NeurondB: Advanced AI Database Extension for PostgreSQL
Production-grade vector search, machine learning inference, hybrid retrieval, and complete RAG pipeline support—all within PostgreSQL
neurondb-demo
NeurondB Interactive Demo Terminal
PostgreSQL extension for AI/ML with vector search, hybrid retrieval, and ONNX inference
Select a tab above and click "Run Demo" to begin
Speed:
Ready
PostgreSQL 16-18Vector SearchML InferenceHNSW IndexingHybrid SearchRAG Pipeline
Key Features

Why neurondb

Vector Search & Indexing

Multiple vector types (float32, float16, int8, binary), 10+ distance metrics, HNSW and IVF indexing with automatic tuning, 2x-32x quantization.

ML & Embeddings

Text, image, and multimodal embedding generation with caching. ONNX runtime for model inference, batch processing, and fine-tuning support.

Hybrid Search

Combine vector and full-text search with weighted scoring. Multi-vector support, faceted search, and temporal decay for relevance.

Reranking

Cross-encoder reranking, LLM-powered scoring (GPT/Claude), ColBERT late interaction models, and ensemble strategies.

RAG Pipeline

Complete Retrieval Augmented Generation pipeline in-database. Document processing, retrieval, generation, and best practices.

Background Workers

neuranq (async job queue), neuranmon (auto-tuner), neurandefrag (index maintenance). Production-ready with tenant isolation.

Analytics

K-means and DBSCAN clustering, PCA/UMAP dimensionality reduction, outlier detection, and embedding quality metrics.

Performance & Security

SIMD-optimized operations, intelligent query planning, encryption, differential privacy, row-level security, and comprehensive monitoring.

PostgreSQL Native

Built with PostgreSQL C coding standards. Pure SQL interface, 100+ functions, PostgreSQL 16-18 compatible.

Depth

Feature Matrix

CapabilityDescriptionPerformanceProduction Ready
Vector SearchHNSW indexing, multiple distance metrics, quantizationSub-millisecond on millions✓
ML InferenceONNX runtime, batch processing, embedding generationHigh-throughput batch ops✓
Hybrid SearchVector + FTS, multi-vector, faceted, temporalOptimized query planning✓
RerankingCross-encoder, LLM, ColBERT, ensembleGPU-accelerated support✓
Background WorkersQueue executor, auto-tuner, index maintenanceNon-blocking async ops✓
RAG PipelineComplete in-database RAG with document processingEnd-to-end optimization✓
Comparison

Feature Comparison

FeatureNeurondBpgvectorpgvectorscalepgai
Vector IndexingHNSW + IVFHNSW + IVFStreamingDiskANNNo (uses pgvector)
ML InferenceONNX RuntimeNoneNoneOpenAI/Ollama API
Embedding GenerationIn-databaseExternalExternalExternal API
Hybrid SearchNative (Vector+FTS)ManualManualManual
RerankingCross-encoder, LLM, ColBERTNoneNoneNone
Background WorkersQueue, Tuner, DefragNoneNoneNone
RAG PipelineComplete In-DBNoneNonePartial (API calls)
Quantization2x-32x (FP16, INT8, Binary)Binary onlyBinary onlyNone
AnalyticsClustering, PCA, UMAP, OutliersNoneNoneNone
Multi-TenancyTenant isolation + quotasManualManualManual
Auto-TuningBackground workerNoneNoneNone
PostgreSQL Versions16, 17, 1812-1815-1816-18
LicenseApache 2.0PostgreSQLTimescale LicensePostgreSQL