feat(worker): 四类进度用 manifest 真实总量，未知保持 NULL

为什么：P02 要求分项计数/字节来自真实清单，不能永远 NULL 或写 0。

cmd/worker/library_handlers.go

@@ -13,6 +13,7 @@ import (
 	"knowledge-governance/internal/governance"
 	"knowledge-governance/internal/jobs"
 	"knowledge-governance/internal/libraryrelease"
+	"knowledge-governance/internal/librarysnapshot"
 	postgresstore "knowledge-governance/internal/platform/postgres"
 )
 
@@ -37,6 +38,13 @@ func runLibraryRelease(ctx context.Context, store *postgresstore.Store, config a
 	now := time.Now().UTC()
 	_ = store.AdvanceLibraryJob(ctx, payload.LibraryJobID, governance.LibraryJobRunning, "prebackup", "", "", "", "", now)
 	recorder := libraryStageRecorder{store: store, jobID: payload.LibraryJobID, scope: payload.Scope}
+	// 四类进度的真实总量来自快照清单（manifest）：读不到就保持 NULL，不写 0。
+	if archive, openErr := librarysnapshot.Open(config.LibraryArchiveRoot, 256*1024*1024); openErr == nil {
+		if manifest, loadErr := archive.Load(payload.TargetSnapshotID); loadErr == nil {
+			recorder.totals = stageSizesFrom(manifest)
+		}
+		archive.Close()
+	}
 
 	// 崩溃接管：作业里已经持久化过预备份快照，就直接复用它继续，
 	// 绝不重新备份——重新备份会把"半写状态"当成新基线，覆盖真正可回滚的那份。

cmd/worker/main.go

@@ -17,6 +17,7 @@ import (
 	"knowledge-governance/internal/governance"
 	"knowledge-governance/internal/jobs"
 	"knowledge-governance/internal/libraryagent"
+	"knowledge-governance/internal/librarysnapshot"
 	postgresstore "knowledge-governance/internal/platform/postgres"
 	"knowledge-governance/internal/platform/weknora"
 	"knowledge-governance/internal/versionhistory"
@@ -28,19 +29,55 @@ import (
 
 // libraryStageRecorder 把执行阶段写进持久作业，页面刷新/worker 重启后仍可复现。
 type libraryStageRecorder struct {
-	store *postgresstore.Store
-	jobID string
-	scope string
+	store  *postgresstore.Store
+	jobID  string
+	scope  string
+	totals map[string]stageSize
 	// mappingDigest 是本次读到的生产索引映射，随预备份一起登记成版本证据。
 	mappingDigest string
 }
 
 func (r libraryStageRecorder) Stage(ctx context.Context, stage, state, detail string) {
-	if err := r.store.UpdateLibraryStage(ctx, r.jobID, stage, state, nil, nil, nil, nil, detail, time.Now().UTC()); err != nil {
+	// 真实总量来自快照 manifest：已知就写真实数字，未知就保持 NULL（不写 0、不编百分比）。
+	var total, completed, bytesTotal, bytesDone *int64
+	if size, ok := r.totals[stage]; ok {
+		total, bytesTotal = int64Ptr(size.count), int64Ptr(size.bytes)
+		if state == "succeeded" {
+			completed, bytesDone = total, bytesTotal
+		}
+	}
+	if err := r.store.UpdateLibraryStage(ctx, r.jobID, stage, state, total, completed, bytesTotal, bytesDone,
+		detail, time.Now().UTC()); err != nil {
 		log.Printf("library stage %s/%s update failed: %v", r.jobID, stage, err)
 	}
 }
 
+func int64Ptr(value int64) *int64 { return &value }
+
+// stageSize 是某一类数据在快照里的真实总量（行数/字节）。
+type stageSize struct {
+	count int64
+	bytes int64
+}
+
+// stageSizesFrom 从 manifest 汇总四类进度的真实总量：
+// database/resources/index/wiki 各自的行数与字节，全部来自归档部件清单。
+func stageSizesFrom(manifest librarysnapshot.Manifest) map[string]stageSize {
+	sizes := map[string]stageSize{}
+	kindFor := map[string]string{"database": "database", "resource": "resources", "index": "index", "wiki": "wiki"}
+	for _, part := range manifest.Parts {
+		stage, ok := kindFor[part.Kind]
+		if !ok {
+			continue
+		}
+		current := sizes[stage]
+		current.count += part.Count
+		current.bytes += part.Bytes
+		sizes[stage] = current
+	}
+	return sizes
+}
+
 // libraryCaptureJobType is shared with the API which enqueues it.
 const libraryCaptureJobType = "library.capture"