fix: add a grace period before GC deletes orphaned blobs

FindOrphanedBlobs returned any unreferenced blob, so a concurrent dedup
upload (which inserts the blob row before its artifact/local_files row)
could have its S3 object deleted mid-flight. Restrict collection to blobs
older than a 1h grace period.

Refs #71

internal/database/artifacts.go

@@ -109,16 +109,22 @@ func (db *DB) InsertAccessLog(ctx context.Context, remoteName, path string, cach
 	return err
 }
 
-func (db *DB) FindOrphanedBlobs(ctx context.Context) ([]models.Blob, error) {
+// FindOrphanedBlobs returns blobs no longer referenced by any artifact or
+// local file, restricted to those created before now()-minAge. The age cutoff
+// is a grace period that avoids a TOCTOU race with in-flight dedup uploads,
+// which insert the blob row before the referencing artifact/local_files row.
+func (db *DB) FindOrphanedBlobs(ctx context.Context, minAge time.Duration) ([]models.Blob, error) {
+	cutoff := time.Now().Add(-minAge)
 	rows, err := db.Pool.Query(ctx, `
 		SELECT b.content_hash, b.s3_key, b.size_bytes, b.content_type, b.created_at
 		FROM blobs b
-		WHERE b.content_hash NOT IN (
+		WHERE b.created_at < $1
+		  AND b.content_hash NOT IN (
 			SELECT content_hash FROM artifacts
 			UNION
 			SELECT content_hash FROM local_files
 		)
-	`)
+	`, cutoff)
 	if err != nil {
 		return nil, err
 	}

internal/database/stats.go

@@ -30,6 +30,15 @@ func (db *DB) GetOverviewStats(ctx context.Context) (*models.OverviewStats, erro
 		return nil, err
 	}
 
+	err = db.Pool.QueryRow(ctx, `
+		SELECT COALESCE(SUM(size_bytes), 0)
+		FROM access_log
+		WHERE cache_hit = TRUE AND created_at > NOW() - INTERVAL '30 days'
+	`).Scan(&stats.BandwidthSaved30d)
+	if err != nil {
+		return nil, err
+	}
+
 	return &stats, nil
 }
 

internal/gc/gc.go

@@ -9,6 +9,11 @@ import (
 	"git.unkin.net/unkin/artifactapi/internal/storage"
 )
 
+// blobGracePeriod is how old an orphaned blob must be before GC will delete
+// it. This avoids racing in-flight dedup uploads that insert the blob row
+// before the referencing artifact/local_files row exists.
+const blobGracePeriod = 1 * time.Hour
+
 type Collector struct {
 	db       *database.DB
 	store    *storage.S3
@@ -38,7 +43,7 @@ func (c *Collector) Run(ctx context.Context) {
 func (c *Collector) sweep(ctx context.Context) {
 	start := time.Now()
 
-	orphaned, err := c.db.FindOrphanedBlobs(ctx)
+	orphaned, err := c.db.FindOrphanedBlobs(ctx, blobGracePeriod)
 	if err != nil {
 		slog.Error("gc: find orphaned blobs", "error", err)
 		return

internal/virtual/helm_merger.go

@@ -65,11 +65,12 @@ func (m *HelmMerger) MergeIndexes(members []MemberIndex, proxyBaseURL string) ([
 							if baseHost != "" && extractHost(u) != baseHost {
 								continue
 							}
+							relPath := extractPathRelativeToBase(u, member.BaseURL)
 							ver.URLs[i] = fmt.Sprintf("%s/api/v1/%s/%s/%s",
 								strings.TrimRight(proxyBaseURL, "/"),
 								routePrefix,
 								member.RemoteName,
-								extractPath(u))
+								relPath)
 						} else {
 							ver.URLs[i] = fmt.Sprintf("%s/api/v1/%s/%s/%s",
 								strings.TrimRight(proxyBaseURL, "/"),
@@ -102,6 +103,18 @@ func extractHost(rawURL string) string {
 	return rest[:slashIdx]
 }
 
+func extractPathRelativeToBase(rawURL, baseURL string) string {
+	fullPath := extractPath(rawURL)
+	basePath := extractPath(baseURL)
+	if basePath != "" {
+		basePath = strings.TrimRight(basePath, "/") + "/"
+		if strings.HasPrefix(fullPath, basePath) {
+			return fullPath[len(basePath):]
+		}
+	}
+	return fullPath
+}
+
 func extractPath(rawURL string) string {
 	idx := strings.Index(rawURL, "://")
 	if idx == -1 {

ui/src/pages/Dashboard.tsx

@@ -50,6 +50,11 @@ export function Dashboard() {
           value={formatNumber(stats.total_blobs_deduped)}
           sub="shared blobs"
         />
+        <StatsCard
+          label="Bandwidth Saved"
+          value={formatBytes(stats.bandwidth_saved_30d)}
+          sub="last 30 days"
+        />
       </div>
 
       {health && (