Home/architecture
Explore The Architecture

How The Verification Architecture Scales

Verity routes claims through a staged pipeline that balances speed, cost, and evidentiary quality. Simple claims resolve quickly. Ambiguous claims escalate automatically to deeper passes.

Routing strategy
5-pass escalation
Model pool
22 models active
Cost profile
Escalate only when needed
Section 1

1) Pass-Based Orchestration

Requests start with efficient model paths and escalate only when confidence or agreement is insufficient. This keeps common cases fast while preserving depth for hard claims.

  • Pass 1 handles straightforward claims with low-latency model combinations.
  • Intermediate passes add broader model families and richer source retrieval.
  • Final passes prioritize depth, contradiction analysis, and robustness checks.
Section 2

2) Evidence And Consensus Layer

The architecture separates model inference from evidence synthesis. This prevents one model from dominating a verdict without corroborating source support.

  • Source retrieval and source scoring are run alongside model reasoning.
  • Consensus is weighted by source authority and contradiction pressure.
  • Final outputs include structured rationale rather than only a verdict label.
Section 3

3) Reliability Under Production Load

The pipeline is designed for operational teams who need predictable behavior during peak volume and high-stakes publishing windows.

  • Queue-aware processing smooths spikes without discarding verification fidelity.
  • Health checks and provider fallback keep runs resilient across outages.
  • Audit artifacts are retained so post-publication reviews remain reproducible.

See The Pipeline On A Real Claim

Submit a claim and watch the architecture route it through passes, aggregate sources, and produce a transparent confidence result.