SuperAPI AI Directory

avl-compare

Data & analytics

View repository

Updated 11-03 10:01

About this skill

AVL Compare

Purpose

Deterministic AVL comparison (incumbents, regional diversity, dual-source position) and the #5 recommendation banding with named conditions.

When to use

After risk-score in every run.

Inputs

scored.json + approved vendor list export.

Steps

  1. Validate against the contract.
  2. Run scripts/avl_match.py --scored scored.json --avl avl.json --out compared.json
  3. Quote the outcome and conditions verbatim; conditions are engine-derived from the escalations, never invented or dropped.

Output

compared.json - the {avl, recommendation} hop.

Grounding requirements

AVL facts cite the vendor-list export; the outcome cites rule #5.

Constraints

  • The model never upgrades an outcome; a do_not_qualify stands unless the commodity council overrides in writing (#5).

Escalation / uncertainty

Concentration (#3.1) and PPAP (#4.1) conditions are mandatory when flagged.