17 lines
4.5 KiB
Plaintext
17 lines
4.5 KiB
Plaintext
# Ralph Progress Log
|
|
Started: 2026-04-19
|
|
---
|
|
- Established native GPUI shell parity direction by reading PRD.md, Electron source surfaces, and visual screenshots before editing.
|
|
- Completed native shell, onboarding, overview, sidebar, and messages parity work.
|
|
- Completed approvals, documents, runs, inbox, settings, and workspaces parity passes.
|
|
- Added native storage-backed verification covering reviews, approvals, runs, and documents.
|
|
- Remaining work is concentrated in interaction depth, animation/state parity, and full runtime contract completion.
|
|
- Standing constraint: all code is written directly in the current context; no subagents or agent teams are used.
|
|
- Completed real settings/workspace state mutations in native Rust path: theme, locale, execution mode, and primary workspace.
|
|
- Added native_settings_workspace_e2e.rs to verify mutation persistence across reopen.
|
|
- Completed first native motion/state parity layer using GPUI animation wrappers for onboarding, inbox, and runs.
|
|
- Added runtime contract persistence entities for run events and artifacts, exposed them in ProfileSnapshot, and rendered them in the native runs inspector.
|
|
- Added runtime_contract_snapshot_e2e.rs to verify runtime event and artifact persistence across reopen.
|
|
- Runtime contract integration advanced: RuntimeEvent ingestion now persists message/tool/artifact/run-finished events into storage and exposes them in the runs inspector.
|
|
- Wired a live native runtime session into the GPUI Runs surface. Selected runs now launch Pi Mono through env-configured stdio sidecar settings, runtime tool calls create approval rows, approval decisions send ToolApproval back into the live session, runtime messages project into agent_messages, and run lifecycle still persists into run_events/artifacts/runs.\n- Closed the final native verification matrix. The workspace now has explicit storage-backed e2e coverage for onboarding bootstrap, approvals/documents/runs visibility, settings/workspace mutations, live runtime contract flows, and inbox attention derivation.\n- Reopened Ralph parity work after a direct source audit. The current native app has real runtime/data flows, but several surfaces still diverge from Electron page grammar: onboarding is too collapsed, dashboard misses banner/attention/chart structure, inbox filters are not real, run detail is dominated by native control slabs not present in source, approvals flatten board/user/manager semantics, and documents still lack work-product/editor hierarchy. Added US-010 through US-014 to close these parity gaps honestly.
|
|
- Rebuilt native onboarding away from the old static split shell. It now has a real multi-stage company/mission/launch flow, screenshot-aligned narrative/summary composition, real connector readiness derived from snapshot data, and a storage bootstrap transaction that creates company, workspace, goal, CEO, kickoff project, kickoff task, and queued run. Remaining US-010 work is the finer original wizard-tab parity.\n- Rebuilt the native dashboard away from the placeholder metrics/activity shell. It now has real recovery and budget banners, attention rows, a performance section, active work, and recent activity, all driven by persisted local data. Remaining US-011 work is the inbox-side parity and finer dashboard detail matching.\n- Restored more source grammar across inbox, approvals, and runs. Inbox now has real interactive filters driving a unified attention stream; approvals again distinguish board/user/manager review groupings; and run detail no longer lets live runtime controls dominate the main header region.\n- Restored document-center grammar toward the source surface: the native page now separates documents prose from work products and uses persisted artifacts as the work-product list instead of hiding them behind the runs inspector only.\n- Restored the communication shell toward the source product grammar: grouped channels, unread badges, channel header, message stream, composer, and side inspector. Combined with the earlier document-center rewrite, the remaining work is now the final parity audit rather than major page-grammar gaps.\n- Closed the final parity remediation slice. Native onboarding now restores explicit wizard-tab semantics (nodecheck/providers/company/modelpool/executionmode/mission/launch), runs now use real queue filtering/search state instead of static shells, messages now use real search plus unread filtering instead of static filter chrome, and the remaining off-source empty-state copy in runs/documents/messages/onboarding was eliminated. Final cargo clippy and cargo test both pass.\n |