Analytics cookies only (Google Analytics, PostHog). No ad networks, no cross-site tracking. See our privacy policy.
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.
Runs on a VPS from $24/mo (Kamatera 8GB, 8 GB RAM) · prices checked Nov 2026
Fits home hardware (2,168 MB measured idle, RAM only): Raspberry Pi 4 (4 GB) (arm64 untested), Raspberry Pi 5 (8 GB) (arm64 untested), Repurposed laptop (8 GB), Intel N100 mini PC (16 GB), AMD Ryzen mini PC (32 GB).
Pick Vespa when you need vector search, BM25 text search and structured filtering in one engine, with ranking expressions you write in the schema, and you can accept running a heavier JVM and C++ system with an application package to deploy. It is Apache-2.0 licensed. Unlike a typical vector database, it does nothing useful until you deploy an application (a `services.xml` plus schemas), after which you feed documents and query them over HTTP. Our snippet follows the project's local quick start: it runs `vespaengine/vespa:8.781.11` (released 2026-11-02, with the image tag confirmed on Docker Hub), publishes the query and document port 8080, binds the config server's port 19071 to localhost, and deploys the upstream `album-recommendation` sample app, fetched at a pinned commit of `vespa-engine/sample-apps`, by posting a tarball to the config server's `prepareandactivate` endpoint instead of installing the `vespa` CLI. Upstream's Docker guide says that, as a rule of thumb, a single-node application requires a minimum of 4 GB for the container, which is our 4096 MB floor. In this catalog, Qdrant, Weaviate and Chroma are vector databases you can use right after starting a container, and Milvus standalone needs etcd and MinIO. Vespa asks for more setup and memory, and in return the same application package format scales from this one node to a cluster once you add nodes in `services.xml` and `hosts.xml`.
Paid link — we earn a commission if you shop through it.
Run these commands on your server:
# Vespa - search and vector engine - adapted from the local quick start (Docker container, album-recommendation sample app deployed over the config server's HTTP API instead of the vespa CLI), pinned to 8.781.11
cat > docker-compose.yml <<'EOF'
services:
vespa:
image: vespaengine/vespa:8.781.11
hostname: vespa-container
restart: unless-stopped
ports:
- "8080:8080"
- "127.0.0.1:19071:19071"
volumes:
- vespavar:/opt/vespa/var
- vespalogs:/opt/vespa/logs
volumes:
vespavar:
vespalogs:
EOF
docker compose up -d
timeout 300 bash -c 'until curl -fsS -o /dev/null http://localhost:19071/state/v1/health; do sleep 3; done'
# Fetch the sample application package files (pinned commit of vespa-engine/sample-apps) and deploy them
mkdir -p app/schemas
curl -fsSL https://raw.githubusercontent.com/vespa-engine/sample-apps/2b54d6b1f6c5db9f837aabf1c501991a31bb34d5/album-recommendation/app/services.xml -o app/services.xml
curl -fsSL https://raw.githubusercontent.com/vespa-engine/sample-apps/2b54d6b1f6c5db9f837aabf1c501991a31bb34d5/album-recommendation/app/schemas/music.sd -o app/schemas/music.sd
tar czf app.tgz -C app .
curl -fsS --header "Content-Type: application/x-gzip" --data-binary @app.tgz http://localhost:19071/application/v2/tenant/default/prepareandactivate
timeout 400 bash -c 'until curl -fsS http://localhost:8080/state/v1/health | grep -q "\"up\""; do sleep 3; done'
# Feed one document and query it back
cat > doc.json <<'EOF'
{"fields": {"album": "A Head Full of Dreams", "artist": "Coldplay", "year": 2015, "category_scores": {"pop": 1, "rock": 0.2, "jazz": 0}}}
EOF
timeout 120 bash -c 'until curl -fsS -X POST -H "Content-Type: application/json" --data-binary @doc.json http://localhost:8080/document/v1/mynamespace/music/docid/a-head-full-of-dreams; do sleep 3; done'
curl -fsS --get --data-urlencode "yql=select * from music where true" http://localhost:8080/search/
# → http://SERVER_IP:8080 (query and document API of the sample app; the config server on 19071 is bound to localhost only, and neither port has authentication)Vespa's Docker documentation and the sample application flag these before you rely on it:
Vespa used 2168 MB of RAM at idle on GCP e2-standard-2 · Ubuntu 26.04 LTS · Docker 29.9.3 (first boot, measured Nov 2026). We list a minimum of 4096 MB.
The Vespa install snippet is marked verified on Ubuntu 26.04 (Nov 2026).