consensus-engine

Multi-agent consensus validation spawns two or more parallel validators against identical journey lists to collect independent evidence. It synthesizes per-journey verdicts into a unified consensus report with a confidence tier, increasing release confidence for high-stakes features.

Multi-agent consensus validation is an orchestration process that runs multiple independent validators on the same journeys to increase confidence. You need it for regression reviews, pre-ship gates, and flaky-behavior detection when single-validator results are insufficient.

Launch at least two parallel validators against the same journey list. Each validator records independent verdicts in an isolated evidence subdirectory, and the consensus engine synthesizes a unified report only after all validators complete their runs.

Yes, the consensus engine triggers root-cause analysis when validator disagreements arise. This happens after all validators complete and before finalizing the report, ensuring conflicting evidence is investigated before a final consensus verdict is synthesized.

The best approach is spawning at least two parallel validators that receive identical journey lists and record independent verdicts in isolated evidence subdirectories. The engine enforces the official contract and synthesizes a single consensus report with a confidence tier after completion.

Consensus validation isolates evidence subdirectories to ensure each validator records independent verdicts without cross-contamination. This isolation is enforced by the engine's contract, allowing accurate synthesis of per-journey verdicts into a unified consensus report.