Skip to content

ROADMAP.md

Latest commit

 

History

History
65 lines (48 loc) · 3.75 KB

ROADMAP.md

File metadata and controls

65 lines (48 loc) · 3.75 KB

Project Roadmap

Current Architecture Baseline

  • Transport -> Connection -> Runtime -> Orchestration -> Views is the intended current direction
  • CodexConnection is the only app-server boundary
  • the separate client layer has been removed
  • Runtime is a sort of "diverter" intended to help with server reqs/notifs we don't need user input on.
  • the connection schema generator and verifier currently report 497/497 exported schema types represented
  • until codex reaches v1.x.x, the schema target is the latest released upstream codex app-server contract

Transition Note

The protocol and schema boundary is substantially ahead of the app shell.

Checked-in source still shows migration seams:

  • CodaxApp uses CodaxOrchestrator
  • several view files still contain placeholders or CodaxViewModel references
  • orchestration tests still target the older CodaxViewModel design
  • narrative docs had previously overstated the completion of SwiftData-backed views, inspector flows, and toolbar behavior

What's Done

  • Local codex app-server process launch and stdio transport
  • JSON-RPC request/response routing in CodexConnection
  • typed CodexConnection request API for every ClientRequest method in codex-schemas/v0.114.0
  • typed ServerNotificationEnvelope for every ServerNotification method
  • typed ServerRequestEnvelope and ServerRequestResponse for every ServerRequest method
  • generated connection-owned Swift representation for every exported schema in codex-schemas/v0.114.0
  • connection schema verification gate via node Tools/update_connection_schema_progress.js --verify
  • generated schema output ignored in .gitignore
  • CodexRuntimeCoordinator owns transport startup and typed event forwarding
  • CodaxOrchestrator exists as the current checked-in app-facing observable object
  • the app entrypoint uses NavigationSplitView

What's In Progress

App-layer migration

  • Finish migrating app behavior from CodaxViewModel assumptions to CodaxOrchestrator
  • Unify the SwiftUI environment surface so views no longer reference removed or stale app-layer types
  • Reconcile orchestration tests with the current app-layer architecture
  • Finish gutting initial SwiftData kickstart, impl slimmer CoreData persistence for persisting "Project" abstraction(s).

UI/UX/AX

  • Add approval, elicitation, and auth-refresh UX
  • Add exec-backed, UI-selectable, git workflow presets for common flows, including commit and commit-and-push UX with approval, output, and failure handling
  • Persist durable project, config, and catalog projections where delayed rendering matters
  • Account and login polish around pending-login, browser handoff (needs impl w/ webkit authview), and auth-refresh flows
  • Improve item, diff, plan, and realtime rendering
  • Dive into accessibility-first UI/UX refinement
  • Full SwiftUI/AppKit AX support. This gets done right for every accessibility need. I'll dive into CarbonAX so deep my dreams have kPrefixes, if I have to.
  • Finish experimenting and evaluating different TTS implementations and flows

Automation and quality

  • Add automated regen and verification of the connection schema, should fail CI on coverage drift
  • Add automated verification for the UI-facing app-server coverage tracker
  • Add targeted tests for the full generated connection boundary where useful

Immediate Priority

The highest-confidence part of the codebase is still the protocol stack: transport, connection typing, schema generation, and runtime forwarding. The highest-value next work is the migration above that boundary: aligning orchestration, views, tests, and documentation around one app-layer design before claiming richer UX completion.