noetl-server 2.19.4

NoETL Control Plane - Async Rust server for workflow orchestration
Documentation
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
//! Execution state reconstruction from events.
//!
//! Provides state reconstruction for event-sourced workflow execution.

use std::collections::HashMap;
use std::time::Instant;

use chrono::{DateTime, Utc};
use serde::{Deserialize, Serialize};

use crate::db::models::Event;

/// High-level execution state.
#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
#[serde(rename_all = "snake_case")]
pub enum ExecutionState {
    /// Execution has not started yet.
    Initial,
    /// Execution is in progress.
    InProgress,
    /// Execution completed successfully.
    Completed,
    /// Execution failed.
    Failed,
    /// Execution was cancelled.
    Cancelled,
}

impl std::fmt::Display for ExecutionState {
    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
        match self {
            Self::Initial => write!(f, "initial"),
            Self::InProgress => write!(f, "in_progress"),
            Self::Completed => write!(f, "completed"),
            Self::Failed => write!(f, "failed"),
            Self::Cancelled => write!(f, "cancelled"),
        }
    }
}

impl From<&str> for ExecutionState {
    fn from(s: &str) -> Self {
        match s.to_lowercase().as_str() {
            "initial" | "pending" => Self::Initial,
            "in_progress" | "running" => Self::InProgress,
            "completed" | "success" => Self::Completed,
            "failed" | "error" => Self::Failed,
            "cancelled" | "canceled" => Self::Cancelled,
            _ => Self::Initial,
        }
    }
}

/// State of a single workflow step.
#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
#[serde(rename_all = "snake_case")]
pub enum StepState {
    /// Step has not been entered yet.
    Pending,
    /// Step has been entered (step.enter).
    Entered,
    /// Command has been issued.
    CommandIssued,
    /// Command has been claimed by a worker.
    CommandClaimed,
    /// Command execution has started.
    CommandStarted,
    /// Step completed successfully.
    Completed,
    /// Step failed.
    Failed,
    /// Step was skipped.
    Skipped,
}

impl std::fmt::Display for StepState {
    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
        match self {
            Self::Pending => write!(f, "pending"),
            Self::Entered => write!(f, "entered"),
            Self::CommandIssued => write!(f, "command_issued"),
            Self::CommandClaimed => write!(f, "command_claimed"),
            Self::CommandStarted => write!(f, "command_started"),
            Self::Completed => write!(f, "completed"),
            Self::Failed => write!(f, "failed"),
            Self::Skipped => write!(f, "skipped"),
        }
    }
}

/// Step information including state and result.
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct StepInfo {
    pub name: String,
    pub state: StepState,
    #[serde(skip_serializing_if = "Option::is_none")]
    pub result: Option<serde_json::Value>,
    #[serde(skip_serializing_if = "Option::is_none")]
    pub error: Option<String>,
    #[serde(skip_serializing_if = "Option::is_none")]
    pub entered_at: Option<DateTime<Utc>>,
    #[serde(skip_serializing_if = "Option::is_none")]
    pub completed_at: Option<DateTime<Utc>>,
    pub attempt: i32,

    // -------- Iterator fan-out (Phase D R3b) --------
    //
    // A step with `step.loop` fans out into N iteration commands at
    // dispatch time.  The orchestrator emits ONE `step.enter` (which
    // records `iterations_expected` here) and N `command.issued`
    // events, each with a per-iteration `command_id` of the shape
    // `<exec>:<step>:<event>:i<index>` and `iteration_index` in
    // meta.  Workers that act on those commands echo `command_id`
    // forward in their emitted events but do NOT necessarily echo
    // `iteration_index` (worker contract is per-command, not
    // per-iteration), so `apply_event` deduplicates `command.completed`
    // events by `command_id` instead of by iteration_index.  The
    // step's `state` flips to `Completed` once we've seen
    // `iterations_expected` distinct command_ids complete.
    //
    // Non-looped steps leave these at their defaults and behave
    // exactly as before.
    /// Total iterations expected when the step is a `step.loop` step.
    /// `None` for non-looped steps; set from the `step.enter` event
    /// context at fan-out time.
    #[serde(default, skip_serializing_if = "Option::is_none")]
    pub iterations_expected: Option<i32>,

    /// Distinct iteration command_ids observed as `command.completed`
    /// (dedup so a dual-worker race emitting two `command.completed`
    /// for the same command_id only counts once).  Always empty for
    /// non-looped steps.
    #[serde(default, skip_serializing_if = "std::collections::HashSet::is_empty")]
    pub iteration_command_ids: std::collections::HashSet<String>,

    /// Per-iteration result payloads collected in dispatch order.
    /// Used to assemble the aggregate result the next step sees in
    /// its render context.  Empty for non-looped steps.
    #[serde(default, skip_serializing_if = "Vec::is_empty")]
    pub iteration_results: Vec<serde_json::Value>,
}

impl StepInfo {
    /// Create a new step info in pending state.
    pub fn new(name: &str) -> Self {
        Self {
            name: name.to_string(),
            state: StepState::Pending,
            result: None,
            error: None,
            entered_at: None,
            completed_at: None,
            attempt: 0,
            iterations_expected: None,
            iteration_command_ids: std::collections::HashSet::new(),
            iteration_results: Vec::new(),
        }
    }

    /// True if this step was dispatched as a `step.loop` fan-out.
    pub fn is_iterator(&self) -> bool {
        self.iterations_expected.is_some()
    }

    /// Number of distinct iterations that have completed.  Always
    /// `0` for non-iterator steps.
    pub fn iterations_completed(&self) -> i32 {
        self.iteration_command_ids.len() as i32
    }
}

/// Complete workflow state reconstructed from events.
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct WorkflowState {
    pub execution_id: i64,
    pub catalog_id: i64,
    pub state: ExecutionState,
    pub steps: HashMap<String, StepInfo>,
    #[serde(skip_serializing_if = "Option::is_none")]
    pub workload: Option<serde_json::Value>,
    #[serde(skip_serializing_if = "Option::is_none")]
    pub path: Option<String>,
    #[serde(skip_serializing_if = "Option::is_none")]
    pub version: Option<String>,
    #[serde(skip_serializing_if = "Option::is_none")]
    pub started_at: Option<DateTime<Utc>>,
    #[serde(skip_serializing_if = "Option::is_none")]
    pub completed_at: Option<DateTime<Utc>>,
    #[serde(skip_serializing_if = "Option::is_none")]
    pub parent_execution_id: Option<i64>,
}

/// Pull a `command_id` out of an event row.  Workers echo
/// `command_id` forward through their emitted events; depending on
/// the lifecycle slot it may live on `meta`, on `result.context`
/// (constraint-compliant envelope), or — in older shapes — on
/// `result.data`.  Returns the first match.  Used by R3b iterator
/// state aggregation to deduplicate `command.completed` events.
fn extract_command_id(event: &Event) -> Option<String> {
    if let Some(meta) = &event.meta {
        if let Some(s) = meta.get("command_id").and_then(|v| v.as_str()) {
            return Some(s.to_string());
        }
    }
    if let Some(result) = &event.result {
        if let Some(s) = result
            .get("context")
            .and_then(|c| c.get("command_id"))
            .and_then(|v| v.as_str())
        {
            return Some(s.to_string());
        }
        if let Some(s) = result
            .get("data")
            .and_then(|d| d.get("command_id"))
            .and_then(|v| v.as_str())
        {
            return Some(s.to_string());
        }
    }
    None
}

impl WorkflowState {
    /// Create a new workflow state.
    pub fn new(execution_id: i64, catalog_id: i64) -> Self {
        Self {
            execution_id,
            catalog_id,
            state: ExecutionState::Initial,
            steps: HashMap::new(),
            workload: None,
            path: None,
            version: None,
            started_at: None,
            completed_at: None,
            parent_execution_id: None,
        }
    }

    /// Reconstruct workflow state from a list of events.
    pub fn from_events(events: &[Event]) -> Option<Self> {
        let start = Instant::now();

        if events.is_empty() {
            return None;
        }

        // Get execution_id and catalog_id from first event
        let first = &events[0];
        let mut state = Self::new(first.execution_id, first.catalog_id);

        // Process events in order
        for event in events {
            state.apply_event(event);
        }

        let duration = start.elapsed();
        let event_count = events.len();

        // Log performance metrics for state reconstruction
        tracing::info!(
            target: "noetl.performance",
            execution_id = %first.execution_id,
            phase = "state_reconstruction",
            event_count = %event_count,
            step_count = %state.steps.len(),
            duration_ms = %duration.as_millis(),
            "State reconstructed from events"
        );

        // Warn if reconstruction is slow (potential bottleneck)
        if duration.as_millis() > 100 || event_count > 50 {
            tracing::warn!(
                target: "noetl.performance",
                execution_id = %first.execution_id,
                event_count = %event_count,
                duration_ms = %duration.as_millis(),
                "Slow state reconstruction detected - consider optimizing event loading"
            );
        }

        Some(state)
    }

    /// Apply a single event to update the workflow state.
    pub fn apply_event(&mut self, event: &Event) {
        match event.event_type.as_str() {
            "playbook_started" => {
                self.state = ExecutionState::InProgress;
                self.started_at = Some(event.created_at);
                self.parent_execution_id = event.parent_execution_id;

                // Extract workload from context
                if let Some(context) = &event.context {
                    if let Some(workload) = context.get("workload") {
                        self.workload = Some(workload.clone());
                    }
                    if let Some(path) = context.get("path").and_then(|v| v.as_str()) {
                        self.path = Some(path.to_string());
                    }
                    if let Some(version) = context.get("version").and_then(|v| v.as_str()) {
                        self.version = Some(version.to_string());
                    }
                }
            }
            "playbook_completed" | "playbook.completed" => {
                self.state = ExecutionState::Completed;
                self.completed_at = Some(event.created_at);
            }
            "playbook_failed" | "playbook.failed" => {
                self.state = ExecutionState::Failed;
                self.completed_at = Some(event.created_at);
            }
            "playbook.cancelled" => {
                self.state = ExecutionState::Cancelled;
                self.completed_at = Some(event.created_at);
            }
            "step.enter" | "step_enter" | "step_started" => {
                if let Some(name) = &event.node_name {
                    let step = self
                        .steps
                        .entry(name.clone())
                        .or_insert_with(|| StepInfo::new(name));
                    step.state = StepState::Entered;
                    step.entered_at = Some(event.created_at);
                    // R3b iterator fan-out: orchestrator stamps the
                    // iteration total onto the step.enter event so
                    // state reconstruction knows how many
                    // command.completed events to wait for before
                    // marking the step truly Completed.  The
                    // orchestrator emits `EventToEmit { context:
                    // Some(...) }`, which `trigger_orchestrator`
                    // persists by wrapping it inside the
                    // constraint-compliant `{status, context}`
                    // result envelope (per noetl/server#29) — so
                    // the canonical storage location is
                    // `event.result.context.iterations_expected`.
                    // Older callers may have populated the event
                    // row's `context` column directly; we accept
                    // both shapes.  Workers' own per-iteration
                    // step.enter events don't carry this key and
                    // leave the previously-set value alone.
                    let total = event
                        .result
                        .as_ref()
                        .and_then(|r| r.get("context"))
                        .and_then(|c| c.get("iterations_expected"))
                        .and_then(|v| v.as_i64())
                        .or_else(|| {
                            event
                                .context
                                .as_ref()
                                .and_then(|c| c.get("iterations_expected"))
                                .and_then(|v| v.as_i64())
                        });
                    if let Some(total) = total {
                        step.iterations_expected = Some(total as i32);
                    }
                }
            }
            "command.issued" => {
                if let Some(name) = &event.node_name {
                    let step = self
                        .steps
                        .entry(name.clone())
                        .or_insert_with(|| StepInfo::new(name));
                    step.state = StepState::CommandIssued;
                }
            }
            "command.claimed" => {
                if let Some(name) = &event.node_name {
                    let step = self
                        .steps
                        .entry(name.clone())
                        .or_insert_with(|| StepInfo::new(name));
                    step.state = StepState::CommandClaimed;
                }
            }
            "command.started" | "action_started" => {
                if let Some(name) = &event.node_name {
                    let step = self
                        .steps
                        .entry(name.clone())
                        .or_insert_with(|| StepInfo::new(name));
                    step.state = StepState::CommandStarted;
                    if let Some(attempt) = event.attempt {
                        step.attempt = attempt;
                    }
                }
            }
            "command.completed" | "action_completed" | "step.exit" | "step_completed" => {
                if let Some(name) = &event.node_name {
                    let step = self
                        .steps
                        .entry(name.clone())
                        .or_insert_with(|| StepInfo::new(name));

                    // R3b iterator-aware completion: if this step is
                    // a loop step (iterations_expected set), count
                    // each distinct `command_id` (sourced from meta
                    // or result.context) toward completion.  Workers
                    // emit multiple events per command (claimed →
                    // started → call.done → completed), and a
                    // dual-worker race may even emit two
                    // `command.completed` events for the same
                    // command_id — both are deduped by the HashSet.
                    // Only flip state to Completed once we've seen
                    // `iterations_expected` distinct command_ids
                    // complete.  Non-iterator steps continue to
                    // complete on the first command.completed.
                    if let Some(expected) = step.iterations_expected {
                        let command_id = extract_command_id(event);
                        if let Some(cid) = command_id {
                            // First time we've seen this iteration?
                            // Append its result in arrival order.
                            if step.iteration_command_ids.insert(cid) {
                                if let Some(result) = event.result.clone() {
                                    step.iteration_results.push(result);
                                }
                            }
                        }
                        if step.iterations_completed() >= expected {
                            step.state = StepState::Completed;
                            step.completed_at = Some(event.created_at);
                            // Aggregate result = list of per-iteration
                            // results in arrival order (may not match
                            // dispatch index in parallel mode — see
                            // R3b follow-up).
                            step.result = Some(serde_json::Value::Array(
                                step.iteration_results.clone(),
                            ));
                        }
                        // Mid-iteration: leave step.state at whatever
                        // command.started / command.claimed last set
                        // it to so `is_step_completed` returns false.
                    } else {
                        // Plain (non-iterator) step.
                        step.state = StepState::Completed;
                        step.completed_at = Some(event.created_at);
                        // Only overwrite step.result with command.completed's
                        // envelope if the user-data hasn't been written yet
                        // (e.g. by an earlier `call.done`).  command.completed
                        // carries only `{status, command_id}`, no data — so
                        // overwriting would lose the rich payload that
                        // next.arcs / step.when need.  See noetl/ai-meta#60
                        // for the orchestrator-template gap that surfaced
                        // this.
                        if step.result.is_none() {
                            step.result = event.result.clone();
                        }
                    }
                }
            }
            "call.done" | "action_done" => {
                // The worker emits `call.done` between
                // `command.started` and `command.completed` to carry
                // the user-code result.  Capture step.result here so
                // the orchestrator's template context (built via
                // `build_context`) can expose `{{ step_name.field }}`
                // for next.arcs / step.when evaluation.
                //
                // The state stays at CommandStarted — `command.completed`
                // (above) flips to Completed.  This event's purpose
                // here is data attachment only.
                if let Some(name) = &event.node_name {
                    let step = self
                        .steps
                        .entry(name.clone())
                        .or_insert_with(|| StepInfo::new(name));
                    // For iterator steps the iteration-aware branch
                    // in command.completed builds the per-iteration
                    // result array; leave it alone here.  Plain steps
                    // get their data attached.
                    if step.iterations_expected.is_none() {
                        if let Some(result) = event.result.clone() {
                            step.result = Some(result);
                        }
                    }
                }
            }
            "command.failed" | "action_failed" | "step_failed" => {
                if let Some(name) = &event.node_name {
                    let step = self
                        .steps
                        .entry(name.clone())
                        .or_insert_with(|| StepInfo::new(name));
                    step.state = StepState::Failed;
                    step.completed_at = Some(event.created_at);
                    // Extract error from result.  Two shapes seen in
                    // the wild — top-level `result.error` and the
                    // nested `result.context.error` (the worker's
                    // standard envelope wraps the tool's
                    // `{status, error, ...}` output under
                    // `result.context`).  Try the top-level form
                    // first, then fall back to the nested form so
                    // step.error gets populated regardless of which
                    // tool emitted the failure.  See
                    // noetl/ai-meta#58 for the orchestrator-side
                    // failure-termination fix that depends on this.
                    if let Some(result) = &event.result {
                        let err_value = result
                            .get("error")
                            .and_then(|v| v.as_str())
                            .or_else(|| {
                                result
                                    .get("context")
                                    .and_then(|c| c.get("error"))
                                    .and_then(|v| v.as_str())
                            });
                        if let Some(error) = err_value {
                            step.error = Some(error.to_string());
                        }
                    }
                }
            }
            _ => {}
        }
    }

    /// Get the result for a specific step.
    pub fn get_step_result(&self, step_name: &str) -> Option<&serde_json::Value> {
        self.steps.get(step_name).and_then(|s| s.result.as_ref())
    }

    /// Get all step results as a map.
    pub fn get_all_results(&self) -> HashMap<String, serde_json::Value> {
        self.steps
            .iter()
            .filter_map(|(name, info)| info.result.clone().map(|r| (name.clone(), r)))
            .collect()
    }

    /// Check if a step has completed (successfully or with failure).
    pub fn is_step_done(&self, step_name: &str) -> bool {
        self.steps
            .get(step_name)
            .map(|s| {
                matches!(
                    s.state,
                    StepState::Completed | StepState::Failed | StepState::Skipped
                )
            })
            .unwrap_or(false)
    }

    /// Check if a step completed successfully.
    pub fn is_step_completed(&self, step_name: &str) -> bool {
        self.steps
            .get(step_name)
            .map(|s| matches!(s.state, StepState::Completed))
            .unwrap_or(false)
    }

    /// Check if a step failed.
    pub fn is_step_failed(&self, step_name: &str) -> bool {
        self.steps
            .get(step_name)
            .map(|s| matches!(s.state, StepState::Failed))
            .unwrap_or(false)
    }

    /// Get the names of all completed steps.
    pub fn completed_steps(&self) -> Vec<&str> {
        self.steps
            .iter()
            .filter(|(_, info)| matches!(info.state, StepState::Completed))
            .map(|(name, _)| name.as_str())
            .collect()
    }

    /// Get the names of all running steps.
    pub fn running_steps(&self) -> Vec<&str> {
        self.steps
            .iter()
            .filter(|(_, info)| {
                matches!(
                    info.state,
                    StepState::Entered
                        | StepState::CommandIssued
                        | StepState::CommandClaimed
                        | StepState::CommandStarted
                )
            })
            .map(|(name, _)| name.as_str())
            .collect()
    }

    /// Check if there are any running steps.
    pub fn has_running_steps(&self) -> bool {
        !self.running_steps().is_empty()
    }

    /// Build a context map for template rendering.
    pub fn build_context(&self) -> serde_json::Value {
        let mut context = serde_json::Map::new();

        // Add workload variables.  Each key is exposed both at the
        // top level (so `{{ skip_middle }}` works) AND under the
        // `workload` namespace (so `{{ workload.skip_middle }}`
        // works) — matches the Python reference shape and the
        // generate_initial_commands path in handlers/execute.rs.
        // Without the `workload` namespace, step.when expressions
        // that reference `workload.X` raise an undefined-value
        // template error during transition evaluation.
        if let Some(serde_json::Value::Object(wl)) = &self.workload {
            for (k, v) in wl {
                context.insert(k.clone(), v.clone());
            }
            context.insert(
                "workload".to_string(),
                serde_json::Value::Object(wl.clone()),
            );
        }

        // Add step results under TWO shapes, matching the Python
        // reference and the canonical v10 playbook YAML:
        //
        // - `steps.<name>` carries the FULL envelope as-stored
        //   (back-compat for `{{ steps.eval_flag.status }}`-style
        //   references and admin tooling that wants the wrapper
        //   metadata).
        // - `<name>` at the TOP level carries the UNWRAPPED user
        //   data — the dict the tool's user code assigned to
        //   `result = {...}`.  This is the shape next.arcs /
        //   step.when guards read via `{{ eval_flag.is_hot }}`
        //   (no `steps.` / no `.data.` prefix needed).
        //
        // The envelope shape stored on `info.result` after wrapping
        // by `apply_event` is:
        //   { status, context: { result: { status, context: {
        //       data: <USER_DATA>, status, stdout, stderr, ... } } } }
        // — `extract_user_data` walks the envelope and returns the
        // inner `data` value.  See noetl/ai-meta#60 for the e2e
        // finding that surfaced this orchestrator template gap.
        let mut steps = serde_json::Map::new();
        for (name, info) in &self.steps {
            if let Some(result) = &info.result {
                steps.insert(name.clone(), result.clone());
                if let Some(user_data) = extract_user_data(result) {
                    context.insert(name.clone(), user_data);
                }
            }
        }
        context.insert("steps".to_string(), serde_json::Value::Object(steps));

        // Add execution metadata
        context.insert(
            "execution_id".to_string(),
            serde_json::json!(self.execution_id.to_string()),
        );
        context.insert(
            "catalog_id".to_string(),
            serde_json::json!(self.catalog_id.to_string()),
        );

        if let Some(path) = &self.path {
            context.insert("path".to_string(), serde_json::json!(path));
        }
        if let Some(version) = &self.version {
            context.insert("version".to_string(), serde_json::json!(version));
        }

        serde_json::Value::Object(context)
    }
}

/// Unwrap a step result envelope to the inner user data dict.
///
/// The wrap layers come from `apply_event`'s standard envelope:
///
/// ```text
/// outer = {
///   status: "COMPLETED",
///   context: {
///     result: {
///       status: "success",
///       context: {
///         data: <USER_DATA>,
///         status, stdout, stderr, ...
///       }
///     },
///     ...
///   }
/// }
/// ```
///
/// Returns the inner `data` value when the wrapper shape matches.
/// Falls back to the outer value (or any partially-unwrapped layer)
/// when the wrapper is absent — handles tooling that emitted a
/// flat result without going through the worker's envelope path.
/// Returns None only when the input is JSON null.
///
/// Tracks noetl/ai-meta#60 — without this unwrap, v10 playbooks
/// that reference `{{ step_name.field }}` in next.arcs / step.when
/// see an undefined value because the envelope's `status` /
/// `context` keys swallowed the user fields.
fn extract_user_data(result: &serde_json::Value) -> Option<serde_json::Value> {
    if result.is_null() {
        return None;
    }
    // Try outer.context.result.context.data — the standard
    // wrapper shape.  Each step is optional so a partial
    // unwrap still yields a useful value for back-compat.
    let inner = result
        .get("context")
        .and_then(|v| v.get("result"))
        .and_then(|v| v.get("context"))
        .and_then(|v| v.get("data"));
    if let Some(data) = inner {
        return Some(data.clone());
    }
    // Single-layer wrappers (e.g. {status, context}).
    if let Some(ctx) = result.get("context") {
        if let Some(data) = ctx.get("data") {
            return Some(data.clone());
        }
        return Some(ctx.clone());
    }
    Some(result.clone())
}

#[cfg(test)]
mod tests {
    use super::*;

    fn make_event(event_type: &str, node_name: Option<&str>) -> Event {
        Event {
            id: 1,
            execution_id: 12345,
            catalog_id: 67890,
            event_id: 1,
            parent_event_id: None,
            parent_execution_id: None,
            event_type: event_type.to_string(),
            node_id: None,
            node_name: node_name.map(|s| s.to_string()),
            node_type: None,
            status: "".to_string(),
            context: None,
            meta: None,
            result: None,
            worker_id: None,
            attempt: None,
            created_at: Utc::now(),
        }
    }

    #[test]
    fn test_execution_state_display() {
        assert_eq!(ExecutionState::Initial.to_string(), "initial");
        assert_eq!(ExecutionState::InProgress.to_string(), "in_progress");
        assert_eq!(ExecutionState::Completed.to_string(), "completed");
    }

    #[test]
    fn test_execution_state_from_str() {
        assert_eq!(ExecutionState::from("initial"), ExecutionState::Initial);
        assert_eq!(ExecutionState::from("RUNNING"), ExecutionState::InProgress);
        assert_eq!(ExecutionState::from("completed"), ExecutionState::Completed);
        assert_eq!(ExecutionState::from("FAILED"), ExecutionState::Failed);
    }

    #[test]
    fn test_workflow_state_from_events() {
        let events = vec![
            {
                let mut e = make_event("playbook_started", None);
                e.context = Some(serde_json::json!({
                    "workload": {"key": "value"},
                    "path": "test/playbook",
                    "version": "1"
                }));
                e
            },
            make_event("step.enter", Some("step1")),
            make_event("command.issued", Some("step1")),
            {
                let mut e = make_event("command.completed", Some("step1"));
                e.result = Some(serde_json::json!({"output": "success"}));
                e
            },
        ];

        let state = WorkflowState::from_events(&events).unwrap();
        assert_eq!(state.execution_id, 12345);
        assert_eq!(state.state, ExecutionState::InProgress);
        assert!(state.is_step_completed("step1"));
        assert_eq!(
            state.get_step_result("step1"),
            Some(&serde_json::json!({"output": "success"}))
        );
    }

    #[test]
    fn test_workflow_state_build_context() {
        let mut state = WorkflowState::new(12345, 67890);
        state.workload = Some(serde_json::json!({"var1": "value1"}));
        state.path = Some("test/path".to_string());

        let mut step_info = StepInfo::new("step1");
        step_info.result = Some(serde_json::json!({"output": "result1"}));
        state.steps.insert("step1".to_string(), step_info);

        let context = state.build_context();
        assert_eq!(context.get("var1").and_then(|v| v.as_str()), Some("value1"));
        assert_eq!(
            context.get("path").and_then(|v| v.as_str()),
            Some("test/path")
        );
        assert!(context.get("steps").is_some());
    }

    #[test]
    fn test_step_state_transitions() {
        let mut state = WorkflowState::new(1, 1);

        state.apply_event(&make_event("step.enter", Some("step1")));
        assert_eq!(state.steps.get("step1").unwrap().state, StepState::Entered);

        state.apply_event(&make_event("command.issued", Some("step1")));
        assert_eq!(
            state.steps.get("step1").unwrap().state,
            StepState::CommandIssued
        );

        state.apply_event(&make_event("command.completed", Some("step1")));
        assert_eq!(
            state.steps.get("step1").unwrap().state,
            StepState::Completed
        );
    }

    #[test]
    fn test_iterator_step_aggregates_completion() {
        // Simulate the events an iterator step produces:
        //   step.enter (iterations_expected=3)
        //   command.completed (iteration_index=0)
        //   command.completed (iteration_index=1)
        //   command.completed (iteration_index=2)
        //
        // The step's state stays "not completed" until all 3
        // iterations land, then flips to Completed with an
        // aggregated array result.
        let mut state = WorkflowState::new(1, 1);

        let mut enter = make_event("step.enter", Some("looped"));
        enter.context = Some(serde_json::json!({
            "iterations_expected": 3,
            "iterator_var": "item",
        }));
        state.apply_event(&enter);
        let after_enter = state.steps.get("looped").unwrap();
        assert_eq!(after_enter.state, StepState::Entered);
        assert_eq!(after_enter.iterations_expected, Some(3));
        assert_eq!(after_enter.iterations_completed(), 0);

        for (idx, payload) in [(0, "a"), (1, "b"), (2, "c")] {
            let mut ev = make_event("command.completed", Some("looped"));
            ev.meta = Some(serde_json::json!({
                "command_id": format!("e:looped:0:i{}", idx),
                "iteration_index": idx,
                "iteration_total": 3,
            }));
            ev.result = Some(serde_json::json!({ "value": payload }));
            state.apply_event(&ev);
        }

        let info = state.steps.get("looped").unwrap();
        assert_eq!(info.state, StepState::Completed);
        assert_eq!(info.iterations_completed(), 3);
        // Aggregate result is the per-iteration array in arrival order.
        let agg = info.result.as_ref().unwrap();
        assert_eq!(agg.as_array().map(|a| a.len()), Some(3));
        let values: Vec<String> = agg
            .as_array()
            .unwrap()
            .iter()
            .map(|v| v.get("value").unwrap().as_str().unwrap().to_string())
            .collect();
        assert_eq!(values, vec!["a", "b", "c"]);
    }

    #[test]
    fn test_iterator_step_dedupes_duplicate_command_completed() {
        // Two `command.completed` events for the same command_id
        // (dual-worker race) should NOT double-count the iteration.
        let mut state = WorkflowState::new(1, 1);
        let mut enter = make_event("step.enter", Some("looped"));
        enter.context = Some(serde_json::json!({
            "iterations_expected": 2,
        }));
        state.apply_event(&enter);

        for _ in 0..2 {
            let mut ev = make_event("command.completed", Some("looped"));
            ev.meta = Some(serde_json::json!({
                "command_id": "e:looped:0:i0",
            }));
            ev.result = Some(serde_json::json!({"i": 0}));
            state.apply_event(&ev);
        }
        // Only 1 distinct command_id seen.
        let info = state.steps.get("looped").unwrap();
        assert_eq!(info.iterations_completed(), 1);
        assert_ne!(info.state, StepState::Completed);

        // Now the second iteration's command_id completes.
        let mut ev = make_event("command.completed", Some("looped"));
        ev.meta = Some(serde_json::json!({
            "command_id": "e:looped:0:i1",
        }));
        ev.result = Some(serde_json::json!({"i": 1}));
        state.apply_event(&ev);
        let info = state.steps.get("looped").unwrap();
        assert_eq!(info.iterations_completed(), 2);
        assert_eq!(info.state, StepState::Completed);
    }

    #[test]
    fn test_iterator_step_partial_completion_stays_running() {
        // Two of three iterations done — step should NOT be
        // Completed yet (state is whatever the last event left it
        // at, but `is_step_completed` returns false).
        let mut state = WorkflowState::new(1, 1);

        let mut enter = make_event("step.enter", Some("looped"));
        enter.context = Some(serde_json::json!({
            "iterations_expected": 3,
        }));
        state.apply_event(&enter);

        for idx in 0..2 {
            let mut ev = make_event("command.completed", Some("looped"));
            ev.meta = Some(serde_json::json!({
                "command_id": format!("e:looped:0:i{}", idx),
            }));
            ev.result = Some(serde_json::json!({"i": idx}));
            state.apply_event(&ev);
        }

        let info = state.steps.get("looped").unwrap();
        assert_ne!(info.state, StepState::Completed);
        assert_eq!(info.iterations_completed(), 2);
        assert!(!state.is_step_completed("looped"));
    }

    #[test]
    fn test_iterator_partial_with_worker_step_exit_does_not_complete() {
        // Reproduces the R3b kind-val symptom: orchestrator emits
        // step.enter(iterations_expected=3), 3 command.issued events
        // fire, then ONE iteration's worker lifecycle arrives
        // (command.claimed/started + worker's step.enter + call.done
        // + step.exit + command.completed).  The looped step must
        // NOT be marked Completed after only 1 iteration, even
        // though step.exit AND command.completed both go through
        // the iteration-aware match arm and both carry the same
        // command_id in result.context.
        let mut state = WorkflowState::new(1, 1);

        // 1. Orchestrator's initial step.enter — populates
        //    iterations_expected.  In production the orchestrator
        //    persists this via `trigger_orchestrator`, which wraps
        //    `EventToEmit.context` in a `{status, context}` result
        //    envelope (per noetl/server#29's chk_event_result_shape
        //    constraint).  So the canonical storage location is
        //    `event.result.context.iterations_expected`, NOT the
        //    event row's `context` column.  Earlier tests used the
        //    `event.context` shape; we accept both via the
        //    apply_event fallback, so this test uses the
        //    production shape.
        let mut enter = make_event("step.enter", Some("looped"));
        enter.result = Some(serde_json::json!({
            "status": "ENTERED",
            "context": {
                "iterations_expected": 3,
                "iterator_var": "item",
            },
        }));
        state.apply_event(&enter);

        // 2. Three command.issued events — each with a distinct
        //    per-iteration command_id in meta.
        for idx in 0..3 {
            let mut ev = make_event("command.issued", Some("looped"));
            ev.meta = Some(serde_json::json!({
                "command_id": format!("exec:looped:e0:i{}", idx),
                "iteration_index": idx,
                "iteration_total": 3,
            }));
            state.apply_event(&ev);
        }

        // 3. One iteration's worker lifecycle (only iter i2).
        let cid = "exec:looped:e0:i2".to_string();
        let mut claimed = make_event("command.claimed", Some("looped"));
        claimed.meta = Some(serde_json::json!({"command_id": cid}));
        state.apply_event(&claimed);

        let mut started = make_event("command.started", Some("looped"));
        started.meta = Some(serde_json::json!({"command_id": cid}));
        state.apply_event(&started);

        // Worker's per-iteration step.enter — no iterations_expected
        // in context, so iterations_expected must stay Some(3).
        let mut worker_enter = make_event("step.enter", Some("looped"));
        worker_enter.context = Some(serde_json::json!({"status": "started"}));
        state.apply_event(&worker_enter);

        // call.done — not in any match arm, no state change.
        let call_done = make_event("call.done", Some("looped"));
        state.apply_event(&call_done);

        // step.exit — IS in the command.completed arm.  Carries
        // command_id in result.context.
        let mut step_exit = make_event("step.exit", Some("looped"));
        step_exit.result = Some(serde_json::json!({
            "status": "COMPLETED",
            "context": { "command_id": cid.clone(), "status": "COMPLETED" }
        }));
        state.apply_event(&step_exit);

        // command.completed — same command_id (dedupes via HashSet).
        let mut completed = make_event("command.completed", Some("looped"));
        completed.result = Some(serde_json::json!({
            "status": "COMPLETED",
            "context": { "command_id": cid.clone(), "worker_id": "w" }
        }));
        state.apply_event(&completed);

        let info = state.steps.get("looped").unwrap();
        assert_eq!(info.iterations_expected, Some(3));
        assert_eq!(
            info.iterations_completed(),
            1,
            "only ONE distinct command_id observed across step.exit + command.completed; \
             iteration_command_ids = {:?}",
            info.iteration_command_ids
        );
        assert_ne!(
            info.state,
            StepState::Completed,
            "looped must NOT be Completed after only 1 of 3 iterations; state = {:?}",
            info.state
        );
        assert!(!state.is_step_completed("looped"));
    }

    #[test]
    fn test_plain_step_unaffected_by_iterator_logic() {
        // A plain step (no iterations_expected) continues to
        // complete on the first command.completed, same as before.
        let mut state = WorkflowState::new(1, 1);
        state.apply_event(&make_event("step.enter", Some("plain")));
        let mut ev = make_event("command.completed", Some("plain"));
        ev.result = Some(serde_json::json!({"ok": true}));
        state.apply_event(&ev);
        let info = state.steps.get("plain").unwrap();
        assert_eq!(info.state, StepState::Completed);
        assert_eq!(info.iterations_expected, None);
        assert_eq!(info.iterations_completed(), 0);
        assert_eq!(info.result, Some(serde_json::json!({"ok": true})));
    }

    #[test]
    fn test_extract_user_data_unwraps_standard_envelope() {
        // Standard wrapper shape emitted by apply_event after
        // the worker's PythonTool result-capture (noetl/tools#17).
        // The orchestrator's template context needs the inner
        // `data` exposed so `{{ step_name.field }}` resolves.
        let envelope = serde_json::json!({
            "status": "COMPLETED",
            "context": {
                "result": {
                    "status": "success",
                    "context": {
                        "data": {"is_hot": true, "message": "hot"},
                        "status": "success",
                        "stdout": "",
                        "stderr": "",
                    },
                },
                "call_index": 0,
            },
        });
        let data = extract_user_data(&envelope).expect("unwrap should succeed");
        assert_eq!(
            data.get("is_hot").and_then(|v| v.as_bool()),
            Some(true)
        );
        assert_eq!(
            data.get("message").and_then(|v| v.as_str()),
            Some("hot")
        );
    }

    #[test]
    fn test_extract_user_data_handles_flat_result() {
        // Back-compat: a tool that emitted a flat result without
        // going through the wrapper.  No `context.result.context.data`
        // path → fall back through the partial-unwrap branches and
        // ultimately return the input.
        let flat = serde_json::json!({"is_hot": false});
        let data = extract_user_data(&flat).expect("flat result preserved");
        assert_eq!(data, flat);
    }

    #[test]
    fn test_extract_user_data_null_returns_none() {
        let null = serde_json::Value::Null;
        assert!(extract_user_data(&null).is_none());
    }

    #[test]
    fn test_build_context_exposes_step_data_at_top_level() {
        // noetl/ai-meta#60 — workflow YAML uses `{{ eval_flag.is_hot }}`
        // (no `steps.` prefix), so the build_context must expose
        // each step's unwrapped data at the top level alongside the
        // back-compat `steps.<name>` shape.
        let mut state = WorkflowState::new(1, 1);
        state.workload = Some(serde_json::json!({"temp": 30}));
        let mut info = StepInfo::new("eval_flag");
        info.result = Some(serde_json::json!({
            "status": "COMPLETED",
            "context": {
                "result": {
                    "status": "success",
                    "context": {
                        "data": {"is_hot": true, "message": "hot"},
                    },
                },
            },
        }));
        state.steps.insert("eval_flag".to_string(), info);

        let ctx = state.build_context();
        // Top-level: `eval_flag.is_hot` resolves to the user data.
        let eval_flag = ctx.get("eval_flag").expect("top-level step data exposed");
        assert_eq!(
            eval_flag.get("is_hot").and_then(|v| v.as_bool()),
            Some(true)
        );
        // Back-compat: `steps.eval_flag` still holds the full envelope.
        let steps = ctx.get("steps").expect("steps namespace present");
        assert!(
            steps.get("eval_flag").is_some(),
            "back-compat steps namespace populated"
        );
        // Workload still at top level (from earlier build_context behavior).
        assert_eq!(ctx.get("temp").and_then(|v| v.as_i64()), Some(30));
    }
}