Managed PostgreSQL
on dedicated NVMe, with pgvector.

Managed PostgreSQL means someone else provisions, secures, backs up, monitors, and upgrades your database while you keep full SQL access. You connect with a standard connection string and use Postgres exactly as you would self-hosted, without owning failover, patching, backup verification, or capacity planning. On Rivestack the database also runs on dedicated local NVMe with pgvector pre-installed, which matters for AI and vector workloads.

Each is good at something different. Rivestack focuses on dedicated NVMe performance, pgvector tuned out of the box, and flat, predictable pricing from $29/month. RDS is the default for AWS-native estates, Supabase bundles a full app platform, Neon is built for serverless and branching, and Aiven spans multiple clouds. If your workload is PostgreSQL with vector search on a budget, the NVMe + flat-price combination is hard to match.

Yes, that is the core use case here. With pgvector, PostgreSQL stores embeddings and runs HNSW/IVFFlat nearest-neighbour search next to your relational data, so RAG, semantic search, and recommendations work in plain SQL. Dedicated NVMe keeps retrieval latency low as the index outgrows memory. See our PostgreSQL for AI page for the full picture.

Dedicated clusters support PostgreSQL 16, 17, and 18 and currently default to 17. pgvector and pgvectorscale are enabled by default, alongside pg_trgm, pg_stat_statements, pgcrypto, and uuid-ossp. The dashboard lists other extensions you can add.

Yes. Solo and HA plans include daily backups via pgBackRest with point-in-time recovery (7-day retention on Solo and 14-day on HA clusters) stored off-site in S3-compatible object storage. You can also trigger manual backups and restore to any second in the window. (Free tier databases are not backed up.)

Clusters run 1 to 3 nodes managed by Patroni with asynchronous streaming replication. With 2 or more healthy nodes, Patroni can promote a replica and the load balancer keeps the same endpoint. Existing sessions disconnect, so applications must retry; failover time varies and recent unreplicated transactions can be lost. Single-node Solo does not include failover.

Yes. For Supabase, Neon, RDS, Aiven, or self-hosted Postgres, migration is pg_dump / pg_restore for smaller databases or logical replication for always-on systems, typically with no application changes. If you are upgrading a Rivestack free database to Solo, we migrate the data across automatically in about three minutes.

Pricing is flat per node: free to start, $29/month for a dedicated Solo VM, and HA clusters from $49/node. There is no per-query, per-IOP, or egress metering, so the bill on day 30 matches the quote on day 1, unlike instance-plus-storage-plus-IOPS pricing on hyperscaler DBaaS.

In the region you choose: EU Central (Falkenstein, Germany), US East (Virginia), or Singapore. EU databases and their backups stay within the EU for GDPR-aligned data residency. Database and backup traffic is encrypted in transit; contact us before relying on a specific encryption-at-rest or compliance control.

Free databases allow 15 direct PostgreSQL connections. Dedicated plans connect through PgBouncer in transaction pooling mode and accept 500 (Solo), 2,000 (Starter), 5,000 (Growth) or 10,000 (Scale) client connections, with 20 to 40 PostgreSQL transaction slots per database; extra clients queue briefly instead of failing. The docs explain the pooling model with examples for Python, Node.js, and Go.