Repository navigation
Deferred: escalation API and identity model Phase B #166
Description
Context
Intentionally deferred items from #146 (Identity model). Phase A (columns, StepMetadata, ConversationMessage, store methods) is done. Phase B is escalation.
Items
-
escalate()API — fork a workflow for human escalation, create child withparent_workflow_idlink -
list_by_parent()helper — query escalation chains without raw SQL - Cap escalation depth — enforce max 2 levels via
parent_workflow_iddepth check - Reference vs copy — child loads parent turns via session_id/parent_workflow_id join (per review: reference, don't copy)
Related
- Identity and association model for workflows, conversations, and escalations #146 (closed) — Identity model Phase A
- ChatWorkflowRunner: multi-turn chat as dynamic workflow #145 (closed) — ChatWorkflowRunner (escalation target)
Activity
beesarmy commentedon Aug 24, 2026
Triage
Summary: Phase B of the identity model — escalation API, parent workflow links, and depth capping. Deferred from #146.
Affected components:
src/cloud_agents/workflow/executor/chat/runner.py— would needescalate()methodsrc/cloud_agents/storage/run_state_store.py— would needlist_by_parent()querysrc/cloud_agents/workflow/temporal_models.py—parent_workflow_idfield exists in schema (via Alembic migration)Related issues/PRs:
Assessment: Well-scoped deferred work. The "reference vs copy" decision (load parent turns via join, not duplication) is the right call from the review. This is lower priority than the production-readiness items in #167 — escalation is a feature, not a blocker.
Suggested labels:
area/chat,feature🤖 Triaged by beesarmy via Claude Code