Start pgvector from a versioned template
Use the PostgreSQL 16 pgvector template and create the application database, credentials, extension, tables, and indexes required by the search workload.
Data services · plans from $10/month
Deploy pgvector.Keep vectors beside durable relational data.
Start PostgreSQL with the pgvector extension, connect an embedding application, verify vector writes and nearest-neighbor queries, and test persistence.
Keep the repositoryInspect build and logsCustom domains and TLS
Adios deploy
Candidate release
pgvector
SOURCE
Git
REGION
de
ROUTE
HTTPS
01Source received
02Build completed
03Runtime started
04Health check passed
Promoted route
production.adios.run
A production path for
pgvectorPostgreSQL 16EmbeddingsVector searchPersistence
The production path
A working pgvector project still needs a safe release.
The application or service is only one part of production. Build evidence, runtime state, health, secrets, logs, routes, and the promoted version should remain inspectable together.
Make persistence an explicit decision
Vectors, source records, and indexes live on persistent PostgreSQL storage. Define backup, restore, re-embedding, and index-rebuild expectations before launch.
Keep service state visible to operators
Measure ingestion and query behavior with representative dimensions and row counts, then inspect application and database failures together.
From source to release
Three steps keep the deployment path reviewable.
Use the source and production behavior the project already has. The manifest records what the platform should build or provision and how the result becomes ready.
- 01
Start with source or a template
Bring the existing repository, or inspect and deploy one of the exact starter variants linked below.
$adios login - 02
Review the deployment contract
Keep commands, runtime or service version, health behavior, and secret references in adios.yaml.
$git diff -- adios.yaml - 03
Deploy and inspect the result
Follow build and runtime evidence, verify the candidate, and open the promoted route or service connection.
$adios up
adios.yaml
Your projectname: vector-db
template: pgvector:16
env:
POSTGRES_USER: app
POSTGRES_PASSWORD: secret://POSTGRES_PASSWORD
POSTGRES_DB: searchCreate the vector extension and application schema through a reviewed migration after the service is available.
Deployable starting points
Start pgvector from a template when the repository is not ready.
Deploy the exact PostgreSQL 16 pgvector variant, then connect it to a FastAPI, Node.js, or other embedding application.
Data services
pgvector 16
PostgreSQL 15, 16, and 17 templates, plus PostgreSQL 16 with pgvector.
SQLAdios managed
- Template key
- pgvector:16
- Runtime
- postgres
- Repository
- templates
- Source path
- pgvector/16
git clone https://github.com/adiosdotdev/templates.git
cd templates/pgvector/16
adios upBefore production
Verify the workload.Then promote it.
The safest first release starts with a reproducible build or service configuration and a preview that exercises the dependencies production will actually use.
Ready when…
- Embedding dimensions and distance operators are chosen.
- Credentials are stored as secrets.
- Schema and index migrations are repeatable.
- Backups include source data needed for recovery.
Preview when…
- Index type or distance metric is changing.
- The embedding model changes vector dimensions.
- Bulk ingestion competes with production queries.
Questions, answered
What to know before deploying pgvector.
Review the runtime or service boundary, template path, failure behavior, and production checks before creating the first release.
Can I deploy pgvector from an official template?
Yes. Choose the pgvector template that matches the version or configuration you need, store credentials as Adios secrets, and deploy it from the console or with adios up.
Does the pgvector template use persistent storage?
Vectors, source records, and indexes live on persistent PostgreSQL storage. Define backup, restore, re-embedding, and index-rebuild expectations before launch.
Does the template include PostgreSQL and pgvector?
Yes. The pgvector:16 template starts PostgreSQL 16 with the vector extension available. Your application still owns schema, migrations, embedding generation, and query design.
Can FastAPI connect to the pgvector template?
Yes. Inject the PostgreSQL connection string into the FastAPI runtime and verify ingestion, vector queries, dependency failure, and reconnect behavior.
Where should database and broker passwords be stored?
Store sensitive values in Adios secrets and reference them with secret://NAME. Do not place production credentials directly in adios.yaml or commit them to Git.
What should I verify before production traffic?
Verify authentication, application connectivity, write and read behavior, restart persistence, backup or recovery expectations, capacity, and the failure behavior of every dependent application.
Related deployment paths
Connect the rest of the stack.
Deploy PostgreSQL
Choose a PostgreSQL version, keep database credentials out of Git, attach persistent storage, connect the application, and verify data after a restart.
Deploy FastAPI
Deploy a FastAPI service with its ASGI import target, dependency install, runtime port, health endpoint, secrets, and promoted release tied to source.
Deploy Python
Run a Python web app or worker with the dependency file, process command, health route, and secret-backed configuration kept beside the source.
Deploy Adios workflows
Deploy scheduled jobs, webhook processors, approval gates, maintenance tasks, and operational automation from a versioned workflow manifest.
The first release
Deploy pgvector with the source and evidence attached.
Start from the repository or a template, review the deployment contract, and inspect what becomes the promoted production version.