01
Allowlisted Signal Categories
Kalpae only ingests domains approved by category policy before evidence extraction runs.
- Macro: 5 primary, 4 corroborative, 3 context domains.
- Policy: 5 primary, 4 corroborative, 3 context domains.
- Technology: 5 primary, 4 corroborative, 3 context domains.
- Climate: 5 primary, 4 corroborative, 3 context domains.
02
CRE Processing Runtime
Chainlink CRE jobs are claimed from queued windows and executed continuously using deterministic run controls.
- No research run has been recorded yet.
- Failed runs recorded: 0.
- Anchor failures recorded: 0.
- Pending review updates: 0.
- No evidence document has been ingested yet.
- No run commitment has been anchored yet.
03
Evidence Weighting
Each accepted source contributes by direction, relevance, confidence, domain tier, and novelty factor.
- Primary tier weight: 1.00.
- Corroborative tier weight: 0.60.
- Context tier weight: 0.30.
- Delta caps and review thresholds are enforced in CRE before probabilities are applied.
04
Conflict And Fallback Handling
When data is missing or contradictory, runs are constrained by retries and explicit review gates.
- Only allowlisted domains are considered eligible for ingestion.
- Duplicate evidence hashes are ignored to prevent replay impact.
- Large deltas are routed to pending review rather than auto-applied.
- Run failures are retried up to configured limits with scheduled backoff.