Meridian

Layer: Data (during agent execution)
Scale: 0-100 with Platinum/Gold/Silver/Bronze tiers
Production Tier: Transaction-Grade Version: 2.0

Purpose

Meridian determines whether a data source is fit for real-time AI decision flows. It provides a consistent trust score and operational recommendation before data is consumed by agents, so weak or risky sources can be routed to review, degraded, or blocked.

Scoring Dimensions

1. Scarcity

Measures how replaceable a source is for the same decision objective. Lower replaceability generally implies higher strategic value and tighter dependency controls.

2. Quality

Measures signal integrity and operational fitness, including accuracy, freshness, completeness, structure, and verification maturity.

3. Decision Impact

Measures how strongly the source influences downstream outcomes, confidence, and business risk.

4. Defensibility

Measures durability and protection posture, including replication difficulty, licensing posture, and governance readiness.

Public note: exact formulas, weighting logic, and calibration constants are intentionally withheld.

How It Works

1Runtime tool and data request

→

2Collect source evidence

→

3Meridian evaluation

→

4Policy gating decision

→

5Audit log and telemetry

Emits

score (0-100)confidence (0-1)tiergating recommendation

Input Schema

Field	Type	Required	Description
source_id	string	yes	Data source identifier.
source_metadata	object	yes	Source type, owner, and provenance metadata.
quality_signals	object	yes	Accuracy/freshness/completeness evidence.
substitution_signals	object	no	Alternate source and equivalence evidence.
impact_signals	object	yes	Outcome and decision contribution signals.
defensibility_signals	object	no	Replication, legal, and network moat signals.
policy_context	object	no	Runtime gating policy context.

Output Schema

Field	Type	Description
framework	string	meridian
version	string	Scoring specification version.
source_id	string	Evaluated source identifier.
score	number	Meridian score from 0 to 100.
tier	string	platinum, gold, silver, or bronze.
confidence	number	Confidence in score quality (0 to 1).
gating_recommendation	string	allow, review, degrade, or block.
quality_flags	string[]	Major quality caveats or evidence gaps.

Score Interpretation

Meridian Tiering Bar

Operational priority by score range for runtime data gating.

80-100

Platinum

Interpretation: Strong source for high-stakes runtime decisions.

Typical action: Allow for sensitive production paths.

60-79

Gold

Interpretation: Reliable source with manageable caveats.

Typical action: Allow with targeted monitoring.

40-59

Silver

Interpretation: Usable with caution and tighter policy controls.

Typical action: Route through review or degraded mode.

0-39

Bronze

Interpretation: Weak source for sensitive decision flows.

Typical action: Block from sensitive execution paths.

Worked Example

Scenario: an AI risk assistant must choose one of three sanctions-screening data feeds at runtime.

Operational outcome:

1. Feed A is selected for primary runtime use.
2. Feed B is retained as conditional fallback with additional monitoring.
3. Feed C is excluded from sensitive workflows until evidence quality improves.

Illustrative note: values and scores above are example outputs for documentation only.

---

Meridian v2.0 - Runtime quality gating for AI data consumption
Copyright 2024-2026 VaryOn Works, Inc.