The AGIsGEM verification engine.
AGIsGEM is building a controlled workflow for organising supplied agricultural evidence, identifying gaps and producing explainable, versioned decision-support reports.
Data ingestion layer
The founding pilot begins with customer-supplied contracts, entity records, asset schedules and operational evidence. Approved registry, saleyard, carbon and marketplace connections will be introduced only after access, testing and legal review. Each report states which sources were supplied, independently checked or not available.
Verification engine — the G.A.M.E. framework
Each contract or asset is processed through a three-layer model:
- Perception — extract and normalise the underlying records.
- Cognitive Auditor — apply rule-based and statistical checks: provenance integrity, weight & health consistency, sale cross-validation, counterparty risk flags.
- Action Layer — emit a Logic Score, risk flags, and a structured proof artefact.
Logic Score
The Logic Score is a 0–100 composite reflecting verification confidence across provenance, integrity, and risk dimensions. Each score is published with its rubric version, the data window it was computed against, and the specific flags driving its result. A score is not a recommendation — it is a transparent, reproducible signal.
Proof anchoring
The target workflow hashes final reports and may pin approved artefacts to IPFS and reference them on Base. Until this capability is tested on an issued pilot report, it remains a planned proof layer rather than a live verification claim.
Outputs
- Logic Score PDF + machine-readable JSON
- Unique report ID and version history
- Optional hash, IPFS and API outputs as each capability is commissioned and tested
- Public status entry for badges approved for publication