You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Commit db0e084
Browse filesBrowse the repository at this point in the historyBrowse files
docs: retire plans/phase3-platform-plugin-progress.md; track remaining work in issues (#982)
The remaining Phase 3 Apple PlatformPlugin work (steps b + d) is now filed as
GitHub issues under umbrella #972, so the standalone progress plan is redundant
and a staleness hazard (it already drifted once re: cost.runnerRoundTrips).
- Remove plans/phase3-platform-plugin-progress.md.
- Repoint its references at the durable sources: perfect-shape.md (x3) and
ADR-0009 now link the Phase 3 tracking issue #972; the plugin.ts step-b facet
note points at ADR-0009 (+ issue #974). Design rationale stays in
perfect-shape.md and ADR-0009; live status lives in the issues.
implementation state is tracked in [phase3-platform-plugin-progress.md](./phase3-platform-plugin-progress.md).
296
+
implementation state is tracked in the [Phase 3 tracking issue #972](https://github.com/callstack/agent-device/issues/972).
|**1 · command spine**| (a) **invert the import graph** — `commandRegistry` becomes root, `command-catalog`/`capabilities`/`daemon-registry`/`batch-policy` derive (parity-tested, no deletion yet); (b) promote each family's facet → `CommandDescriptor` additively; (c) replace the 24-arm switch with the total map, arm-by-arm |**med** (the import-cycle inversion is the real first-week blocker: `command-catalog` has ~95 importers and the facet imports `AgentDeviceClient` today) | finishes a proven seam; add-command → ~2 files; enables everything below |
400
400
|**2 · typed results** (the parity oracle) | (a) `CommandResultMap` with `Record` default, migrate per-command from real runner payloads; (b) graft `TypedError`; fold the disowned 'generic' family into `handlers/`**last**; (c) kill the `client-types.ts` mirror (~550 LOC) | med (must be per-command, never a big-bang retype of 203 files) | the safety net the platform unwind needs; biggest single LOC win |
401
-
|**3 · platform plugin** (now safe) | (a) define `PlatformPlugin`, **lazy** factories (cold-start benchmark guards latency); (b) move capability columns onto plugin grants, **porting every `supports()` closure verbatim**, pinned by the table-equivalence test before deletion; (c) ✅ unwind macOS and the OS-agnostic Apple engine out of `platforms/ios`; (d) finish Apple plugin facets, tvOS leaf, final `Platform` collapse, and watchOS sentinel last. The **Apple plugin is the first instance** and owns the `AppleOS` leaves — see ADR 0009 and [phase3-platform-plugin-progress.md](./phase3-platform-plugin-progress.md)|**high** (touches shared platform routing and the XCTest runner) | add-platform wiring → ~3 files; kills the 231-branch smear |
401
+
|**3 · platform plugin** (now safe) | (a) define `PlatformPlugin`, **lazy** factories (cold-start benchmark guards latency); (b) move capability columns onto plugin grants, **porting every `supports()` closure verbatim**, pinned by the table-equivalence test before deletion; (c) ✅ unwind macOS and the OS-agnostic Apple engine out of `platforms/ios`; (d) finish Apple plugin facets, tvOS leaf, final `Platform` collapse, and watchOS sentinel last. The **Apple plugin is the first instance** and owns the `AppleOS` leaves — see ADR 0009 and the [Phase 3 tracking issue #972](https://github.com/callstack/agent-device/issues/972)|**high** (touches shared platform routing and the XCTest runner) | add-platform wiring → ~3 files; kills the 231-branch smear |
402
402
|**4 · agent-cost** (opt-in) | (a) `ResponseView.toView` with `default`==today; `responseLevel` knob defaulting to `default`; (b) typed `BatchStepResult` → intermediate steps digest; per-command MCP `outputSchema`; generalize zero-load fast-paths | med (wire-shape risk vs Maestro — strictly opt-in) | the north-star-#2 token/latency wins |
403
403
|**5 · layering + legacy** (quiet windows) | intent-folder moves + utils extraction as pure path codemods; at next major drop the ~175 LOC of legacy aliases/barrels | low-per-step, high-diff | scoped ownership + import lint; merge-pain risk → land fast, small |
404
404
@@ -433,7 +433,7 @@ grants; bundling the folder reorg with the registry work (maximizes diff-noise a
433
433
## 8. Before / after — the command axis
434
434
435
435
(The platform-axis decision lives in ADR 0009; current implementation status lives in
0 commit comments