| name | vllm-backport-score-rank |
| description | Score and rank backport candidates using a composite formula based on verdict, severity, scope, risk, and self-containedness. Use after the agent completes semantic analysis to produce a prioritized ranked list.
|
| compatibility | Requires python3 |
| allowed-tools | Bash Read |
| metadata | {"author":"ZhanqiuHu","version":"1.0","tags":"vllm, backport, triage"} |
Score and Rank
Applies a deterministic scoring formula to produce reproducible rankings.
Scoring Formula
| Field | Values and Points |
|---|
| verdict | must_backport=30, likely_relevant=20, needs_review=10, likely_skip/skip=0 |
| severity | critical=25, moderate=15, low=5 |
| affected_scope | all_users=20, specific_models=12, specific_feature=8, edge_case=3 |
| backport_risk | safe=15, moderate=8, risky=0 |
| self_contained | true=10, false=0 |
Max score: 100. Sorted by score desc, then files_in_release desc, then change_size asc.
Each PR also gets backport_ease: ai-fixable if self_contained AND risk is safe/moderate.
Usage
python3 scripts/score-and-rank.py \
--input artifacts/backport-triage/analyzed.json \
--output artifacts/backport-triage/ranked.json
Input
analyzed.json — candidates with agent-added fields: verdict, severity,
affected_scope, backport_risk, self_contained.
Output
ranked.json — filtered (removes SKIP/already_backported), scored, sorted,
with rank, score, change_size, backport_ease added.