Introduction

Driftbase keeps the copies of your data honest. It reads committed changes from Postgres logical replication and applies them, in commit order, to search indexes and caches: Elasticsearch, OpenSearch, Meilisearch, Typesense, Redis and Valkey.

You can run it two ways, and they use the same binary:

• The drift CLI, on your own infrastructure. Apache-2.0, a single static binary, no runtime dependencies.
• Driftbase Cloud, where we run the pipeline for you in us-east-1, us-west-2, eu-central-1 or ap-northeast-1, with monitoring, alerts and a dashboard.

How it fits together

A source is a Postgres database with wal_level = logical. Driftbase creates one publication and one replication slot per source.

A sink is a destination: an index, a collection or a key space.

A pipeline connects one source to one sink. It says which tables to read, how to shape each row into a document (a SQL query, a column list, or the table as is) and what to do on delete.

postgres (orders, order_items) ──▶ drift ──▶ meilisearch: orders
                                        └──▶ redis: order:{id}

Where to go next

• New here? Getting started takes about ten minutes.
• Install the CLI on macOS, Linux, Windows or in Docker.
• Every option in driftbase.toml is in the configuration reference.
• Guides for Meilisearch, Redis caches and backfills.