Table editor
Browse, filter, insert, edit, and delete rows without leaving the database console.
Open the consoleManaged pgvector · 3 regions
Managed pgvector on dedicated NVMe Postgres.
Run vectors and relational data in one always-on PostgreSQL database. Tuned pgvector, automated backups, and fixed pricing from $29/month, with migration help included.
No cold starts · no per-query fees · no second vector database
Migrating or unsure about capacity? Get a free workload plan.FIG 0.1
One database, not two
Vectors, metadata, filters, joins, and transactions live beside your application rows. No ingestion pipeline into a second store.
FIG 0.2
Tuned pgvector
HNSW and StreamingDiskANN indexes on dedicated local NVMe, sized and tuned per plan — not a default extension install.
FIG 0.3
Measured, not promised
3.5 ms p50 at recall@10 0.90+ and ~4,465 QPS, measured with EXPLAIN ANALYZE on Growth. Always-on compute, no cold starts.
2–5 ms
p50 at recall@10 0.90+
1M
1536d vectors served hot on Scale
5.0
Product Hunt review score
3 regions
EU Central, US East, Singapore
“Their support is as lightning fast as their database.”
Serhii Boiko
@sboyko · Product Hunt
“Postgres is online in seconds with pgvector ready for AI.”
Ilyas Amraoui
@ilyas_amraoui · Product Hunt
“Databases are production and development ready to connect, saving me setup time.”
Red
@red17 · Product Hunt
“Simple Postgres with the nice schema witter which just removes every need to hustle with sql.”
Pratyay Mustafi
@pratyay_mustafi · Product Hunt
01 — Switch without guessing
Bring the workload. We will map the move.
The answer should depend on your vectors, filters, traffic, and tolerance for downtime — not a generic sales calculator.
Supabase bill climbing?
Compare like-for-like compute, storage, and the cost of keeping vector indexes hot.
See the Supabase comparison
Neon cold starts hurting RAG?
Move a steady latency-sensitive workload to always-on compute and local NVMe.
See the Neon comparison
Pinecone plus Postgres?
Put vectors, metadata, filters, joins, and transactions back in one database.
See the Pinecone comparison
Self-hosting became a second job?
Hand off Patroni, backups, PITR, monitoring, upgrades, and pgvector tuning.
What managed pgvector includes
- The right plan for your row count and dimensions
- Expected QPS, latency, and index memory
- A side-by-side cost check against your provider
- A realistic migration and cutover path
02 — Compare
Focused Postgres instead of a broader platform bill.
Calculate your workload cost| Rivestack Starter | Supabase | Neon | |
|---|---|---|---|
| Monthly cost | $49 | ~$75* | Usage-based |
| Compute | 2 vCPU · 4 GB | 2 vCPU · 4 GB | Serverless |
| Storage | Local NVMe | Cloud SSD | Cloud SSD |
| pgvector | Tuned HNSW | Extension available | Extension available |
| Backups | Daily + 14d PITR | Daily | Included |
| Measured vector perf | ~1,185 QPS · 3.2 ms p50 | Not compared | Not compared |
| Terraform | Included | Included | Not listed |
* Supabase comparison: Pro ($25) + Medium compute ($60) − $10 included compute credit. Rivestack performance is measured on Starter at recall@10 0.90 with four clients; no competitor performance benchmark is implied. Full Rivestack methodology.
Supabase and Neon are excellent broader platforms. If you need bundled auth, storage, edge functions, realtime, or scale-to-zero, keep them. If the workload is always-on PostgreSQL with pgvector, compare Rivestack vs. Supabase and Rivestack vs. Neon in detail.
03 — Semantic search
No ingestion pipeline to a second database.
pgvector + HNSW on your own dedicated NVMe Postgres. Your text, vectors, metadata, and application rows stay together.
- One-click sample dataset. Run a real semantic query before writing integration code.
- Embeddings generated for you. Text becomes a vector through the table editor or embed API.
- Numbers measured by Postgres. Query time comes from EXPLAIN ANALYZE, not a UI timer.
In-database query time, not end-to-end embedding or network latency.
04 — Included tools
Use standard Postgres. Manage it without assembling another platform.
AI-assisted SQL
Turn a plain-language request into SQL, inspect it, then run it against your database.
Read the docsTerraform provider
Create and resize clusters alongside the rest of your infrastructure, without dashboard drift.
Provision with TerraformPostgreSQL 16–18pgvectorpgvectorscalePatronipgBackRestPrometheusTerraform
05 — Pricing
Fixed monthly pricing.
No query, egress, or vector-count charges. Prices below are for EU Central.
Shared
$0/mo
For prototypes and testing pgvector with your own data.
Create a free sandbox- Shared CPU · 256 MB RAM
- 2 GB storage
- ~100K vectors (1536d)
- No credit card
Solo
$29/mo
A single-tenant PostgreSQL VM for a real app that does not need HA yet.
Start Solo — $29/mo- 1 vCPU · 2 GB RAM
- 30 GB local NVMe
- Daily backups + 7d PITR
- Founder-assisted pilot available
Production
from $49/node/mo
Dedicated capacity that can grow from one node to automatic failover.
Start Production- 2–8 vCPU per node
- 55–295 GB local NVMe
- HNSW + StreamingDiskANN
- 14d PITR · monitoring · Terraform
$29 founder-assisted pilot
A dedicated database plus a human for the first search.
Start Solo at the normal $29 price and we will help configure embeddings, tune the results, and get a working search API onto your data. No extra consulting fee.
06 — FAQ
The honest version.
Will Rivestack actually be cheaper than Supabase for pgvector?
For a like-for-like 2 vCPU / 4 GB database, Supabase Pro plus Medium compute is about $75/month after the included compute credit. Rivestack Starter is $49/month in EU Central. Supabase includes a broader platform, so compare against the features you actually use and send us the workload if you want an honest cost check.
Does Rivestack avoid Neon-style cold starts?
Yes. Dedicated Rivestack plans run PostgreSQL on always-on compute with local NVMe. There is no scale-to-zero wake-up before the first query.
Can pgvector really replace Pinecone?
For many workloads, yes: HNSW on tuned PostgreSQL keeps vectors beside relational data and supports SQL filters, joins, and transactions. The right answer depends on vector count, dimensions, filters, QPS, and latency targets, which is why the workload review is free.
How does migration work?
The dashboard migration workflow uses PostgreSQL dump and restore. Duration and write downtime depend on database size, network throughput, extensions, and write activity. Test the migration first and plan a maintenance window; Rivestack does not promise a zero-downtime cutover when the workload cannot support one.
Where is my data hosted?
Rivestack is a French company. Selecting EU Central keeps provisioned database nodes in Germany; US East and Singapore are separate choices. Data residency is not itself a compliance certification, so contact us for a DPA or a specific control review.
Is the price fixed?
Yes. Dedicated plans are a fixed monthly price per node, without per-query, egress, or vector-count charges. Resizing or adding a node changes the bill; query traffic does not.