feat(backend): report stored pairwise distances in the duplicates response

ListVisible already loads each pair's Hamming distance, but clusterPairs threw
it away. Thread it through: Clusters now returns a Cluster carrying the stored
pairwise distances (indexed once per page), and the list endpoint emits them as
{a, b, distance}. Pairs linked only transitively have no stored distance and are
omitted. Lets the UI show how close each file is to the kept one without a
client-side hash compare (phash exceeds JS's safe-integer range).

backend/internal/handler/duplicate_handler.go

@@ -47,12 +47,16 @@ func (h *DuplicateHandler) List(c *gin.Context) {
 	}
 
 	items := make([]gin.H, len(clusters))
-	for i, files := range clusters {
-		fs := make([]fileJSON, len(files))
-		for j, f := range files {
+	for i, cl := range clusters {
+		fs := make([]fileJSON, len(cl.Files))
+		for j, f := range cl.Files {
 			fs[j] = toFileJSON(f)
 		}
-		items[i] = gin.H{"files": fs}
+		dists := make([]gin.H, len(cl.Distances))
+		for j, d := range cl.Distances {
+			dists[j] = gin.H{"a": d.A, "b": d.B, "distance": d.Distance}
+		}
+		items[i] = gin.H{"files": fs, "distances": dists}
 	}
 	respondJSON(c, http.StatusOK, gin.H{
 		"items":  items,

backend/internal/integration/server_test.go

@@ -1661,6 +1661,11 @@ type dupListResponse struct {
 				ID string `json:"id"`
 			} `json:"tags"`
 		} `json:"files"`
+		Distances []struct {
+			A        string `json:"a"`
+			B        string `json:"b"`
+			Distance int    `json:"distance"`
+		} `json:"distances"`
 	} `json:"items"`
 	Total int `json:"total"`
 }
@@ -1703,6 +1708,9 @@ func TestDuplicateDetection(t *testing.T) {
 	require.Equal(t, 1, list.Total, "expected one duplicate cluster: %s", resp)
 	require.Len(t, list.Items, 1)
 	require.Len(t, list.Items[0].Files, 2)
+	// The pair's stored distance rides along; identical 1×1 uploads are distance 0.
+	require.Len(t, list.Items[0].Distances, 1, "the pair's distance should be reported")
+	assert.Equal(t, 0, list.Items[0].Distances[0].Distance)
 
 	// --- resolve: keep f1, union tags from f2, trash f2 ----------------------
 	resp = h.doJSON("POST", "/files/duplicates/resolve", map[string]any{

backend/internal/service/duplicate_index.go

@@ -103,6 +103,31 @@ func buildPairs(entries []domain.PHashEntry, threshold int, onProgress func(done
 	return pairs
 }
 
+// orderedPair returns the two ids in canonical (a < b by UUID byte order) order,
+// matching how the pairs table keys a distance so a lookup hits regardless of the
+// argument order.
+func orderedPair(a, b uuid.UUID) [2]uuid.UUID {
+	if bytes.Compare(a[:], b[:]) > 0 {
+		return [2]uuid.UUID{b, a}
+	}
+	return [2]uuid.UUID{a, b}
+}
+
+// clusterDistances returns the stored Hamming distance for every pair of files in
+// the cluster that has one. Pairs present only transitively have no stored
+// distance and are left out.
+func clusterDistances(files []domain.File, distByPair map[[2]uuid.UUID]int) []PairDistance {
+	var out []PairDistance
+	for i := 0; i < len(files); i++ {
+		for j := i + 1; j < len(files); j++ {
+			if d, ok := distByPair[orderedPair(files[i].ID, files[j].ID)]; ok {
+				out = append(out, PairDistance{A: files[i].ID, B: files[j].ID, Distance: d})
+			}
+		}
+	}
+	return out
+}
+
 // clusterPairs groups pairs into connected components (transitive closure) via
 // union-find. Every returned cluster has at least two files; clusters and the ids
 // within them are sorted by UUID for stable pagination.

backend/internal/service/duplicate_service.go

@@ -125,10 +125,27 @@ func NewDuplicateService(
 	}
 }
 
+// Cluster is a group of near-duplicate files together with the pairwise Hamming
+// distances known between them. Distances are read from the stored pairs, so two
+// files linked into the cluster only transitively (through an intermediate) may
+// have no direct distance — that pair is simply omitted.
+type Cluster struct {
+	Files     []domain.File
+	Distances []PairDistance
+}
+
+// PairDistance is the stored Hamming distance between two files of a cluster.
+type PairDistance struct {
+	A        uuid.UUID
+	B        uuid.UUID
+	Distance int
+}
+
 // Clusters returns a page of duplicate clusters visible to the caller. Pairs are
 // read from the precomputed table (no all-pairs scan here) and grouped into
-// connected components; pagination is over whole clusters.
-func (s *DuplicateService) Clusters(ctx context.Context, limit, offset int) (clusters [][]domain.File, total int, err error) {
+// connected components; pagination is over whole clusters. Each cluster carries
+// the stored pairwise distances so callers can show how close the files are.
+func (s *DuplicateService) Clusters(ctx context.Context, limit, offset int) (clusters []Cluster, total int, err error) {
 	userID, isAdmin, _ := domain.UserFromContext(ctx)
 
 	pairs, err := s.pairs.ListVisible(ctx, userID, isAdmin)
@@ -142,14 +159,20 @@ func (s *DuplicateService) Clusters(ctx context.Context, limit, offset int) (clu
 		offset = 0
 	}
 	if offset >= len(groups) {
-		return [][]domain.File{}, total, nil
+		return []Cluster{}, total, nil
 	}
 	end := offset + limit
 	if end > len(groups) || limit <= 0 {
 		end = len(groups)
 	}
 
-	out := make([][]domain.File, 0, end-offset)
+	// Index the stored distances once; each page cluster looks up its own pairs.
+	distByPair := make(map[[2]uuid.UUID]int, len(pairs))
+	for _, p := range pairs {
+		distByPair[orderedPair(p.FileA, p.FileB)] = p.Distance
+	}
+
+	out := make([]Cluster, 0, end-offset)
 	for _, ids := range groups[offset:end] {
 		files := make([]domain.File, 0, len(ids))
 		for _, id := range ids {
@@ -164,7 +187,7 @@ func (s *DuplicateService) Clusters(ctx context.Context, limit, offset int) (clu
 			files = append(files, *f)
 		}
 		if len(files) >= 2 {
-			out = append(out, files)
+			out = append(out, Cluster{Files: files, Distances: clusterDistances(files, distByPair)})
 		}
 	}
 	return out, total, nil