Skip to content

Expose per-edge confidence in affected output (+ affected --json mode) #2352

Description

@HigorAlves

Summary

graphify affected "X" is excellent for reverse blast-radius, but its text output omits the per-edge confidence tag that already exists in graph.json (EXTRACTED / INFERRED / AMBIGUOUS). Tools that turn blast-radius into assertions — e.g. a code reviewer flagging "you forgot to update caller Z" — need to gate on trustworthy edges, and can't today.

Current output (graphify 0.9.30)

- applyToolsAction() [calls] cli/internal/cli/appactions.go:L165

The relation is shown; the confidence is not.

Request

  1. Include each edge's confidence inline, e.g. - applyToolsAction() [calls, EXTRACTED] cli/internal/cli/appactions.go:L165.
  2. Add an affected --json mode mirroring god-nodes --json, emitting {id, label, relation, confidence, confidence_score, source_file, source_location} per affected node.

Why

It lets a consumer floor a finding's confidence by the weakest edge on the path and drop AMBIGUOUS-only leads — turning affected from a "read these candidates" accelerator into a safe finding source. Without it, blast-radius findings can't be trusted for correctness-critical uses like review.

Verified against graphify 0.9.30.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions