feat(sandbox): add checkpoint and rollback hooks #462

2026-02-27T04:29:08Z

freemo commented

2026-02-27 04:29:08 +00:00

Summary

Adds sandbox checkpoint and rollback hooks integrated into plan execute/apply flows with metadata capture. Checkpoints preserve sandbox state at key points, and rollback restores to a specific checkpoint with metadata preserved.

Changes

SandboxCheckpoint Model & CheckpointManager

SandboxCheckpoint frozen Pydantic model with ULID-based checkpoint_id, sandbox_id, plan_id, phase, created_at, metadata, snapshot_path
CheckpointManager class managing checkpoint lifecycle: create_checkpoint, rollback_to, list_checkpoints, delete_checkpoint
Uses filesystem-based snapshots for state preservation

Plan Execute/Apply Integration

Pre-execute checkpoint creation in plan_executor.py
Post-execute checkpoint on success
Pre-apply checkpoint in plan_apply_service.py
Automatic rollback attempt on execute/apply failure
Optional injection — hooks silently skip if no CheckpointManager

Documentation

New docs/reference/sandbox.md with checkpoint lifecycle diagram and rollback behavior

Tests

Behave: 12 scenarios in features/sandbox_checkpoints.feature (all pass)
Robot: 5 test cases in robot/sandbox_checkpoint_smoke.robot (all pass)
ASV: 3 benchmarks in benchmarks/sandbox_checkpoint_bench.py

Quality Gates

lint: PASS
typecheck: PASS (0 errors, 0 warnings)
unit_tests: PASS
integration_tests: PASS

Closes #183

## Summary Adds sandbox checkpoint and rollback hooks integrated into plan execute/apply flows with metadata capture. Checkpoints preserve sandbox state at key points, and rollback restores to a specific checkpoint with metadata preserved. ## Changes ### SandboxCheckpoint Model & CheckpointManager - `SandboxCheckpoint` frozen Pydantic model with ULID-based `checkpoint_id`, `sandbox_id`, `plan_id`, `phase`, `created_at`, `metadata`, `snapshot_path` - `CheckpointManager` class managing checkpoint lifecycle: `create_checkpoint`, `rollback_to`, `list_checkpoints`, `delete_checkpoint` - Uses filesystem-based snapshots for state preservation ### Plan Execute/Apply Integration - Pre-execute checkpoint creation in `plan_executor.py` - Post-execute checkpoint on success - Pre-apply checkpoint in `plan_apply_service.py` - Automatic rollback attempt on execute/apply failure - Optional injection — hooks silently skip if no CheckpointManager ### Documentation - New `docs/reference/sandbox.md` with checkpoint lifecycle diagram and rollback behavior ### Tests - **Behave**: 12 scenarios in `features/sandbox_checkpoints.feature` (all pass) - **Robot**: 5 test cases in `robot/sandbox_checkpoint_smoke.robot` (all pass) - **ASV**: 3 benchmarks in `benchmarks/sandbox_checkpoint_bench.py` ## Quality Gates - lint: PASS - typecheck: PASS (0 errors, 0 warnings) - unit_tests: PASS - integration_tests: PASS Closes #183

freemo referenced this pull request

2026-02-27 04:29:19 +00:00

feat(sandbox): add checkpoint and rollback hooks #183

freemo force-pushed feature/m4-checkpoints from 0f4eb80678 to 63e6d23946

2026-02-27 19:35:46 +00:00

Compare

freemo scheduled this pull request to auto merge when all checks succeed 2026-02-27 19:35:59 +00:00

freemo force-pushed feature/m4-checkpoints from 63e6d23946 to 8fc08d1069

2026-02-27 21:09:00 +00:00

Compare

freemo force-pushed feature/m4-checkpoints from 8fc08d1069 to 8096744b13

2026-02-27 21:34:07 +00:00