docs(runtime): record checkpoint artifact id rollout

This commit is contained in:
Saravana Achu Mac 2026-04-04 14:27:34 -07:00
parent 59ae0e1943
commit e6fb20f86b
3 changed files with 2 additions and 5 deletions

View File

@ -365,7 +365,6 @@ Current state:
Next conforming steps:
1. document which Cowork transitions are direct observations versus projections derived by `cowork-service`
2. decide whether session/task IPC should eventually expose first-class checkpoint artifact IDs in addition to refs
Conformance bar:

View File

@ -270,12 +270,11 @@ These should be resolved before claiming the ecosystem docs are fully implementa
### 6.1 Remaining Direct Runtime TODOs
- Cowork: decide whether checkpoint summaries should grow first-class artifact IDs in addition to the new artifact and memory refs.
- FlowMonk: decide whether the new backend editing surfaces need dedicated product UI affordances now or can remain API-first.
### 6.2 Explicit Blockers And Questions
- Question: should Cowork eventually expose first-class checkpoint artifact IDs on session/task IPC projections, or are the current ref arrays enough for the current roadmap?
- Question: should FlowMonk approvals and todos stay API-first for now, or is a dedicated product UI the next priority?
---

View File

@ -120,9 +120,8 @@ Observed baseline:
- Cowork session and task IPC projections now also expose canonical event IDs directly, and checkpoint summaries preserve artifact and memory refs when present.
- FlowMonk now emits direct runtime projections for planning sessions, tasks, todos, runs, approvals, actions, and checkpoints with persisted native runtime records plus direct PATCH editing surfaces for approvals and todos.
- remaining follow-through is now evolutionary rather than blocking:
- Cowork can later expose first-class checkpoint artifact IDs beyond the current ref arrays
- FlowMonk can later add richer UI affordances on top of the new backend editing surfaces
## 8. Explicit Blockers And Questions
- Question: should Cowork eventually expose first-class checkpoint artifact IDs on session/task IPC projections, or are the current artifact/memory ref arrays enough for Phase 5?
- Question: should FlowMonk approvals and todos remain API-first for now, or should a dedicated product UI become the next runtime slice?