Dedicated PostgreSQL 18 with pgvector on NVMe storage. Pre-tuned HNSW indexes, automated backups, 14-day PITR, HA failover. 2,000 QPS vector search at sub-4ms latency on the $29/mo plan. Free tier to start. No per-query fees, no surprise bills.

Maker comment (post this as the first comment after launch):

Hey Product Hunt! I built Rivestack because every managed PostgreSQL option with pgvector is either expensive (AWS, GCP) or gives you shared resources (Supabase and Neon free tiers).
The problem: if you want dedicated pgvector for a RAG app or semantic search, you're looking at $100-150/month minimum on most platforms. That's steep when you're prototyping or running a small production workload.
Rivestack gives you a dedicated PostgreSQL 18 instance with pgvector pre-configured on NVMe storage for $35/month. Same 2 vCPU / 4 GB RAM that costs ~$105 on Supabase.
How? We run on dedicated NVMe servers in certified data centers instead of hyperscalers, and pass the savings to you. The stack is Patroni for HA, pgBackRest for backups, Prometheus for monitoring — battle-tested open-source tools.
To prove it works, I built ask.rivestack.io — semantic search over 30 days of Hacker News, running on a Rivestack free tier cluster. Every query hits pgvector with HNSW indexes and returns in under 50ms.
What's live today:

Free tier (shared cluster, 2 GB storage)
Starter at $35/mo (2 vCPU, 4 GB RAM, 55 GB NVMe)
Growth at $59/mo (4 vCPU, 8 GB RAM, 135 GB NVMe)
EU and US-East regions
Terraform provider

I'd love feedback — especially on what's missing before you'd trust a newer provider with production data. That's the hardest problem I'm solving and I want to hear what would make you comfortable.
Try the demo: https://ask.rivestack.io
Docs: https://docs.rivestack.io

Rivestack - Managed pgvector on NVMe. 3× cheaper than Supabase.

Replies