feat(decopilot): durable agent loop with crash recovery and resume#2807

Merged

tlgimenes merged 10 commits intomainfrom

tlgimenes/agent-durability

Mar 23, 2026

Commits on Mar 23, 2026

feat(decopilot): add durable agent loop with crash recovery and resume

tlgimenes
and
claude
committed
fix(decopilot): auto-recover all orphaned runs, not just automations

tlgimenes
and
claude
committed
fix(decopilot): address PR review feedback for durable agent runs

tlgimenes
and
claude
committed
fix(decopilot): remove stale run_started_at fallback from claim CAS

tlgimenes
and
claude
committed
feat(decopilot): transparent orphan resume in GET /attach

tlgimenes
and
claude
committed
fix(decopilot): handle stale-pod orphans and improve resume UX

tlgimenes
and
claude
committed
fix(decopilot): throw 403 on model permission failure in attach orphan path

tlgimenes
and
claude
committed
fix(decopilot): move stale-pod recovery to CAS claim instead of separate method

tlgimenes
and
claude
committed
feat(decopilot): add DB CAS on RUN_STARTED and per-pod NATS KV heartbeat

tlgimenes
and
claude
committed
fix: renumber durable-agent-runs migration to 050 after rebase

tlgimenes
and
claude
committed