v0.6.1 — author.* commands proxy through attune-author orchestration (D2)

Highlights

Phase D2 of the architecture-realignment spec, gui side. The six author.* executors moved to attune-author 0.9.0; this release wires the gui to dispatch through them via attune_author.orchestration.run_command(...).

• init, status, maintain, lookup, regen, setup now register as proxy CommandSpecs
• Two host-only behaviours stay in the gui via the new _author_proxy(name, *, pre_resolve_workspace=False, invalidate_after=False) helper:
  • pre_resolve_workspace=True — _resolve_project_paths runs before dispatch so the workspace fallback (which the orchestration runtime intentionally lacks) still works
  • invalidate_after=True — attune_gui.routes.rag.invalidate(project_root) runs after dispatch (Phase D4 will replace this with public pipeline_for(corpus_id))
• commands.py shrinks by ~480 lines; test_commands.py shrinks by ~440 lines (executor coverage now lives in attune-author's tests)
• Net diff: +213 / −951

Dependencies

Pins attune-author[ai]>=0.9.0,<0.10 (was >=0.8.1,<0.9). Pairs with attune-author v0.9.0.

Architecture-realignment progress

A · B1 · B2 · C · D1 · D2 done · D3 (help.* 3 commands) · D4 (gui sweep — pipeline_for + error envelope) remaining.