Skip to content

Index

state

State management backends.

Classes

StateBackend

Bases: ABC

Abstract base class for state storage backends.

Implementations handle persistence of job state for resumable execution.

Attributes
supports_execution_history property 
supports_execution_history

Whether this backend persists execution history.

Backends that override record_execution should also override this to return True. Callers can check this to avoid silent data loss when a backend unexpectedly lacks execution recording.

Functions
load abstractmethod async 
load(job_id)

Load state for a job.

Parameters:

Name Type Description Default
job_id str

Unique job identifier

 required

Returns:

Type Description
CheckpointState | None

CheckpointState if found, None otherwise
Source code in src/marianne/state/base.py 
18
19
20
21
22
23
24
25
26
27
28
@abstractmethod
async def load(self, job_id: str) -> CheckpointState | None:
    """Load state for a job.

    Args:
        job_id: Unique job identifier

    Returns:
        CheckpointState if found, None otherwise
    """
    ...
save abstractmethod async 
save(state)

Save job state.

Parameters:

Name Type Description Default
state CheckpointState

Checkpoint state to persist

 required
Source code in src/marianne/state/base.py 
30
31
32
33
34
35
36
37
@abstractmethod
async def save(self, state: CheckpointState) -> None:
    """Save job state.

    Args:
        state: Checkpoint state to persist
    """
    ...
delete abstractmethod async 
delete(job_id)

Delete state for a job.

Parameters:

Name Type Description Default
job_id str

Unique job identifier

 required

Returns:

Type Description
bool

True if deleted, False if not found
Source code in src/marianne/state/base.py 
39
40
41
42
43
44
45
46
47
48
49
@abstractmethod
async def delete(self, job_id: str) -> bool:
    """Delete state for a job.

    Args:
        job_id: Unique job identifier

    Returns:
        True if deleted, False if not found
    """
    ...
list_jobs abstractmethod async 
list_jobs()

List all jobs with state.

Returns:

Type Description
list[CheckpointState]

List of all checkpoint states
Source code in src/marianne/state/base.py 
51
52
53
54
55
56
57
58
@abstractmethod
async def list_jobs(self) -> list[CheckpointState]:
    """List all jobs with state.

    Returns:
        List of all checkpoint states
    """
    ...
get_next_sheet abstractmethod async 
get_next_sheet(job_id)

Get the next sheet to process for a job.

Parameters:

Name Type Description Default
job_id str

Unique job identifier

 required

Returns:

Type Description
int | None

Next sheet number, or None if complete
Source code in src/marianne/state/base.py 
60
61
62
63
64
65
66
67
68
69
70
@abstractmethod
async def get_next_sheet(self, job_id: str) -> int | None:
    """Get the next sheet to process for a job.

    Args:
        job_id: Unique job identifier

    Returns:
        Next sheet number, or None if complete
    """
    ...
mark_sheet_status abstractmethod async 
mark_sheet_status(job_id, sheet_num, status, error_message=None)

Update status of a specific sheet.

Parameters:

Name Type Description Default
job_id str

Unique job identifier

 required
sheet_num int

Sheet number to update

 required
status SheetStatus

New status

 required
error_message str | None

Optional error message for failures

 None
Source code in src/marianne/state/base.py 
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
@abstractmethod
async def mark_sheet_status(
    self,
    job_id: str,
    sheet_num: int,
    status: SheetStatus,
    error_message: str | None = None,
) -> None:
    """Update status of a specific sheet.

    Args:
        job_id: Unique job identifier
        sheet_num: Sheet number to update
        status: New status
        error_message: Optional error message for failures
    """
    ...
close async 
close()

Release any resources held by the backend (Q018/#37).

Optional method — backends that hold persistent connections or file handles should override this. The default no-op is safe for backends that use per-operation connections (e.g. aiosqlite context managers).

Source code in src/marianne/state/base.py 
90
91
92
93
94
95
96
async def close(self) -> None:  # noqa: B027 — intentional concrete no-op default
    """Release any resources held by the backend (Q018/#37).

    Optional method — backends that hold persistent connections or file
    handles should override this. The default no-op is safe for backends
    that use per-operation connections (e.g. aiosqlite context managers).
    """
record_execution async 
record_execution(job_id, sheet_num, attempt_num, prompt=None, output=None, exit_code=None, duration_seconds=None)

Record an execution attempt in history.

Optional method -- backends that support execution recording (e.g. SQLite) override this. The default logs at WARNING level so silent data loss is traceable, then returns None.

Parameters:

Name Type Description Default
job_id str

Job identifier.

 required
sheet_num int

Sheet number.

 required
attempt_num int

Attempt number within the sheet.

 required
prompt str | None

The prompt sent to the backend.

 None
output str | None

The output received.

 None
exit_code int | None

Exit code from the execution.

 None
duration_seconds float | None

Execution duration.

 None

Returns:

Type Description
int | None

The ID of the inserted record, or None if not supported.
Source code in src/marianne/state/base.py 
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
async def record_execution(
    self,
    job_id: str,
    sheet_num: int,
    attempt_num: int,
    prompt: str | None = None,
    output: str | None = None,
    exit_code: int | None = None,
    duration_seconds: float | None = None,
) -> int | None:
    """Record an execution attempt in history.

    Optional method -- backends that support execution recording (e.g. SQLite)
    override this. The default logs at WARNING level so silent data
    loss is traceable, then returns None.

    Args:
        job_id: Job identifier.
        sheet_num: Sheet number.
        attempt_num: Attempt number within the sheet.
        prompt: The prompt sent to the backend.
        output: The output received.
        exit_code: Exit code from the execution.
        duration_seconds: Execution duration.

    Returns:
        The ID of the inserted record, or None if not supported.
    """
    _logger.warning(
        "record_execution_noop",
        extra={"job_id": job_id, SHEET_NUM_KEY: sheet_num},
    )
    return None
infer_state_from_artifacts async 
infer_state_from_artifacts(job_id, workspace, artifact_pattern)

Infer last completed sheet from artifact files.

Fallback when state file is missing - checks for output files. Based on the fallback logic in run-sheet-review.sh.

Parameters:

Name Type Description Default
job_id str

Unique job identifier

 required
workspace str

Workspace directory path

 required
artifact_pattern str

Glob pattern for sheet artifacts (e.g., "sheet*-security-report.md")

 required

Returns:

Type Description
int | None

Inferred last completed sheet number, or None
Source code in src/marianne/state/base.py 
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
async def infer_state_from_artifacts(
    self,
    job_id: str,
    workspace: str,
    artifact_pattern: str,
) -> int | None:
    """Infer last completed sheet from artifact files.

    Fallback when state file is missing - checks for output files.
    Based on the fallback logic in run-sheet-review.sh.

    Args:
        job_id: Unique job identifier
        workspace: Workspace directory path
        artifact_pattern: Glob pattern for sheet artifacts (e.g., "sheet*-security-report.md")

    Returns:
        Inferred last completed sheet number, or None
    """
    import re
    from pathlib import Path

    workspace_path = Path(workspace)
    if not workspace_path.exists():
        return None

    # Find all matching artifact files
    artifacts = list(workspace_path.glob(artifact_pattern))
    if not artifacts:
        return None

    # Extract sheet numbers and find max
    sheet_nums = []
    for artifact in artifacts:
        match = re.search(r"sheet(\d+)", artifact.name)
        if match:
            sheet_nums.append(int(match.group(1)))

    return max(sheet_nums) if sheet_nums else None

JsonStateBackend 

JsonStateBackend(state_dir)

Bases: StateBackend

JSON file-based state storage.

Stores each job's state in a separate JSON file within the state directory. File naming: {state_dir}/{job_id}.json

Initialize JSON backend.

Parameters:

Name Type Description Default
state_dir Path

Directory to store state files

 required
Source code in src/marianne/state/json_backend.py 
44
45
46
47
48
49
50
51
def __init__(self, state_dir: Path):
    """Initialize JSON backend.

    Args:
        state_dir: Directory to store state files
    """
    self.state_dir = Path(state_dir)
    self.state_dir.mkdir(parents=True, exist_ok=True)
Functions
load async 
load(job_id)

Load state from JSON file.

Automatically detects and recovers zombie jobs (RUNNING status but process dead). When a zombie is detected, the state is updated to PAUSED and saved before returning.

Source code in src/marianne/state/json_backend.py 
 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
async def load(self, job_id: str) -> CheckpointState | None:
    """Load state from JSON file.

    Automatically detects and recovers zombie jobs (RUNNING status but
    process dead). When a zombie is detected, the state is updated to
    PAUSED and saved before returning.
    """
    state_file = self._get_state_file(job_id)
    if not state_file.exists():
        _logger.debug("state_file_not_found", job_id=job_id, path=str(state_file))
        return None

    try:
        with open(state_file) as f:
            data = json.load(f)
        state = CheckpointState.model_validate(data)

        # Detect zombie state (running but process dead) without saving.
        # Callers that need persistence (e.g. runner lifecycle) save
        # explicitly — load() remains read-only for dashboard/CLI callers.
        if state.is_zombie():
            _logger.warning(
                "zombie_detected_on_load",
                job_id=job_id,
                pid=state.pid,
                status=state.status.value,
            )
            state.mark_zombie_detected(
                reason="Detected on state load - process no longer running"
            )

        _logger.debug(
            "checkpoint_loaded",
            job_id=job_id,
            status=state.status.value,
            last_completed_sheet=state.last_completed_sheet,
            total_sheets=state.total_sheets,
        )
        return state
    except (json.JSONDecodeError, ValueError) as e:
        error_type = "json_decode" if isinstance(e, json.JSONDecodeError) else "validation"
        _logger.error(
            "checkpoint_corruption_detected",
            job_id=job_id,
            path=str(state_file),
            error_type=error_type,
            error=str(e),
        )
        raise StateCorruptionError(
            job_id=job_id,
            path=str(state_file),
            error_type=error_type,
            detail=str(e),
        ) from e
save async 
save(state)

Save state to JSON file.

Source code in src/marianne/state/json_backend.py 
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
async def save(self, state: CheckpointState) -> None:
    """Save state to JSON file."""
    state.updated_at = utc_now()
    state_file = self._get_state_file(state.job_id)

    # Write atomically using temp file + rename
    temp_file = state_file.with_suffix(".json.tmp")
    with open(temp_file, "w") as f:
        json.dump(
            state.model_dump(mode="json"),
            f,
            indent=2,
            default=str,  # Handle datetime serialization
        )
        f.flush()
        os.fsync(f.fileno())
    temp_file.replace(state_file)

    _logger.info(
        "checkpoint_saved",
        job_id=state.job_id,
        status=state.status.value,
        last_completed_sheet=state.last_completed_sheet,
        total_sheets=state.total_sheets,
        path=str(state_file),
    )
delete async 
delete(job_id)

Delete state file.

Source code in src/marianne/state/json_backend.py 
141
142
143
144
145
146
147
async def delete(self, job_id: str) -> bool:
    """Delete state file."""
    state_file = self._get_state_file(job_id)
    if state_file.exists():
        state_file.unlink()
        return True
    return False
list_jobs async 
list_jobs()

List all jobs with state files.

Optimized: reads raw JSON and sorts by updated_at before doing full Pydantic validation, avoiding expensive model_validate on every file just to determine sort order.

Source code in src/marianne/state/json_backend.py 
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
async def list_jobs(self) -> list[CheckpointState]:
    """List all jobs with state files.

    Optimized: reads raw JSON and sorts by updated_at before doing full
    Pydantic validation, avoiding expensive model_validate on every file
    just to determine sort order.
    """
    # Phase 1: Read raw JSON and extract sort key (no Pydantic overhead)
    raw_entries: list[tuple[str, dict[str, Any]]] = []  # (updated_at_str, data)
    for state_file in self.state_dir.glob("*.json"):
        if state_file.suffix == ".json" and not state_file.name.endswith(".tmp"):
            try:
                with open(state_file) as f:
                    data = json.load(f)
                # Extract updated_at for sorting without full validation
                updated_at = data.get("updated_at", "")
                raw_entries.append((updated_at, data))
            except (json.JSONDecodeError, OSError) as exc:
                _logger.warning("corrupt_state_file", path=str(state_file), error=str(exc))
                continue

    # Phase 2: Sort by raw updated_at string (ISO format sorts lexicographically)
    raw_entries.sort(key=lambda e: e[0], reverse=True)

    # Phase 3: Validate sorted entries
    states: list[CheckpointState] = []
    for _, data in raw_entries:
        try:
            states.append(CheckpointState.model_validate(data))
        except ValueError as exc:
            _logger.warning("invalid_state_data", error=str(exc))
            continue
    return states
get_next_sheet async 
get_next_sheet(job_id)

Get next sheet from state.

Source code in src/marianne/state/json_backend.py 
183
184
185
186
187
188
async def get_next_sheet(self, job_id: str) -> int | None:
    """Get next sheet from state."""
    state = await self.load(job_id)
    if state is None:
        return 1  # Start from beginning if no state
    return state.get_next_sheet()
mark_sheet_status async 
mark_sheet_status(job_id, sheet_num, status, error_message=None)

Update sheet status in state.

Source code in src/marianne/state/json_backend.py 
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
async def mark_sheet_status(
    self,
    job_id: str,
    sheet_num: int,
    status: SheetStatus,
    error_message: str | None = None,
) -> None:
    """Update sheet status in state."""
    state = await self.load(job_id)
    if state is None:
        raise ValueError(f"No state found for job {job_id}")

    if status == SheetStatus.COMPLETED:
        state.mark_sheet_completed(sheet_num)
    elif status == SheetStatus.FAILED:
        state.mark_sheet_failed(sheet_num, error_message or "Unknown error")
    elif status == SheetStatus.IN_PROGRESS:
        state.mark_sheet_started(sheet_num)

    await self.save(state)

InMemoryStateBackend 

InMemoryStateBackend()

Bases: StateBackend

In-memory state backend for testing.

Tracks state changes in a dict without filesystem I/O.

Source code in src/marianne/state/memory.py 
17
18
def __init__(self) -> None:
    self.states: dict[str, CheckpointState] = {}

SQLiteStateBackend 

SQLiteStateBackend(db_path)

Bases: StateBackend

SQLite-based state storage with execution history.

Stores job state in a SQLite database with queryable tables for: - jobs: Job-level metadata and status - sheets: Per-sheet state including attempts and errors - execution_history: Detailed record of each execution attempt

Supports schema migrations for future upgrades.

Initialize SQLite backend.

Parameters:

Name Type Description Default
db_path str | Path

Path to SQLite database file

 required
Source code in src/marianne/state/sqlite_backend.py 
43
44
45
46
47
48
49
50
51
52
def __init__(self, db_path: str | Path) -> None:
    """Initialize SQLite backend.

    Args:
        db_path: Path to SQLite database file
    """
    self.db_path = Path(db_path)
    self.db_path.parent.mkdir(parents=True, exist_ok=True)
    self._initialized = False
    self._init_lock = asyncio.Lock()
Functions
load async 
load(job_id)

Load state for a job from SQLite.

Automatically detects and recovers zombie jobs (RUNNING status but process dead). When a zombie is detected, the state is updated to PAUSED and saved before returning.

Source code in src/marianne/state/sqlite_backend.py 
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
async def load(self, job_id: str) -> CheckpointState | None:
    """Load state for a job from SQLite.

    Automatically detects and recovers zombie jobs (RUNNING status but
    process dead). When a zombie is detected, the state is updated to
    PAUSED and saved before returning.
    """
    await self._ensure_initialized()

    async with self._connect() as db:
        db.row_factory = aiosqlite.Row

        # Load job record
        cursor = await db.execute(
            "SELECT * FROM jobs WHERE id = ?", (job_id,)
        )
        job_row = await cursor.fetchone()

        if not job_row:
            _logger.debug("state_not_found", job_id=job_id)
            return None

        # Load sheet records
        cursor = await db.execute(
            "SELECT * FROM sheets WHERE job_id = ? ORDER BY sheet_num",
            (job_id,),
        )
        sheet_rows = await cursor.fetchall()

        # Reconstruct CheckpointState
        sheets: dict[int, SheetState] = {}
        for row in sheet_rows:
            sheet = SheetState(
                sheet_num=row[SHEET_NUM_KEY],
                status=SheetStatus(row["status"]),
                started_at=self._str_to_datetime(row["started_at"]),
                completed_at=self._str_to_datetime(row["completed_at"]),
                attempt_count=row["attempt_count"],
                exit_code=row["exit_code"],
                error_message=row["error_message"],
                error_category=row["error_category"],
                validation_passed=self._int_to_bool(row["validation_passed"]),
                validation_details=self._json_loads(row["validation_details"]),
                completion_attempts=row["completion_attempts"],
                passed_validations=self._json_loads(row["passed_validations"]) or [],
                failed_validations=self._json_loads(row["failed_validations"]) or [],
                last_pass_percentage=row["last_pass_percentage"],
                execution_mode=row["execution_mode"],
                outcome_data=self._json_loads(row["outcome_data"]),
                confidence_score=row["confidence_score"],
                learned_patterns=self._json_loads(row["learned_patterns"]) or [],
                similar_outcomes_count=row["similar_outcomes_count"],
                success_without_retry=bool(row["success_without_retry"]),
                outcome_category=row["outcome_category"],
                execution_duration_seconds=(
                    row["execution_duration_seconds"]
                    if "execution_duration_seconds" in row.keys()  # noqa: SIM118
                    else None
                ),
                exit_signal=(
                    row["exit_signal"]
                    if "exit_signal" in row.keys()  # noqa: SIM118
                    else None
                ),
                exit_reason=(
                    row["exit_reason"]
                    if "exit_reason" in row.keys()  # noqa: SIM118
                    else None
                ),
            )
            sheets[sheet.sheet_num] = sheet

        # Handle config_path which may not exist in older schemas
        try:
            config_path_value = job_row["config_path"]
        except (IndexError, KeyError):
            config_path_value = None

        state = CheckpointState(
            job_id=job_row["id"],
            job_name=job_row["name"],
            config_hash=job_row["config_hash"],
            config_snapshot=self._json_loads(job_row["config_snapshot"]),
            config_path=config_path_value,
            created_at=self._str_to_datetime(job_row["created_at"])
            or utc_now(),
            updated_at=self._str_to_datetime(job_row["updated_at"])
            or utc_now(),
            started_at=self._str_to_datetime(job_row["started_at"]),
            completed_at=self._str_to_datetime(job_row["completed_at"]),
            total_sheets=job_row["total_sheets"],
            last_completed_sheet=job_row["last_completed_sheet"],
            current_sheet=job_row["current_sheet"],
            status=JobStatus(job_row["status"]),
            sheets=sheets,
            pid=job_row["pid"],
            error_message=job_row["error_message"],
            total_retry_count=job_row["total_retry_count"],
            rate_limit_waits=job_row["rate_limit_waits"],
        )

        # Check for zombie state and auto-recover
        if state.is_zombie():
            _logger.warning(
                "zombie_auto_recovery",
                job_id=job_id,
                pid=state.pid,
                status=state.status.value,
            )
            state.mark_zombie_detected(
                reason="Detected on state load - process no longer running"
            )
            # Save the recovered state
            await self.save(state)

        _logger.debug(
            "checkpoint_loaded",
            job_id=job_id,
            status=state.status.value,
            last_completed_sheet=state.last_completed_sheet,
            total_sheets=state.total_sheets,
            sheet_count=len(sheets),
        )
        return state
save async 
save(state)

Save job state to SQLite.

Source code in src/marianne/state/sqlite_backend.py 
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
async def save(self, state: CheckpointState) -> None:
    """Save job state to SQLite."""
    await self._ensure_initialized()

    state.updated_at = utc_now()

    async with self._connect() as db:
        # Wrap all writes in an explicit transaction so that a crash
        # between the job upsert and the last sheet upsert doesn't
        # leave the database in an inconsistent state.
        await db.execute("BEGIN IMMEDIATE")
        try:
            # Upsert job record
            await db.execute(
                """
                INSERT INTO jobs (
                    id, name, description, status, total_sheets,
                    last_completed_sheet, current_sheet, config_hash,
                    config_snapshot, config_path,
                    pid, error_message, total_retry_count, rate_limit_waits,
                    created_at, updated_at, started_at, completed_at
                ) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)
                ON CONFLICT(id) DO UPDATE SET
                    name = excluded.name,
                    status = excluded.status,
                    total_sheets = excluded.total_sheets,
                    last_completed_sheet = excluded.last_completed_sheet,
                    current_sheet = excluded.current_sheet,
                    config_hash = excluded.config_hash,
                    config_snapshot = excluded.config_snapshot,
                    config_path = excluded.config_path,
                    pid = excluded.pid,
                    error_message = excluded.error_message,
                    total_retry_count = excluded.total_retry_count,
                    rate_limit_waits = excluded.rate_limit_waits,
                    updated_at = excluded.updated_at,
                    started_at = excluded.started_at,
                    completed_at = excluded.completed_at
            """,
                (
                    state.job_id,
                    state.job_name,
                    None,  # description - not in CheckpointState currently
                    state.status.value,
                    state.total_sheets,
                    state.last_completed_sheet,
                    state.current_sheet,
                    state.config_hash,
                    self._json_dumps(state.config_snapshot),
                    state.config_path,
                    state.pid,
                    state.error_message,
                    state.total_retry_count,
                    state.rate_limit_waits,
                    self._datetime_to_str(state.created_at),
                    self._datetime_to_str(state.updated_at),
                    self._datetime_to_str(state.started_at),
                    self._datetime_to_str(state.completed_at),
                ),
            )

            # Upsert sheet records
            for sheet in state.sheets.values():
                await db.execute(
                    """
                    INSERT INTO sheets (
                        job_id, sheet_num, status, attempt_count, exit_code,
                        error_message, error_category, validation_passed,
                        validation_details, completion_attempts, passed_validations,
                        failed_validations, last_pass_percentage, execution_mode,
                        outcome_data, confidence_score, learned_patterns,
                        similar_outcomes_count, success_without_retry,
                        outcome_category, started_at, completed_at,
                        execution_duration_seconds, exit_signal, exit_reason
                    ) VALUES (
                        ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?,
                        ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?
                    )
                    ON CONFLICT(job_id, sheet_num) DO UPDATE SET
                        status = excluded.status,
                        attempt_count = excluded.attempt_count,
                        exit_code = excluded.exit_code,
                        error_message = excluded.error_message,
                        error_category = excluded.error_category,
                        validation_passed = excluded.validation_passed,
                        validation_details = excluded.validation_details,
                        completion_attempts = excluded.completion_attempts,
                        passed_validations = excluded.passed_validations,
                        failed_validations = excluded.failed_validations,
                        last_pass_percentage = excluded.last_pass_percentage,
                        execution_mode = excluded.execution_mode,
                        outcome_data = excluded.outcome_data,
                        confidence_score = excluded.confidence_score,
                        learned_patterns = excluded.learned_patterns,
                        similar_outcomes_count = excluded.similar_outcomes_count,
                        success_without_retry = excluded.success_without_retry,
                        outcome_category = excluded.outcome_category,
                        started_at = excluded.started_at,
                        completed_at = excluded.completed_at,
                        execution_duration_seconds = excluded.execution_duration_seconds,
                        exit_signal = excluded.exit_signal,
                        exit_reason = excluded.exit_reason
                """,
                    (
                        state.job_id,
                        sheet.sheet_num,
                        sheet.status.value,
                        sheet.attempt_count,
                        sheet.exit_code,
                        sheet.error_message,
                        sheet.error_category,
                        self._bool_to_int(sheet.validation_passed),
                        self._json_dumps(sheet.validation_details),
                        sheet.completion_attempts,
                        self._json_dumps(sheet.passed_validations),
                        self._json_dumps(sheet.failed_validations),
                        sheet.last_pass_percentage,
                        sheet.execution_mode,
                        self._json_dumps(sheet.outcome_data),
                        sheet.confidence_score,
                        self._json_dumps(sheet.learned_patterns),
                        sheet.similar_outcomes_count,
                        1 if sheet.success_without_retry else 0,
                        sheet.outcome_category,
                        self._datetime_to_str(sheet.started_at),
                        self._datetime_to_str(sheet.completed_at),
                        sheet.execution_duration_seconds,
                        sheet.exit_signal,
                        sheet.exit_reason,
                    ),
                )

            await db.commit()
        except Exception:
            await db.rollback()
            raise

    _logger.info(
        "checkpoint_saved",
        job_id=state.job_id,
        status=state.status.value,
        last_completed_sheet=state.last_completed_sheet,
        total_sheets=state.total_sheets,
        sheet_count=len(state.sheets),
    )
delete async 
delete(job_id)

Delete state for a job.

Source code in src/marianne/state/sqlite_backend.py 
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
async def delete(self, job_id: str) -> bool:
    """Delete state for a job."""
    await self._ensure_initialized()

    async with self._connect() as db:
        # Check if job exists
        cursor = await db.execute(
            "SELECT id FROM jobs WHERE id = ?", (job_id,)
        )
        if not await cursor.fetchone():
            return False

        # Delete cascades to sheets and execution_history
        await db.execute("DELETE FROM jobs WHERE id = ?", (job_id,))
        await db.commit()
        return True
list_jobs async 
list_jobs()

List all jobs with state.

Source code in src/marianne/state/sqlite_backend.py 
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
async def list_jobs(self) -> list[CheckpointState]:
    """List all jobs with state."""
    await self._ensure_initialized()

    async with self._connect() as db:
        cursor = await db.execute(
            "SELECT id FROM jobs ORDER BY updated_at DESC"
        )
        rows = await cursor.fetchall()

    # Load full state for each job
    states = []
    for row in rows:
        state = await self.load(row[0])
        if state:
            states.append(state)

    return states
get_next_sheet async 
get_next_sheet(job_id)

Get the next sheet to process for a job.

Source code in src/marianne/state/sqlite_backend.py 
638
639
640
641
642
643
async def get_next_sheet(self, job_id: str) -> int | None:
    """Get the next sheet to process for a job."""
    state = await self.load(job_id)
    if state is None:
        return 1  # Start from beginning if no state
    return state.get_next_sheet()
mark_sheet_status async 
mark_sheet_status(job_id, sheet_num, status, error_message=None)

Update status of a specific sheet.

Source code in src/marianne/state/sqlite_backend.py 
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
async def mark_sheet_status(
    self,
    job_id: str,
    sheet_num: int,
    status: SheetStatus,
    error_message: str | None = None,
) -> None:
    """Update status of a specific sheet."""
    state = await self.load(job_id)
    if state is None:
        raise ValueError(f"No state found for job {job_id}")

    if status == SheetStatus.COMPLETED:
        state.mark_sheet_completed(sheet_num)
    elif status == SheetStatus.FAILED:
        state.mark_sheet_failed(sheet_num, error_message or "Unknown error")
    elif status == SheetStatus.IN_PROGRESS:
        state.mark_sheet_started(sheet_num)

    await self.save(state)
record_execution async 
record_execution(job_id, sheet_num, attempt_num, prompt=None, output=None, exit_code=None, duration_seconds=None)

Record an execution attempt in history.

Parameters:

Name Type Description Default
job_id str

Job identifier

 required
sheet_num int

Sheet number

 required
attempt_num int

Attempt number within the sheet

 required
prompt str | None

The prompt sent to Claude

 None
output str | None

The output received

 None
exit_code int | None

Exit code from the execution

 None
duration_seconds float | None

Execution duration

 None

Returns:

Type Description
int | None

The ID of the inserted record, or None if insertion failed.
Source code in src/marianne/state/sqlite_backend.py 
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
async def record_execution(
    self,
    job_id: str,
    sheet_num: int,
    attempt_num: int,
    prompt: str | None = None,
    output: str | None = None,
    exit_code: int | None = None,
    duration_seconds: float | None = None,
) -> int | None:
    """Record an execution attempt in history.

    Args:
        job_id: Job identifier
        sheet_num: Sheet number
        attempt_num: Attempt number within the sheet
        prompt: The prompt sent to Claude
        output: The output received
        exit_code: Exit code from the execution
        duration_seconds: Execution duration

    Returns:
        The ID of the inserted record, or None if insertion failed.
    """
    await self._ensure_initialized()

    async with self._connect() as db:
        cursor = await db.execute(
            """
            INSERT INTO execution_history (
                job_id, sheet_num, attempt_num, prompt, output,
                exit_code, duration_seconds, executed_at
            ) VALUES (?, ?, ?, ?, ?, ?, ?, ?)
        """,
            (
                job_id,
                sheet_num,
                attempt_num,
                prompt,
                output,
                exit_code,
                duration_seconds,
                utc_now().isoformat(),
            ),
        )
        await db.commit()
        return cursor.lastrowid or 0
get_execution_history async 
get_execution_history(job_id, sheet_num=None, limit=100)

Get execution history for a job.

Parameters:

Name Type Description Default
job_id str

Job identifier

 required
sheet_num int | None

Optional sheet number filter

 None
limit int

Maximum records to return

 100

Returns:

Type Description
list[dict[str, Any]]

List of execution history records
Source code in src/marianne/state/sqlite_backend.py 
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
async def get_execution_history(
    self,
    job_id: str,
    sheet_num: int | None = None,
    limit: int = 100,
) -> list[dict[str, Any]]:
    """Get execution history for a job.

    Args:
        job_id: Job identifier
        sheet_num: Optional sheet number filter
        limit: Maximum records to return

    Returns:
        List of execution history records
    """
    await self._ensure_initialized()

    async with self._connect() as db:
        db.row_factory = aiosqlite.Row

        if sheet_num is not None:
            cursor = await db.execute(
                """
                SELECT * FROM execution_history
                WHERE job_id = ? AND sheet_num = ?
                ORDER BY executed_at DESC
                LIMIT ?
            """,
                (job_id, sheet_num, limit),
            )
        else:
            cursor = await db.execute(
                """
                SELECT * FROM execution_history
                WHERE job_id = ?
                ORDER BY executed_at DESC
                LIMIT ?
            """,
                (job_id, limit),
            )

        rows = await cursor.fetchall()
        return [dict(row) for row in rows]
get_execution_history_count async 
get_execution_history_count(job_id)

Get total count of execution history records for a job.

Parameters:

Name Type Description Default
job_id str

Job identifier

 required

Returns:

Type Description
int

Total number of execution history records
Source code in src/marianne/state/sqlite_backend.py 
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
async def get_execution_history_count(self, job_id: str) -> int:
    """Get total count of execution history records for a job.

    Args:
        job_id: Job identifier

    Returns:
        Total number of execution history records
    """
    await self._ensure_initialized()

    async with self._connect() as db:
        cursor = await db.execute(
            "SELECT COUNT(*) FROM execution_history WHERE job_id = ?",
            (job_id,),
        )
        row = await cursor.fetchone()
        return row[0] if row else 0
get_job_statistics async 
get_job_statistics(job_id)

Get aggregate statistics for a job.

Parameters:

Name Type Description Default
job_id str

Job identifier

 required

Returns:

Type Description
dict[str, Any]

Dictionary with statistics including:
dict[str, Any]
  • total_executions: Total execution attempts
dict[str, Any]
  • success_rate: Percentage of successful sheets
dict[str, Any]
  • avg_duration: Average execution duration
dict[str, Any]
  • total_retries: Total retry count
Source code in src/marianne/state/sqlite_backend.py 
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
async def get_job_statistics(self, job_id: str) -> dict[str, Any]:
    """Get aggregate statistics for a job.

    Args:
        job_id: Job identifier

    Returns:
        Dictionary with statistics including:
        - total_executions: Total execution attempts
        - success_rate: Percentage of successful sheets
        - avg_duration: Average execution duration
        - total_retries: Total retry count
    """
    await self._ensure_initialized()

    async with self._connect() as db:
        # Get job data
        cursor = await db.execute(
            "SELECT total_sheets, last_completed_sheet, total_retry_count "
            "FROM jobs WHERE id = ?",
            (job_id,),
        )
        job_row = await cursor.fetchone()

        if not job_row:
            return {}

        # Get execution stats
        cursor = await db.execute(
            """
            SELECT
                COUNT(*) as total_executions,
                AVG(duration_seconds) as avg_duration,
                SUM(CASE WHEN exit_code = 0 THEN 1 ELSE 0 END) as successful
            FROM execution_history
            WHERE job_id = ?
        """,
            (job_id,),
        )
        exec_row = await cursor.fetchone()

        total_sheets = job_row[0]
        completed = job_row[1]
        total_retries = job_row[2]

        return {
            "total_sheets": total_sheets,
            "completed_sheets": completed,
            "success_rate": (completed / total_sheets * 100)
            if total_sheets > 0
            else 0.0,
            "total_retries": total_retries,
            "total_executions": exec_row[0] if exec_row else 0,
            "avg_duration_seconds": exec_row[1] if exec_row else None,
            "successful_executions": exec_row[2] if exec_row else 0,
        }
query_jobs async 
query_jobs(status=None, since=None, limit=50)

Query jobs with filters for dashboard.

Parameters:

Name Type Description Default
status JobStatus | None

Optional status filter

 None
since datetime | None

Only return jobs updated since this time

 None
limit int

Maximum results

 50

Returns:

Type Description
list[dict[str, Any]]

List of job summary dictionaries
Source code in src/marianne/state/sqlite_backend.py 
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
async def query_jobs(
    self,
    status: JobStatus | None = None,
    since: datetime | None = None,
    limit: int = 50,
) -> list[dict[str, Any]]:
    """Query jobs with filters for dashboard.

    Args:
        status: Optional status filter
        since: Only return jobs updated since this time
        limit: Maximum results

    Returns:
        List of job summary dictionaries
    """
    await self._ensure_initialized()

    async with self._connect() as db:
        db.row_factory = aiosqlite.Row

        conditions = []
        params: list[Any] = []

        if status:
            conditions.append("status = ?")
            params.append(status.value)

        if since:
            conditions.append("updated_at >= ?")
            params.append(since.isoformat())

        where_clause = " AND ".join(conditions) if conditions else "1=1"
        params.append(limit)

        cursor = await db.execute(
            f"""
            SELECT id, name, status, total_sheets, last_completed_sheet,
                   created_at, updated_at, completed_at, error_message
            FROM jobs
            WHERE {where_clause}
            ORDER BY updated_at DESC
            LIMIT ?
        """,
            params,
        )

        rows = await cursor.fetchall()
        return [dict(row) for row in rows]