Finds the cause. Rules out the rest.
The incident agent turns a signal into a diagnosis you can act on: what broke, what it affects, what it is not, and the one fix that makes sense.
Triaging no changes have reached the topic for 3 minutes.
Confirmed the connector is paused. The task is healthy and the source is still writing.
- Root cause
- Connector paused
- Impact
- Delayed, nothing lost
- Ruled out
Registry downCDC login locked
orders-cdc · resume
- Approved by you · 12:06:14
- Connector and task RUNNING · 12:06:19
- Evidence sealed for one year
Root cause, with receipts
Each cause names the evidence behind it: connector state, slot position, logs, events.
Rules things out
Plausible causes that the evidence contradicts are struck through, so nobody chases them.
Impact in plain words
Delayed or lost? The agent reads the slot and the source to tell you which.
A fix, not a list
One recommended action, ready for a named approver to run from the console.
You decide
The agent waits. Nothing changes until a named person approves.
A record for later
A timeline and a drafted write-up come from the same sealed record, ready for the review.
From signal to decision
- Triage
How long, how much, which pipeline.
- Diagnose
Known causes tested against the evidence.
- Rule out
Contradicted causes are named and struck through.
- Recommend
The one fix that fits, and its risk.
- Approve
A person approves, and the actions agent takes over.
Live in lab What it does today
- Diagnoses paused connectors, failed tasks, registry outages, locked CDC logins and idle sources
- Recommends a fix a named approver can run from the console
- Proved in the lab: diagnosis, approval, resume and verification in under 2 minutes
- A timeline of every incident from the sealed record, ready to export to your incident tool
- A blameless post-incident write-up drafted from the evidence, ready for you to check
- A follow-up issue in Jira with the write-up attached, in one step, live against the Fortis lab's own Jira
- After an incident, checks whether any rows were lost or arrived twice between the source and Kafka (24 of 24 lab trials), by itself once the pipeline recovers
Coming What is coming
- Incident channels in Slack or Microsoft Teams
- Follow-up issues in other trackers
Put an incident agent on your pipeline
Varqeni is in early access. Tell us about your pipelines and we will set up a trial around them.