feat(sync): fix a session's source machine at first ingestion

Resolve the machine for a discovered session from the configured filesystem
root that produced it, and persist that label once. Every later write keeps the
stored value: ordinary reparses, appends, watcher refresh, provider refresh,
and trash handling all retain the database row's machine rather than adopting a
newly configured label. During sync --full the archive store points at the
original archive while e.db is the fresh replacement, so a full rebuild copies
existing labels forward instead of relabeling.

Editing a source label therefore affects only sessions discovered afterward.
Because relabeling is no longer part of sync, this drops the machine-only
update for trashed sessions, the all-machines baseline scan, and the second
ownership baseline index keyed without machine. Watch reconciliation continues
to use the baseline for the machine that originally admitted the session, so
per-event work stays bounded by the changed batch.

The previous approach chased reattribution through each new writer, trash path,
snapshot path, and reconciliation edge, which kept widening the
persistent-archive surface and reintroducing consistency bugs.

Author: mjacobs

cmd/agentsview/archive_query_backend.go

@@ -250,6 +250,7 @@ func (b localArchiveQueryBackend) SessionUsage(
 	if known && !b.skipFreshData {
 		engine := sync.NewEngine(b.database, sync.EngineConfig{
 			AgentDirs:               b.cfg.AgentDirs,
+			SourceMachines:          b.cfg.SourceMachines,
 			IncludeCwdPrefixes:      b.cfg.SyncIncludeCwdPrefixes,
 			Machine:                 b.cfg.LocalMachineName,
 			BlockedResultCategories: b.cfg.ResultContentBlockedCategories,

cmd/agentsview/archive_write_backend.go

@@ -839,6 +839,7 @@ func (b *localArchiveWriteBackend) DuckDBPushWatch(
 
 	engine := syncpkg.NewEngine(b.database, syncpkg.EngineConfig{
 		AgentDirs:               b.appCfg.AgentDirs,
+		SourceMachines:          b.appCfg.SourceMachines,
 		IncludeCwdPrefixes:      b.appCfg.SyncIncludeCwdPrefixes,
 		Machine:                 b.appCfg.LocalMachineName,
 		BlockedResultCategories: b.appCfg.ResultContentBlockedCategories,

cmd/agentsview/main.go

@@ -295,6 +295,7 @@ func runServe(cfg config.Config, opts serveOptions) {
 		var onStartupReconciled func(sync.SyncStats, error)
 		engine = sync.NewEngine(database, sync.EngineConfig{
 			AgentDirs:               cfg.AgentDirs,
+			SourceMachines:          cfg.SourceMachines,
 			IncludeCwdPrefixes:      cfg.SyncIncludeCwdPrefixes,
 			Machine:                 cfg.LocalMachineName,
 			BlockedResultCategories: cfg.ResultContentBlockedCategories,

cmd/agentsview/parse_diff.go

@@ -127,6 +127,7 @@ func doParseDiff(cfg ParseDiffConfig) (failed bool) {
 
 	engine := sync.NewDiffEngine(database, sync.EngineConfig{
 		AgentDirs:               appCfg.AgentDirs,
+		SourceMachines:          appCfg.SourceMachines,
 		IncludeCwdPrefixes:      appCfg.SyncIncludeCwdPrefixes,
 		Machine:                 appCfg.LocalMachineName,
 		BlockedResultCategories: appCfg.ResultContentBlockedCategories,

cmd/agentsview/session_sync.go

@@ -74,6 +74,7 @@ func syncService(
 	}
 	engine := sync.NewEngine(d, sync.EngineConfig{
 		AgentDirs:          cfg.AgentDirs,
+		SourceMachines:     cfg.SourceMachines,
 		IncludeCwdPrefixes: cfg.SyncIncludeCwdPrefixes,
 		Machine:            cfg.LocalMachineName,
 	})

cmd/agentsview/sync.go

@@ -884,6 +884,7 @@ func coordinateLocalSync(
 
 	engine := sync.NewEngine(database, sync.EngineConfig{
 		AgentDirs:               appCfg.AgentDirs,
+		SourceMachines:          appCfg.SourceMachines,
 		IncludeCwdPrefixes:      appCfg.SyncIncludeCwdPrefixes,
 		Machine:                 appCfg.LocalMachineName,
 		BlockedResultCategories: appCfg.ResultContentBlockedCategories,

cmd/agentsview/sync_worker.go

@@ -360,6 +360,7 @@ func openWorkerWriteDB(cfg config.Config) (*db.DB, *writeOwnerLock, error) {
 func workerEngineConfig(cfg config.Config) sync.EngineConfig {
 	return sync.EngineConfig{
 		AgentDirs:               cfg.AgentDirs,
+		SourceMachines:          cfg.SourceMachines,
 		IncludeCwdPrefixes:      cfg.SyncIncludeCwdPrefixes,
 		Machine:                 cfg.LocalMachineName,
 		BlockedResultCategories: cfg.ResultContentBlockedCategories,

cmd/agentsview/sync_worker_test.go

@@ -99,6 +99,28 @@ func TestSyncWorkerStartupModeSyncsAndEmitsTerminalResult(t *testing.T) {
 		"public SyncStats fields must survive the NDJSON protocol")
 }
 
+func TestSyncWorkerStartupUsesConfiguredSourceMachine(t *testing.T) {
+	cfg := testConfigWithClaudeFixture(t)
+	claudeRoot := cfg.AgentDirs[parser.AgentClaude][0]
+	cfg.SourceMachines = map[parser.AgentType]map[string]string{
+		parser.AgentClaude: {claudeRoot: "archivebox"},
+	}
+
+	var out bytes.Buffer
+	require.NoError(t, runSyncWorker(cfg, "startup", &out))
+	assert.Equal(t, "ok", decodeSingleResult(t, &out).Status)
+
+	database, err := db.OpenReadOnly(cfg.DBPath)
+	require.NoError(t, err)
+	defer database.Close()
+	page, err := database.ListSessions(context.Background(), db.SessionFilter{})
+	require.NoError(t, err)
+	require.Len(t, page.Sessions, 3)
+	for _, sess := range page.Sessions {
+		assert.Equal(t, "archivebox", sess.Machine)
+	}
+}
+
 func TestSyncWorkerReportsAbortAsFailure(t *testing.T) {
 	cfg := testConfigWithClaudeFixture(t)
 	ctx, cancel := context.WithCancel(context.Background())

cmd/agentsview/usage.go

@@ -312,6 +312,7 @@ func ensureFreshData(
 	if database.NeedsResync() {
 		engine := sync.NewEngine(database, sync.EngineConfig{
 			AgentDirs:          appCfg.AgentDirs,
+			SourceMachines:     appCfg.SourceMachines,
 			IncludeCwdPrefixes: appCfg.SyncIncludeCwdPrefixes,
 			Machine:            appCfg.LocalMachineName,
 		})
@@ -336,6 +337,7 @@ func ensureFreshData(
 
 	engine := sync.NewEngine(database, sync.EngineConfig{
 		AgentDirs:          appCfg.AgentDirs,
+		SourceMachines:     appCfg.SourceMachines,
 		IncludeCwdPrefixes: appCfg.SyncIncludeCwdPrefixes,
 		Machine:            appCfg.LocalMachineName,
 	})

internal/db/project_identity_test.go

@@ -81,6 +81,63 @@ func TestProjectIdentityPublicationRevisionTracksSnapshotChanges(t *testing.T) {
 	assert.Greater(t, afterDelete, afterInsert)
 }
 
+func TestCopySessionMetadataPreservesRecordedMachineAttribution(
+	t *testing.T,
+) {
+	dir := t.TempDir()
+	sourcePath := filepath.Join(dir, "source.db")
+	source := testDBAtPath(t, sourcePath, "source")
+	require.NoError(t, source.UpsertSession(Session{
+		ID: "resynced-session", Project: "app",
+		Machine: "oldbox", Agent: "claude",
+	}))
+	require.NoError(t, source.UpsertProjectIdentityObservation(
+		t.Context(), export.ProjectIdentityObservation{
+			SessionID: "resynced-session", Project: "app", Machine: "oldbox",
+			RootPath: "/workspace/app", ObservedAt: time.Now().UTC(),
+		},
+	))
+	require.NoError(t, source.Close())
+
+	destinationPath := filepath.Join(dir, "destination.db")
+	destination := testDBAtPath(t, destinationPath, "destination")
+	t.Cleanup(func() { _ = destination.Close() })
+	require.NoError(t, destination.UpsertSession(Session{
+		ID: "resynced-session", Project: "app",
+		Machine: "newbox", Agent: "claude",
+	}))
+	require.NoError(t, destination.UpsertProjectIdentityObservation(
+		t.Context(), export.ProjectIdentityObservation{
+			SessionID: "resynced-session", Project: "app", Machine: "newbox",
+			RootPath: "/workspace/app", ObservedAt: time.Now().UTC(),
+		},
+	))
+
+	require.NoError(t, destination.CopySessionMetadataFrom(sourcePath))
+
+	session, err := destination.GetSessionFull(
+		t.Context(), "resynced-session",
+	)
+	require.NoError(t, err)
+	require.NotNil(t, session)
+	assert.Equal(t, "newbox", session.Machine)
+	snapshots, err := destination.ListSessionProjectIdentitySnapshots(
+		t.Context(),
+	)
+	require.NoError(t, err)
+	require.Len(t, snapshots, 1)
+	assert.Equal(t, "oldbox", snapshots[0].Machine)
+	observations, err := destination.ListProjectIdentityObservations(
+		t.Context(), []string{"app"},
+	)
+	require.NoError(t, err)
+	require.Len(t, observations, 2)
+	assert.ElementsMatch(t, []string{"newbox", "oldbox"}, []string{
+		observations[0].Machine,
+		observations[1].Machine,
+	})
+}
+
 func TestSessionProjectIdentitySnapshotPreservesFirstRootKeyUntilRemoteEvidence(
 	t *testing.T,
 ) {