@@ -75,21 +75,41 @@ fn load_groups_from_sources(
7575 kind : AgentReportKind ,
7676) -> Result < BTreeMap < String , CodexGroup > > {
7777 let file_groups = paths:: collect_deduped_codex_usage_files ( sources) ;
78- let replay_plan = CodexReplayPlan :: new (
79- file_groups
80- . iter ( )
81- . map ( |group| ( group. dir . as_path ( ) , group. files . as_slice ( ) ) ) ,
82- shared. single_thread ,
83- ) ;
78+ let files_by_group = file_groups
79+ . iter ( )
80+ . map ( |group| paths:: filter_codex_usage_files ( & group. dir , & group. files , shared) )
81+ . collect :: < Vec < _ > > ( ) ;
82+ let replay_plan = if shared. since . is_some ( ) || shared. until . is_some ( ) {
83+ CodexReplayPlan :: for_bounded_files (
84+ file_groups
85+ . iter ( )
86+ . zip ( & files_by_group)
87+ . map ( |( group, files) | ( group. dir . as_path ( ) , files. as_slice ( ) ) ) ,
88+ file_groups
89+ . iter ( )
90+ . map ( |group| ( group. dir . as_path ( ) , group. files . as_slice ( ) ) ) ,
91+ shared. single_thread ,
92+ )
93+ } else {
94+ CodexReplayPlan :: new (
95+ file_groups
96+ . iter ( )
97+ . map ( |group| ( group. dir . as_path ( ) , group. files . as_slice ( ) ) ) ,
98+ shared. single_thread ,
99+ )
100+ } ;
84101 let mut groups = BTreeMap :: new ( ) ;
85102 let seen = create_dedupe_shards ( ) ;
86- for group in & file_groups {
103+ for ( group, files) in file_groups. iter ( ) . zip ( & files_by_group) {
104+ if files. is_empty ( ) {
105+ continue ;
106+ }
87107 merge_groups (
88108 & mut groups,
89109 aggregate_files_with_dedupe (
90110 & CodexAggregateRun {
91111 sessions_dir : & group. dir ,
92- files : & group . files ,
112+ files : & files,
93113 shared,
94114 kind,
95115 replay_plan : & replay_plan,
@@ -107,9 +127,17 @@ pub(super) fn load_groups_from_directory(
107127 shared : & SharedArgs ,
108128 kind : AgentReportKind ,
109129) -> Result < BTreeMap < String , CodexGroup > > {
110- let files = paths:: collect_codex_usage_files ( sessions_dir) ;
111- let replay_plan =
112- CodexReplayPlan :: new ( [ ( sessions_dir, files. as_slice ( ) ) ] , shared. single_thread ) ;
130+ let all_files = paths:: collect_codex_usage_files ( sessions_dir) ;
131+ let files = paths:: filter_codex_usage_files ( sessions_dir, & all_files, shared) ;
132+ let replay_plan = if shared. since . is_some ( ) || shared. until . is_some ( ) {
133+ CodexReplayPlan :: for_bounded_files (
134+ [ ( sessions_dir, files. as_slice ( ) ) ] ,
135+ [ ( sessions_dir, all_files. as_slice ( ) ) ] ,
136+ shared. single_thread ,
137+ )
138+ } else {
139+ CodexReplayPlan :: new ( [ ( sessions_dir, all_files. as_slice ( ) ) ] , shared. single_thread )
140+ } ;
113141 let run = CodexAggregateRun {
114142 sessions_dir,
115143 files : & files,
@@ -678,7 +706,7 @@ mod tests {
678706
679707 use crate :: {
680708 CodexModelUsage , PricingMap , cli:: CodexSpeed , model_aliases:: set_model_aliases_for_tests,
681- paths:: CodexUsageSource ,
709+ paths:: CodexUsageSource , replay ,
682710 } ;
683711
684712 #[ test]
@@ -747,6 +775,153 @@ mod tests {
747775 ) ;
748776 }
749777
778+ #[ test]
779+ fn skips_historical_files_before_parsing_but_keeps_long_running_sessions ( ) {
780+ let usage_line = |timestamp : & str , input_tokens : u64 | {
781+ json ! ( {
782+ "timestamp" : timestamp,
783+ "type" : "event_msg" ,
784+ "payload" : {
785+ "type" : "token_count" ,
786+ "info" : {
787+ "model" : "gpt-5" ,
788+ "last_token_usage" : {
789+ "input_tokens" : input_tokens,
790+ "cached_input_tokens" : 0 ,
791+ "output_tokens" : 1 ,
792+ "total_tokens" : input_tokens + 1 ,
793+ } ,
794+ } ,
795+ } ,
796+ } )
797+ . to_string ( )
798+ } ;
799+ let historical = [
800+ json ! ( {
801+ "timestamp" : "2025-01-01T08:00:00.000Z" ,
802+ "type" : "session_meta" ,
803+ "payload" : { "id" : "historical" } ,
804+ } )
805+ . to_string ( ) ,
806+ usage_line ( "2026-03-15T08:01:00.000Z" , 999 ) ,
807+ ]
808+ . join ( "\n " ) ;
809+ let long_running = [
810+ json ! ( {
811+ "timestamp" : "2026-01-01T08:00:00.000Z" ,
812+ "type" : "session_meta" ,
813+ "payload" : { "id" : "long-running" } ,
814+ } )
815+ . to_string ( ) ,
816+ usage_line ( "2026-01-01T08:01:00.000Z" , 10 ) ,
817+ usage_line ( "2026-03-15T08:01:00.000Z" , 100 ) ,
818+ ]
819+ . join ( "\n " ) ;
820+ let fixture = fs_fixture ! ( {
821+ "sessions/2025/01/01/historical.jsonl" : & historical,
822+ "sessions/2026/01/01/long-running.jsonl" : & long_running,
823+ } ) ;
824+ let historical_path = fixture. path ( "sessions/2025/01/01/historical.jsonl" ) ;
825+ let long_running_path = fixture. path ( "sessions/2026/01/01/long-running.jsonl" ) ;
826+ crate :: paths:: set_file_modified (
827+ & historical_path,
828+ parse_ts_timestamp ( "2025-01-01T08:01:00.000Z" ) . unwrap ( ) ,
829+ ) ;
830+ crate :: paths:: set_file_modified (
831+ & long_running_path,
832+ parse_ts_timestamp ( "2026-03-15T08:01:00.000Z" ) . unwrap ( ) ,
833+ ) ;
834+ let shared = SharedArgs {
835+ since : Some ( "20260315" . to_string ( ) ) ,
836+ until : Some ( "20260315" . to_string ( ) ) ,
837+ timezone : Some ( "UTC" . to_string ( ) ) ,
838+ ..SharedArgs :: default ( )
839+ } ;
840+
841+ for single_thread in [ true , false ] {
842+ let _ = replay:: take_observed_file_reads ( ) ;
843+ let bounded = load_groups_from_directory (
844+ & fixture. path ( "sessions" ) ,
845+ & SharedArgs {
846+ single_thread,
847+ ..shared. clone ( )
848+ } ,
849+ AgentReportKind :: Daily ,
850+ )
851+ . unwrap ( ) ;
852+ let group = & bounded[ "2026-03-15" ] ;
853+ assert_eq ! ( group. input_tokens, 100 ) ;
854+ assert_eq ! ( group. output_tokens, 1 ) ;
855+ assert_eq ! ( group. total_tokens, 101 ) ;
856+ if single_thread {
857+ let reads = replay:: take_observed_file_reads ( ) ;
858+ assert ! ( reads. contains( & long_running_path) ) ;
859+ assert ! ( !reads. contains( & historical_path) ) ;
860+ }
861+
862+ let unbounded = load_groups_from_directory (
863+ & fixture. path ( "sessions" ) ,
864+ & SharedArgs {
865+ single_thread,
866+ ..SharedArgs :: default ( )
867+ } ,
868+ AgentReportKind :: Daily ,
869+ )
870+ . unwrap ( ) ;
871+ assert_eq ! ( unbounded[ "2026-03-15" ] . input_tokens, 1_099 ) ;
872+ }
873+ }
874+
875+ #[ test]
876+ fn retains_next_utc_path_day_for_local_until_boundary ( ) {
877+ let late_utc = [
878+ json ! ( {
879+ "timestamp" : "2026-03-16T06:29:00.000Z" ,
880+ "type" : "session_meta" ,
881+ "payload" : { "id" : "late-utc" } ,
882+ } )
883+ . to_string ( ) ,
884+ json ! ( {
885+ "timestamp" : "2026-03-16T06:30:00.000Z" ,
886+ "type" : "event_msg" ,
887+ "payload" : {
888+ "type" : "token_count" ,
889+ "info" : {
890+ "model" : "gpt-5" ,
891+ "last_token_usage" : {
892+ "input_tokens" : 100 ,
893+ "cached_input_tokens" : 0 ,
894+ "output_tokens" : 1 ,
895+ "total_tokens" : 101 ,
896+ } ,
897+ } ,
898+ } ,
899+ } )
900+ . to_string ( ) ,
901+ ]
902+ . join ( "\n " ) ;
903+ let fixture = fs_fixture ! ( {
904+ "sessions/2026/03/16/late-utc.jsonl" : & late_utc,
905+ } ) ;
906+ let late_utc_path = fixture. path ( "sessions/2026/03/16/late-utc.jsonl" ) ;
907+ let shared = SharedArgs {
908+ since : Some ( "20260315" . to_string ( ) ) ,
909+ until : Some ( "20260315" . to_string ( ) ) ,
910+ timezone : Some ( "America/Los_Angeles" . to_string ( ) ) ,
911+ single_thread : true ,
912+ ..SharedArgs :: default ( )
913+ } ;
914+
915+ let _ = replay:: take_observed_file_reads ( ) ;
916+ let groups =
917+ load_groups_from_directory ( & fixture. path ( "sessions" ) , & shared, AgentReportKind :: Daily )
918+ . unwrap ( ) ;
919+
920+ assert_eq ! ( groups[ "2026-03-15" ] . input_tokens, 100 ) ;
921+ assert_eq ! ( groups[ "2026-03-15" ] . total_tokens, 101 ) ;
922+ assert ! ( replay:: take_observed_file_reads( ) . contains( & late_utc_path) ) ;
923+ }
924+
750925 #[ test]
751926 fn dedupes_copied_token_usage_across_session_files ( ) {
752927 let usage_line = json ! ( {
0 commit comments