Remedy The full incident-to-repair architecture
Many incident entry paths. One route to a verified fix.
Human reports, application errors, server failures, API probes and connected telemetry converge only after their evidence has been structured and governed.
1 · Incident entry
How a Remedy incident begins.
The originating signal is preserved, but no untrusted report or raw telemetry is treated as an instruction to change code.
2 · Convergence
The shared repair spine.
Once an incident exists, Remedy correlates the minimum evidence needed to test a root-cause hypothesis and prepare a governed response.
INCIDENTCORRELATEROOT CAUSEREPOSITORY MAPBASELINE SHAISOLATED REPAIRAI REPAIRGATESREVIEWPR / MRDEPLOYVERIFY
Bounded and sanitised
Configuration determines which application, runtime, probe and connected evidence is permitted. Credentials are rejected before repair analysis.
One incident timeline
User impact, service state, deployment identity and repository context are connected without building a general telemetry warehouse.
A testable hypothesis
AI reasoning proposes the likely cause. Reproduction and deterministic evidence challenge it.
Exact baseline commit
The repair maps to authorised Git state rather than whatever code happens to be in a live folder.
Isolated candidate
The engine prepares a focused change in an isolated environment governed by project rules.
Gates before confidence
Builds, tests, static checks and targeted assertions determine whether the candidate progresses.
Branch and PR/MR
Where supported, the candidate enters the normal source-control review process.
Customer-governed release
Policy defines who or what may approve and deploy the tested source state.
Check production again
The original probe or reproduction contract runs against the target environment before VERIFIED is awarded.