pub struct SearchHit {
pub id: PlanPointId,
pub score: f64,
pub payload: Option<HashMap<String, Value>>,
pub collection: Option<String>,
pub vector: Option<PlanVectorStruct>,
}Expand description
Normalized search hit returned inside ExecResponse data.
Fields§
§id: PlanPointIdPoint ID (integer or string/UUID).
score: f64Similarity or rerank score, rounded once at ingestion via
score_f64 so it holds the shortest round-trip decimal (0.95,
not 0.949999988079071): Qdrant’s own JSON text and the native Python
ScoredPoint.score getter both use that form, so the JSON report view
and the typed view agree with plain Serialize (no custom
serialize_with).
Deliberate asymmetry with PlanVectorStruct: scalar scores are
human-read values worth normalizing, while vector stays Vec<f32>
passthrough — exact decimal equality on 128-dim vectors is
meaningless, and doubling their memory for cosmetics is indefensible.
payload: Option<HashMap<String, Value>>Point payload when requested via WITH PAYLOAD.
collection: Option<String>Source collection. Populated by cross-collection operations (e.g. CROSS RERANK) so results are unambiguous when multiple collections share the same point id.
vector: Option<PlanVectorStruct>Vector(s) returned when requested via WITH VECTOR.