Parent POM for Spectrayan SSE project

Opinionated, auto-configured Server-Sent Events (SSE) library for Spring Boot / WebFlux. Provides a reactive SseEmitter service for topic-based streaming, heartbeat keep-alive, configurable headers, MDC propagation, and a ready-to-use controller that exposes /{topic} SSE endpoints out of the box. Requires Java 21+ and Spring Boot 3.x.

Ultra-fast, SIMD-accelerated semantic search engine built on Java Vector API + modern JVM technologies.

Shared utilities: content extraction, text chunking, normalization, and configuration.

SIMD-accelerated math kernels and similarity functions via Java Vector API.

Centralized configuration: properties, typed config records, and parameter types (IndexType, HnswParams, PersistenceMode).

SPI interface for embedding providers. Zero dependencies — implement this to plug in any embedding model.

Panama MemorySegment-based zero-copy vector and document storage.

Ollama embedding provider using java.net.http — zero external dependencies.

HNSW vector index and BM25 keyword index implementations.

Query engine with hybrid search orchestration and RRF fusion ranking.

LLM-powered test validation utilities for Spector modules — LLM-as-Judge assertions.

GPU acceleration via Panama FFM + CUDA for batch vector similarity computation.

Instance-scoped telemetry bus and event types for Spector observability. Provides TelemetryBus (thread-safe, HA-safe), TelemetryScope (ScopedValue carrier), and a sealed TelemetryEvent hierarchy for all telemetry data.

Retrieval-Augmented Generation pipeline: context assembly, attribution, and retrieval orchestration.

Pure ingestion utilities: file discovery, text chunking, title extraction. No dependency on runtime or engine — used as a utility by spector-runtime.

Biologically-inspired cognitive memory for autonomous AI agents. 16 neuroscience mechanisms — from dopamine-driven surprise detection to hippocampal sleep consolidation — running natively on Java Panama with zero-GC, SIMD-accelerated off-heap storage.

Search engine facade, lifecycle management, and ingestion pipeline.

Micrometer-based observability for Spector — metered decorators for engine, memory, and ingestion.

Unified application context that composes SpectorEngine (search) and SpectorMemory (cognitive) into a single configurable runtime. All consumers (MCP, API, CLI) depend on this module.

High-performance Model Context Protocol (MCP) server for Spector. Provides AI agents (Claude Desktop, Cursor, etc.) direct in-process access to Spector's SIMD-accelerated vector search engine via JSON-RPC over stdio or HTTP. Uses the official Anthropic MCP Java SDK for protocol compliance.

Unified Spector node — serves HTTP REST, gRPC, MCP-over-SSE, and Prometheus metrics on a single Armeria (Netty) port. Includes distributed cluster coordination (membership, shard routing, replication) for HA deployments.

JMH benchmarks for Spector performance testing.

Java client SDK for programmatic interaction with Spector REST API.

Spring AI VectorStore implementation backed by Spector engine.

Command-line tool for managing Spector instances.

Single fat JAR containing all Spector modules. Default main class: SpectorMcpMain (MCP server). Use -cp with the appropriate main class to run different tools: - MCP Server: com.spectrayan.spector.mcp.SpectorMcpMain - CLI: com.spectrayan.spector.cli.SpectorCtl

Redis Pub/Sub bridge for the Spectrayan SSE server library. Enables cross-instance SSE event fan-out via Redis. Drop this module in for automatic multi-pod SSE support — no Spring Cloud Stream required.

Spring Cloud Stream bridge for the Spectrayan SSE server library. Enables cross-instance SSE event fan-out through any Spring Cloud Stream binder (Kafka, RabbitMQ, Google Pub/Sub, Pulsar, Azure Event Hubs, etc.). Drop this module in alongside your chosen binder dependency for automatic multi-pod SSE support.