations--Contributing]] | Development setup and PR process |
| [[JDK API Status|Getting-Started--Jdk-Api-Status]] | Vector API, Panama FFM compatibility |
| [[Roadmap]] | What's planned next |
| [[FAQ]] | Common questions answered |

---

💡 How It Works

Spector combines three search modalities — semantic vectors, keyword matching, and cognitive scoring — into a single fused pipeline:

graph LR
    A["🤖 AI Agent"] --> B["📡 MCP Server"]
    B --> C["⚡ SpectorEngine"]
    C --> D["🧠 Hybrid Search"]
    D --> E["🎯 RRF Fusion"]
    E --> F["🤖 LLM Re-ranking"]
    F --> G["✨ Results"]

    H["📄 Document"] --> I["🧩 Chunking"]
    I --> J["🧬 Embedding"]
    J --> C

Loading

What Makes Spector Different

• Embedded deployment — runs as a library inside your JVM. No Docker, no servers, no network hops.
• Agent-native — 13 MCP tools for search, memory, and cognitive operations. Connect Claude Desktop or Cursor in one config line.
• Cognitive memory — the only system combining power-law decay, Two-Factor strengthening (Bjork & Bjork), emotional valence, and Hebbian association in a single scoring formula.
• Zero GC pressure — all vector data and headers live off-heap via Project Panama. The JVM garbage collector never sees memory records.
• SIMD everywhere — vector distance, quantization, and scoring use Java Vector API (AVX2/AVX-512/NEON) for hardware-accelerated computation.

tip: New here? Start with Quick Start to build and run your first search in under 5 minutes. Want to connect an AI agent? See the MCP Server Guide.

---

🌟 Project Stats

Language	Java 25
License	Apache 2.0 · BSL 1.1 (memory module)
Modules	25 Maven modules
Dependencies	Zero (JDK only)
SIMD	AVX2 / AVX-512 / NEON
GPU	CUDA via Panama FFM
MCP	Built-in, 13 agent-ready tools
Distributed	gRPC fan-out + consistent hashing

---

Built with ⚡ by Spectrayan · GitHub · Apache 2.0 · BSL 1.1 (memory)