Deterministic rating engine
The system decides scores and rankings. The LLM narration layer only rephrases an approved structured payload — it never decides, calculates, filters, or ranks.
Admin panel →
Browse, re-score, set editor adjustments, run narrations, bulk re-score.
README ↗
Architecture, three-axis model, hard rules, known limitations.
Active engine version
- weights
- v2
- scoring
- v7
- explain
- v5
- display
- v4
- narrate
- v3
Every rating_runs row carries the triplet (weights / scoring / explain) so a score change can be traced to the layer that produced it.
Raw0–100
System math. Internal scale.
Display0–10
Curved for UI readability.
Confidence0–1
How much we trust the score.