Liberatii is a Gold Sponsor of Postgres Summit US 2026, taking place 30 September to 2 October at Convene, 555 Broadway, New York — and we're bringing Liberatii Advisor to the stage for the first time.

An optimizer, not an agent

Nima Torabi and Vitalii Akimov will present An Optimizer, Not an Agent: Deterministic, Cost-Based Postgres Tuning, including the first live demonstration of Liberatii Advisor.

The distinction in the title is the whole argument. A lot of tooling in this space is an agent: it invokes a collection of tuning scripts and returns a recommendation, and you take it on faith. Advisor works the way a database query optimizer works. It constructs a search space, derives properties, prices the alternatives, and selects a winner.

The consequence is determinism. Given the same workload facts, transformation rules, constraints and prices, Advisor generates the same alternatives, assigns the same costs, and reaches the same decision — including the decision that the original design was already the right one. Selected and rejected alternatives both keep their cost evidence and provenance, so the reasoning is auditable rather than asserted.

Two demonstrations

The first isolates a single application workload, finds a five-join bottleneck, and walks through the alternatives and their costs. Only safe online changes are approved, indexes are applied while traffic continues, and the result is verified at the application boundary. The operator supplies policy — scope, safety, priority — not optimization logic.

The second makes two cost components visible that are usually collapsed together. The same analytical question is run over raw tables on a remote three-node YugabyteDB cluster, against a maintained result on YugabyteDB, and against a fresh materialization beside the application. Separating distributed compute cost from network delivery cost shows why a remote materialized result can still leave a WAN round trip on every request.

Both demonstrations use ordinary SQL and standard Postgres clients. You see the executed SQL and the measured outcome, not a polished dashboard.

Familiar ground

New York has been good to us. We announced our partnership with Datavail from this same conference in 2024, when it was still PGConf NYC. Two years on, the question we were asking then — whether application refactoring is what really holds back Oracle to PostgreSQL migrations — has not gone away, but the answers have got a lot more concrete.

Come and find us in New York, or get in touch beforehand.