Firebreak.RiskScore (Firebreak v0.1.0)

Copy Markdown View Source

A single, transparent supervision-risk number per project (and per supervisor), derived from the model IR — for dashboards and for tracking whether a tree is getting riskier over time.

It is a weighted count of structural hazards, not an absolute grade. Three components, each read straight off the IR:

  • sync_crossings — synchronous cross-tree crossings (a caller in one subtree blocking on a process in another). The highest-signal hazard.
  • blast_radius — children that co-restart together under :one_for_all / :rest_for_one (a wide shared-fate set).
  • intensity_pressure — supervisors running many children (≥5) on a tight restart budget (≤3 restarts), where one crash-loop takes the whole subtree.

The weights are relative, not calibrated absolutes — the value is the trend and the per-supervisor ranking, which point at where to look. Every score ships with its component breakdown, so it's never a black box. A clean app scores 0.

Summary

Types

t()

Functions

json(a)

The risk score as a JSON string (mix firebreak --format score).

score(a)

The risk score for an analysis: total, component breakdown, and per-supervisor ranking.

Types

t()

@type t() :: %{
  score: non_neg_integer(),
  components: %{
    sync_crossings: non_neg_integer(),
    blast_radius: non_neg_integer(),
    intensity_pressure: non_neg_integer()
  },
  weights: map(),
  per_supervisor: [map()]
}

Functions

json(a)

@spec json(Firebreak.Analysis.t()) :: String.t()

The risk score as a JSON string (mix firebreak --format score).

score(a)

@spec score(Firebreak.Analysis.t()) :: t()

The risk score for an analysis: total, component breakdown, and per-supervisor ranking.