~/rivestack $ cat company/about_

Postgres infrastructure
built by the team that runs it.

Rivestack is managed PostgreSQL for AI teams that want fast pgvector, predictable bills, and a database they can still understand. It is built and operated by a small infrastructure team in France, close to both the systems and the customers who depend on them.

talk to the team see what we publish

team.yaml● direct

team: Infrastructure + product
company: OPSAFLOW SAS
home: France · European Union
focus: PostgreSQL + AI workloads
default: EU Central · Germany
support: human, not tiered

France

EU company and operating base

3 regions

EU Central, US East, Singapore

$29/mo

Dedicated NVMe Postgres starts here

# why Rivestack exists

AI teams needed better Postgres, not another platform.

The usual RAG stack puts application data in Postgres and embeddings somewhere else. That creates two bills, two operational surfaces, and a synchronization problem. Or it puts pgvector on general-purpose cloud Postgres and leaves the team to discover HNSW memory limits, storage latency, and connection-pool behavior in production.

Rivestack takes the narrower path: always-on PostgreSQL on local NVMe, pgvector and pgvectorscale ready, sensible production defaults, backups, monitoring, and migration help. One database for rows and vectors, without turning a product team into a database operations team.

# operating principles

The rules behind the product.

Infrastructure trust comes from what a provider does when the honest answer is less convenient than the sales answer.

Postgres should stay portable

Rivestack is standard PostgreSQL. Your schema, extensions, connection string, pg_dump workflow, and SQL stay familiar, so leaving never requires an export from a proprietary database API.

Publish the conditions with the number

Vector benchmarks include dimensions, recall, index settings, node size, concurrency, and the harness used to run them. A fast number without its workload shape is not useful evidence.

A plan should have one price

Compute and included NVMe storage are priced per node. No per-query meter, no vector-operation line item, and no invoice that grows because your product had a good week.

Write the limits down

Rivestack is not a BaaS and not a hyperscale vector database. We document capacity boundaries and will say when Supabase, Pinecone, or a larger platform is the better fit.

# hands-on team

The Rivestack team

Infrastructure, product & support

The team builds Rivestack end to end: database automation, backups, monitoring, failover, the control plane, dashboard, documentation, benchmarks, and customer support.

That means support does not move through layers before it reaches someone who can change the system. The person reading your workload or incident is the person who operates the platform.

hello@rivestack.io

# what runs underneath

Boring infrastructure, deliberately.

The product is built from proven PostgreSQL tooling and automated operational workflows. Rivestack adds the opinionated packaging, workload tuning, and direct accountability around it.

PostgreSQL 16–18
pgvector + pgvectorscale
local NVMe storage
Patroni high availability
pgBackRest backups + PITR
Prometheus monitoring
Terraform provisioning

measured benchmarks security practices sub-processors

# company details

A real EU company.

Legal entity: OPSAFLOW SAS
Registration: SIREN 990 431 215
Registered office: 60 rue François 1er, 75008 Paris, France
General contact: hello@rivestack.io
Legal & DPA: legal@rivestack.io

Postgres infrastructurebuilt by the team that runs it.