Technology

Databases & Storage

The right store for each workload: relational, cache, analytics, vector, object - your data stays portable.

Overview

Where it fits.

We match the store to the query pattern, not the trend. PostgreSQL by default, specialized engines where they earn it, and schemas you can walk away with. Managed or self-hosted is a decision we make together.

Tools

What we reach for.

PostgreSQL
16
primary relational
PostgreSQL Licensedocs →
Valkey
8
cache/KV (open Redis fork)
BSD-3-Clausedocs →
ClickHouse
2x
analytics/OLAP
Apache-2.0docs →
DuckDB
1.x
in-process analytics
MITdocs →
SQLite
3
edge/embedded
Public Domaindocs →
pgvector
0.8
vector search
PostgreSQL Licensedocs →
TimescaleDB
2.x
time-series
Apache-2.0 (community)docs →
MinIO
S3-compatible object storage
AGPL-3.0docs →

Use cases

What it's great for.

Transactional OLTP
PostgreSQL with tuning, replication, backups
Caching & sessions
Valkey with eviction + TTL policies
Analytics & OLAP
ClickHouse / DuckDB columnar queries
Vector + time-series
pgvector embeddings, TimescaleDB metrics

Pairings

Plays well with.

backend developmentdata access layers + migrations
cloud devopsmanaged vs self-hosted + backup/DR
ai dataembeddings + pipeline storage

As a service

Where this stack ships.

FAQ

Questions, answered.

How do you pick a database?

Workload-driven: the query pattern, consistency needs and scale decide. PostgreSQL is the default; specialized engines earn their place.

Managed or self-hosted?

Both are options. Ops maturity, cost, and compliance decide - we lay out the trade-offs before choosing.

Can you migrate databases?

Yes - phased with dual-write, validation gates, and rollback plans. Zero-downtime where the workload allows.

Vector search?

pgvector for moderate scale inside Postgres; we move to a dedicated store only when recall/latency demand it.

How do you avoid lock-in?

Portable schemas, open formats, documented exit. Your data stays yours - always.