pub struct RepoRecall {Show 23 fields
pub repo: String,
pub architecture: f64,
pub entrypoints: f64,
pub behavior: f64,
pub state_authority: f64,
pub contracts: f64,
pub landmarks: f64,
pub tests: f64,
pub overall: f64,
pub precision: f64,
pub layer_precision: BTreeMap<String, f64>,
pub layer_f2: BTreeMap<String, f64>,
pub f2: f64,
pub density: f64,
pub atlas_tokens: usize,
pub resolved_calls: usize,
pub backends_used: Vec<String>,
pub backends_missing: Vec<String>,
pub resolve_error: Option<String>,
pub landmark_items: usize,
pub skipped_reason: Option<String>,
pub missed: Vec<String>,
pub gaps: Vec<GapFinding>,
}Fields§
§repo: String§architecture: f64§entrypoints: f64§behavior: f64§contracts: f64§landmarks: f64§tests: f64§overall: f64Equal-weighted mean of the five startup-required layers (the gate).
precision: f64Mean startup-required precision over the five startup-required
layers: per layer, |atlas entries that match a ground-truth item| /
|atlas entries in the layer| (v3 — a too-much-architecture
detector; see layer_precision).
layer_precision: BTreeMap<String, f64>Per-layer startup-required precision (the five startup layers only).
layer_f2: BTreeMap<String, f64>Per-layer F2 = (5PR)/(4*P+R) over the five startup-required
layers (zero when P+R==0); f2 is the equal-weighted mean.
f2: f64Equal-weighted mean F2 over the five startup-required layers.
density: f64Matched startup-required items per 1000 atlas tokens (architecture_density).
atlas_tokens: usizeRendered atlas token count.
resolved_calls: usizeNumber of call edges upgraded to RESOLVED by the semantic backends
(pyright/tsserver) before scoring; 0 when --no-resolve.
backends_used: Vec<String>Semantic backends that were available and used for resolution
(pyright/tsserver); empty when unavailable or --no-resolve.
backends_missing: Vec<String>Semantic backends unavailable (not installed) — resolution degraded, not fatal.
resolve_error: Option<String>Resolution error text when the backend was found but the resolve pass failed (e.g. LSP handshake) — degraded, not fatal.
landmark_items: usizeNumber of ground-truth items in the (informational) landmarks layer.
skipped_reason: Option<String>When set, the repo was not scored (missing dir / missing ground truth / index failure) and the recall fields are meaningless.
missed: Vec<String>Ground-truth key strings the structured matcher missed
(section:key), for diagnosing misses.
gaps: Vec<GapFinding>Per-item gap classification (populated when --diagnose).