Skip to content

Best Self-Hosted Vector database Tools

A vector database stores embeddings, the number lists that AI models produce for text and images, and finds the nearest ones to a query. It is the retrieval layer behind semantic search and retrieval-augmented generation. The six picks here sit at different points on the scale. Chroma is an easy start: one container, Python, JavaScript and Rust clients, and an index that must fit in RAM. Qdrant is a single Rust container with REST and gRPC APIs, a built-in dashboard and payload filtering. Weaviate stores objects alongside their vectors and can call embedding and generative model providers through optional modules. Milvus is built for scale, and its standalone mode runs a Milvus container plus etcd and MinIO. Those four are Apache-2.0 or BSD-3-Clause at the core, although Weaviate's wl directory holds proprietary enterprise code. pgvector is not a separate database at all: a PostgreSQL extension, under the PostgreSQL License, that adds vector columns and nearest-neighbour search beside your relational data. Vespa, Apache-2.0, takes the most setup: a search and recommendation engine that needs an application package deployed before it answers queries, with a rule-of-thumb 4 GB minimum for a single node. Size by how many vectors you will hold, since memory is the usual limit. A vector API holds your data, so keep it on a private network or behind an authenticating reverse proxy rather than on the open internet.

We earn commissions when you shop through the links below. Full disclosure →

Vector database apps

Vector database

Chroma

An open-source vector database for AI applications: store embeddings with documents and metadata in collections, then query by similarity with metadata and document filters. It runs as a single server with Python, JavaScript/TypeScript and Rust clients and an HTTP API, persisting to local disk. Apache-2.0 licensed, with a hosted Chroma Cloud service from the same team.
RustPython
easyRead guide →
Vector database

Milvus

A vector database built for similarity search over embeddings at scale: collections with multiple index types, scalar filtering, hybrid dense-plus-sparse search, partitions and role-based access, served over gRPC and a REST API with SDKs for Python, Go, Java and Node.js. Apache-2.0 licensed, written in Go and C++; in standalone mode it runs as one Milvus container that depends on etcd for metadata and MinIO (or S3) for the vectors themselves, with the message queue embedded.
GoC++etcdMinIO
moderateRead guide →
Vector database

pgvector

An open-source extension that adds vector similarity search to PostgreSQL, so embeddings live in ordinary tables next to the rest of your data. It stores single-precision, half-precision, binary and sparse vectors, supports exact and approximate (HNSW and IVFFlat) nearest-neighbor search with L2, inner product, cosine, L1, Hamming and Jaccard distances, and inherits Postgres features such as ACID transactions, JOINs and point-in-time recovery. It is not a separate server: any Postgres client library can query it.
CPostgreSQL
easyRead guide →
Vector database

Qdrant

A vector database and vector search engine for the embeddings behind semantic search, recommendations and retrieval-augmented generation: a REST API, a gRPC API, a built-in web dashboard, JSON payload filtering, quantization and per-collection snapshots. Written in Rust and Apache-2.0 licensed, it runs as a single container with no separate datastore.
Rust
easyRead guide →
Vector database

Vespa

An open-source search and big-data serving engine that combines vector search, full-text search (BM25) and structured filtering in one system, with tensor-based ranking you define in the schema. You describe an application as a package (a services.xml plus schemas), deploy it to a config server, then feed documents and query them over HTTP. A single container runs a one-node setup, and the same package format scales out to a cluster once you add nodes to services.xml and hosts.xml.
C++Java
hardRead guide →
Vector database

Weaviate

A vector database that stores objects alongside their vectors and combines vector search with structured filtering, exposed over REST and gRPC. Optional modules call embedding, generative and reranker model providers for you, or you can import pre-vectorized data. Written in Go; everything outside the repository's wl directory is BSD-3-Clause, while the wl directory holds enterprise features unlocked by a licence key.
Go
easyRead guide →

Head-to-heads in this category

Where to host itaffiliate disclosure
Kamaterafree trial
From $4/mo · 1 vCPU / 1 GB / 20 GB · US + EU + Asia
Start free on Kamatera → (opens in new tab)
DigitalOceanalso works on
From $6/mo · 1 vCPU / 1 GB / 25 GB · US + EU + Asia
Deploy on DigitalOcean → (opens in new tab)

Paid link — we earn a commission if you shop through it.

Replace a paid tool

Common questions

How much RAM do self-hosted apps in Vector database need?

Across the 6 apps in Vector database, minimums run from 512 MB (pgvector) to 8192 MB (Milvus).

Which app in Vector database uses the least RAM?

Of the 6 apps in Vector database we measured, Chroma used the least: 11 MB at idle (first boot, measured Nov 2026).

Do self-hosted apps in Vector database support Docker?

All 6 apps in Vector database list Docker support.

Analytics cookies only (Google Analytics, PostHog). No ad networks, no cross-site tracking. See our privacy policy.