StealThis .dev
Recomendaciones Bases Vectoriales y RAG

Vector Databases (Embeddings & RAG)

Stores for embeddings that power semantic search and RAG — from managed services to a Postgres extension.

alternativas (6)

Chroma

Ideal para: Getting started with RAG

Open-source embedding database that's easy to run locally — a friendly default for learning RAG.

  • +Dead-simple API
  • +Runs locally
  • +Open source
  • Less proven at huge scale

Pinecone

Ideal para: Production RAG, no ops

Fully managed, serverless vector database built for production semantic search with zero ops.

  • +Fully managed
  • +Serverless scaling
  • +Fast
  • Proprietary
  • Paid

Qdrant

Ideal para: Filtered search at scale

Open-source vector DB written in Rust with strong metadata filtering and a cloud option.

  • +Fast
  • +Great filtering
  • +Self-host or cloud
  • You run it (or pay cloud)

Weaviate

Ideal para: Hybrid search

Open-source vector database with built-in hybrid (keyword + vector) search and modules.

  • +Hybrid search
  • +Modules & integrations
  • More to configure

Milvus

Ideal para: Massive scale

Open-source vector database engineered to scale to billions of vectors.

  • +Huge scale
  • +Mature ecosystem
  • Heavier to operate

pgvector

Ideal para: Vectors inside Postgres

Postgres extension that adds vector columns and similarity search to a database you already use.

  • +No new infra
  • +SQL + vectors together
  • +Open source
  • Tied to Postgres limits

Comparar

Marca las que quieras comparar

AlternativaTypeHostingOpen sourceBest for
ChromaEmbedded / serverLocal / self-hostYesPrototyping RAG
PineconeManaged serviceCloud onlyNoProduction, zero ops
QdrantServerSelf-host / cloudYesFiltered vector search
WeaviateServerSelf-host / cloudYesKeyword + vector
MilvusServerSelf-host / cloudYesBillions of vectors
pgvectorPostgres extensionYour PostgresYesAdd vectors to Postgres

To build RAG or semantic search you store embeddings (vectors) and query by similarity. Start with Chroma locally to learn, reach for Pinecone when you want a managed service with no ops, pick Qdrant / Weaviate / Milvus when you want open source and control, or use pgvector to add vectors to a Postgres database you already run. Compare by type, hosting, and openness.