pub struct RerankExecutedPayload {Show 13 fields
pub reranker: RerankerKind,
pub served_by_profile_id: Option<String>,
pub model_id: Option<Id128>,
pub query_id: Option<String>,
pub tiers: Vec<String>,
pub ignored_weights: Vec<String>,
pub unidentified_candidates: usize,
pub candidates: Vec<Id128>,
pub reranked: Vec<(Id128, Vec<(String, f32)>)>,
pub final_scores: Vec<(Id128, f32)>,
pub latency_us: u64,
pub hook_applied: bool,
pub hook_target_match: bool,
}Expand description
Payload for a rerank pass event, recording per-candidate scores.
All score values (reranked section scores, final_scores) must be finite.
When the serde feature is enabled, deserialization rejects non-finite scores.
Fields§
§reranker: RerankerKindNative model or weighted feature scoring; legacy events default to native.
served_by_profile_id: Option<String>Brain profile that served this rerank, if any.
model_id: Option<Id128>Model used for reranking; absent for weighted feature scoring.
query_id: Option<String>Opaque caller-supplied query correlation; never generated or parsed as a UUID.
tiers: Vec<String>Positive recognized feature weights, in scoring order.
ignored_weights: Vec<String>Configured keys unknown to weighted scoring, in lexical order.
unidentified_candidates: usizeCandidates without a parseable UUID, omitted from the three ID arrays.
candidates: Vec<Id128>Candidate IDs in input order.
reranked: Vec<(Id128, Vec<(String, f32)>)>Per-candidate named sub-scores from the reranker.
final_scores: Vec<(Id128, f32)>Final aggregated score per candidate.
latency_us: u64Wall-clock latency of the rerank operation in microseconds.
hook_applied: boolWhether a brain hook was applied during this rerank.
hook_target_match: boolWhether the hook matched the intended target.