PelQi in Varqeni

The operator’s voice for production data.

Varqeni watches data platforms around the clock. PelQi is how people talk to it: ask what is wrong, see the evidence, approve a fix, and get the record. Live in the Fortis lab.

Evidence first

Every answer shows its sources.

PelQi reads the live signals that bear on the question, checks every claim in its answer against them, and cites each one. Tap a source to see what it read and when.

  • Connector, slot, heartbeat and cluster state
  • Each claim held against the evidence
  • Nothing lost, only delayed, said only when it is true
PelQi, in Varqeni
Why is the orders pipeline behind?

The orders connector was paused at 12:04. The slot is holding every change, so nothing is lost, only delayed.

connector: PAUSEDslot: growingsource: writing
connector: PAUSEDConnect REST, read 12:04:57state PAUSED, task 0 healthy
Never guess

When it does not know, it says so.

If none of its sources hold the answer, PelQi names what it searched and stops there. A confident guess near production is worse than an honest gap.

  • It lists the sources it searched
  • It says what would let it answer
  • It never fills a gap with a likely story
PelQi, in Varqeni
Who changed the connector config last week?

I cannot tell. None of my sources record who changed it, so I will not guess.

Searched 3 sources. The answer is not in any of them.
Refuses at every tier

Some things it will never do.

Operations that cannot be undone are refused however they are asked, and PelQi offers the safe fix instead. Nothing is sent when it refuses.

  • Dropping replication slots, resetting offsets, deleting topics
  • Changing publications, restarting the database
  • A safer fix offered every time
PelQi, in Varqeni
Drop the replication slot so it stops growing.

I will not do that, at any tier. Dropping the slot throws away every change it holds. Resuming the connector drains it safely.

Never automatedDrop a replication slotReset offsetsDelete or recreate topicsChange a publicationRestart the database
An engineer at a desk late at night with a laptop and phone
Within policy

Approved once. Checked twice.

A named person approves a safe, reversible fix; it runs once, the live state is verified before and after, and the record is sealed for thirty days. In the Fortis lab, Varqeni now resumes a paused connector on its own within policy, after ten of ten closed-loop trials. Every other action still waits for a person.

Measured, not claimed

Proven in the Fortis lab.

Real faults injected into a live data pipeline, diagnosed without the answers. Lab results, not customer results.

30/30blind fault trials diagnosed correctly
5 sfrom a person's approval to a verified fix
0changes sent by any refused action
30 daysevery investigation sealed against edits
Early access

See PelQi in Varqeni

Varqeni runs in the Fortis lab today. Join the waitlist to bring it to your data platform.

PostgreSQL, Apache Kafka, Kubernetes and Prometheus are trademarks of their respective owners. Photos are licensed stock and show models, not PelQi customers.