Core Runtime

Operational Simulation Runtime

Compile real incidents into interactive P1/P0 simulations with deterministic state, evidence gating, and branching operational decisions.

What It Is

The Operational Simulation Runtime is the execution layer of MazeLabs. It takes a compressed, redacted EvidencePack produced by the Evidence Compression Engine and compiles it into a fully interactive incident simulation.

Unlike generic scenario builders or quiz-based training tools, the Runtime drives deterministic state transitions based on operator decisions, evidence inspection, and escalation behavior. Every simulation is bounded by real incident data — not scripted dialogues or hallucinated scenarios.

Runtime Architecture Flow

EvidencePack
Scenario Compiler
Simulation State Machine
Actor Runtime
Decision Engine
Session Tracker
Scoring Hooks
Debrief Generator

Scenario Compiler

The Scenario Compiler transforms a validated EvidencePack into a structured simulation scenario definition. It extracts causal chains, identifies decision points, plants hidden evidence, maps valid investigation checks, defines wrong paths, and establishes recovery criteria.

Scenario Definition Output
{
  "scenario_id": "inc-2026-0510-db-timeout",
  "phases": ["triage", "investigation", "hypothesis", "action", "validation"],
  "evidence_gates": 12,
  "hidden_evidence_count": 4,
  "wrong_paths": ["blame_network", "restart_app_server"],
  "recovery_criteria": {
    "primary": "identify_connection_pool_exhaustion",
    "validation": "confirm_pool_resize_and_monitor"
  }
}

Simulation State Machine

The simulation progresses through deterministic operational states. Each state has entry conditions, available actions, evidence requirements, and transition rules. There are no random outcomes — every result is derived from operator behavior.

1

Triage

Initial alert assessment and incident classification

2

Investigation

Evidence collection, log correlation, and pattern identification

3

Hypothesis

Formulate root-cause theories based on available evidence

4

Action

Execute mitigation steps, apply fixes, and coordinate response

5

Validation

Verify that mitigation resolved the issue without regressions

6

Escalation

Route to specialists, notify stakeholders, and escalate severity

7

Recovery

Restore full service, confirm stability, and document resolution

8

Retrospective

Review decisions, missed signals, and process improvements

Actor Runtime

Each simulation runs multiple AI-driven actors that behave according to their role, the scenario state, and available evidence. Actors generate realistic pressure, provide relevant context, and respond to operator decisions — not from scripts, but from bounded scenario logic.

SRE

Drives investigation, runs checks, correlates system signals

DBA

Investigates database health, queries, replication, and storage

DevOps

Checks deployments, config changes, CI/CD pipeline state

Incident Commander

Coordinates response, manages timeline, and tracks decisions

Support Engineer

Manages customer communication and impact reporting

Stakeholder

Applies pressure, requests ETAs, and demands status updates

Evidence Inspection

During a simulation, operators interact with evidence panels to inspect real operational data — logs, timelines, alerts, tickets, and topology. Evidence is revealed progressively as the investigation deepens, just like a real P1 incident.

Log Viewer
Incident Timeline
Alert Panel
Ticket History
Topology Map
Metric Graphs

Decision Branching

The simulation engine supports full decision branching. Different operator choices lead to different outcomes. Choosing to restart a service before checking logs produces a different simulation path than correlating metrics first. Wrong paths are tracked, scored, and reviewed in the debrief.

Decision Tree Example
Alert received → Check metrics first or Restart immediately
Check metrics → Correlate DB latency spike → Correct path ✓
Restart immediately → Issue recurs in 5 min → Wrong path tracked

Debrief Integration

Every simulation session produces a structured debrief. Missed signals, wrong assumptions, evidence usage, escalation timing, and recovery quality are all captured and replayed for organizational learning.

Session ReplayMissed SignalsDecision TimelineEvidence Usage MapScoring BreakdownRecommended Drills