Attested Computations in Open Knowledge Format (OKF): Ensuring Verifiable AI Knowledge

When AI agents answer complex analytical questions—such as calculating fiscal revenue or gross profit—knowing what a metric means is only half the battle. We also need to guarantee that the number was produced using the exact, sanctioned computation rather than agent improvisation.

In Open Knowledge Format (OKF) v0.2, this challenge is solved through Attested Computations.

---

What is an Attested Computation?

An Attested Computation is a standalone concept (type: Attested Computation) that carries a sanctioned way to compute a value. Instead of embedding raw SQL or scripts directly inside narrative documents, OKF separates the definition from the execution contract.

Why Separate Computations into Concepts?

1. Runtime Binding: The runtime property (e.g., bigquery, postgres, dbt, python) defines precisely how parameters are bound and interpreted.
2. Reusability: A single computation can back multiple metrics, dashboards, and reports across the organization.
3. Independent Trust: Each computation maintains its own verified, stale_after, and attester state, allowing figures like revenue and profit to verify independently.

---

Contract Structure of an Attested Computation

An Attested Computation concept frontmatter and body specify everything required for execution and mechanical verification:

---
type: Attested Computation
title: Revenue for fiscal year
description: Recognized revenue for a fiscal year, per Finance's definition.
status: stable
runtime: bigquery
parameters:
  - { name: year, type: integer, required: true }
executor:
  resource: references/skills/run-on-bq.md
  receipt: [job_id, executed_sql, result]
attester:
  resource: references/attesters/sql-equality.py
generated: { by: reference_agent/gemini-2.5-pro, at: 2026-06-20T22:53:05Z }
verified: { by: human:ahormati, at: 2026-06-25T09:00:00Z }
stale_after: 2026-09-23
sources:
  - id: rev-policy
    resource: https://wiki.acme/finance/revenue-recognition
    title: Revenue recognition policy
---

# Computation

    SELECT SUM(amount) AS revenue
    FROM finance.recognized_revenue
    WHERE fiscal_year = @year

The computation binds only the declared `parameters`, per the recognition policy.[^rev-policy]

[^rev-policy]: Revenue recognition policy

Key Components:

• runtime: Identifies the execution environment (bigquery, dbt, etc.).
• parameters: Typed, named variables the agent is allowed to supply. Agents must not alter the underlying computation query itself.
• executor: Defines execution instructions (resource) and required return evidence (receipt).
• attester: Points to deterministic (no-LLM) code that inspects the receipt and returns a pass/fail verdict.

---

How Consumers Use Attested Computations

The OKF specification separates specification from runtime execution in 6 clear steps:

1. Discover: Agents locate the computation via type: Attested Computation or via links from narrative metrics.
2. Load: The consumer reads the contract and computation block.
3. Parameterize: The agent supplies valid values for declared parameters (e.g., year: 2026).
4. Execute: The executor runs the bound computation and returns a receipt (e.g., BigQuery job_id, executed_sql, and result).
5. Attest: The consumer runs the deterministic attester script over the receipt to verify that the query executed matched the sanctioned computation without unauthorized modifications.
6. Gate: The system surfaces the verified result or refuses stale/failing computations.

---

Verification vs. Attestation

It is crucial to distinguish between doc-level verification and runtime attestation:

• verified: Confirms that the metric definition matches business policy (stored in the bundle, updated periodically).
• Attestation: Confirms that a specific runtime execution produced the value correctly (per-call, un-stored receipt evaluation).

Conclusion

Attested Computations in OKF v0.2 transform AI-driven analytics from a “trust me” black box into a verifiable, deterministic science. By establishing clear contracts between definitions, parameters, executors, and deterministic attesters, organizations can deploy autonomous agents with absolute confidence in their data integrity.