VectorDB
What is VectorDB?
VectorDB provides a unified, production-oriented toolkit for Semantic Search and Retrieval-Augmented Generation across five vector databases, with feature parity between Haystack and LangChain. It ships ready-to-run pipelines for Dense, Sparse, and Hybrid Retrieval, plus advanced RAG capabilities like Reranking, Query Enhancement, Contextual Compression, Parent-Child Retrieval, and Agentic Retrieval Loops. The design is configuration-driven, environment-variable friendly, and built for consistent benchmarking across databases and datasets.Use VectorDB to build, compare, and deploy retrieval systems without re-implementing logic per backend.
Supported vector databases
Pinecone
Managed vector database with namespaces and native sparse-dense hybrid retrieval
Weaviate
Open-source vector search with BM25 hybrid retrieval, collections, and multi-tenancy
Qdrant
High-performance search with payload filtering and scalar or binary quantization
Milvus
Scalable vector database with partition-key isolation and hybrid retrieval
Chroma
Lightweight vector store for local development and rapid prototyping
Key features
Semantic search
Dense vector retrieval with metadata filters and optional answer generation
Hybrid search
Dense + sparse retrieval fused with RRF or weighted scoring
Reranking
Two-stage retrieval using cross-encoders for higher precision
Query enhancement
Multi-query, HyDE, and step-back prompting to improve recall
Contextual compression
Reduce retrieved context via reranking or LLM extraction
Parent document retrieval
Index chunks but return parent documents or context windows
Multi-tenancy
Tenant isolation using database-specific strategies at scale
Agentic RAG
Iterative retrieval loop with search, reflect, and refine steps
Built-in datasets and evaluation
VectorDB includes dataset loaders and standardized evaluation utilities so you can benchmark retrieval quality across databases and frameworks. Supported datasets:- TriviaQA - Open-domain question-answer pairs for general knowledge retrieval
- ARC - Science reasoning questions requiring multi-hop inference
- PopQA - Factoid questions about popular entities
- FactScore - Atomic facts for verification and hallucination detection
- Earnings Calls - Financial transcript Q&A for domain-specific RAG
- Recall@k
- Precision@k
- MRR (Mean Reciprocal Rank)
- NDCG@k
- Hit rate
Get started
Installation
Install VectorDB using uv package manager
Quickstart
Build your first RAG pipeline in minutes