add web service

internal/search/scoring.go

@@ -0,0 +1,67 @@
+package search
+
+import (
+	"math/rand/v2"
+	"time"
+
+	"git.vakhrushev.me/av/remembos/internal/memos"
+)
+
+// Memory is the result of the search algorithm.
+type Memory struct {
+	Memo      *memos.Memo
+	Tier      int
+	YearsAgo  int
+	ShowCount int
+	Date      time.Time
+}
+
+// candidate is an intermediate struct during scoring.
+type candidate struct {
+	memo      *memos.Memo
+	yearsAgo  int
+	showCount int
+}
+
+// weightedSelect picks one candidate using weighted random selection.
+// preferOlder gives higher weight to older memos. maxYearsBack normalizes recency.
+func weightedSelect(candidates []candidate, preferOlder bool, maxYearsBack int) *candidate {
+	if len(candidates) == 0 {
+		return nil
+	}
+	if len(candidates) == 1 {
+		return &candidates[0]
+	}
+
+	scores := make([]float64, len(candidates))
+	var total float64
+
+	for i, c := range candidates {
+		var recencyFactor float64
+		if preferOlder && maxYearsBack > 0 && c.yearsAgo > 0 {
+			recencyFactor = float64(c.yearsAgo) / float64(maxYearsBack)
+		} else {
+			recencyFactor = 1.0
+		}
+
+		showCountPenalty := 1.0 / (1.0 + float64(c.showCount))
+
+		score := recencyFactor * showCountPenalty
+		if score < 0.01 {
+			score = 0.01 // minimum weight
+		}
+		scores[i] = score
+		total += score
+	}
+
+	r := rand.Float64() * total
+	var cumulative float64
+	for i, s := range scores {
+		cumulative += s
+		if r <= cumulative {
+			return &candidates[i]
+		}
+	}
+
+	return &candidates[len(candidates)-1]
+}