Default database / maximum optionality

Start with Postgres

When you are building a system from scratch, Postgres should probably be the default starting point. Not because it solves everything. Because it solves a surprising amount before you need another moving part.

Pick a module Read setup guides

6 day-one substitutes 4 setup guides 1 boring default

Writes Queues Search Jobs

The argument

Most systems do not fail because Postgres was not powerful enough.

They fail because the architecture became distributed before the product needed it. Use Postgres deeply, push it until you have real evidence, then add specialized infrastructure with a reason.

Moving parts avoided

Route the first version through Postgres.

Pick the system you were about to add. The console shows the Postgres-first move, when it is reasonable, and where to go deeper.

day-one route

Kafka -> pgmq

Start with a durable queue inside Postgres when you need simple jobs, events, and retries before a streaming platform.

Open comparison

Comparisons

Instead of a zoo, start with a toolkit.

Strong defaults, real caveats, and graduation signals for the moment Postgres stops being the right shape.

Instead of Kafka pgmq

Use pgmq for simple durable jobs and event handoff before you bring in a streaming platform.

queuesjobsevents Instead of MongoDB JSONB

Use JSONB when part of your data is still changing but the rest still wants SQL, joins, and constraints.

documentsschemaindexes Instead of Time-series DB TimescaleDB

Use TimescaleDB when time-based data still belongs near your product data and SQL is enough.

analyticsmetricshypertables Instead of Vector DB pgvector

Use pgvector for early semantic search when the search still needs product data, filters, and permissions.

aisearchembeddings Instead of Geospatial DB PostGIS

Use PostGIS for nearby search, boundaries, and location queries while the data still belongs with the product.

geomapsindexes Instead of Cron service pg_cron

Use pg_cron for recurring SQL jobs before you add a separate scheduler.

jobsmaintenanceschedules

Setup guides

Enough detail to try it this week.

Small, practical walkthroughs for trying the Postgres-first move without turning the site into a book.

JSONB Use JSONB without giving up relational shape

Store flexible fields in JSONB while keeping important columns, indexes, and constraints normal.

pgmq Set up a durable queue with pgmq

Create a Postgres-backed queue for small jobs, retries, and simple event handoff.

pgvector Add semantic search with pgvector

Store embeddings beside source records so search can use normal filters and permissions.

PostGIS Run nearby search with PostGIS

Use geography points and GiST indexes for practical nearby search inside Postgres.