[PsychedelicShayna]

What

Finish the visible NeoCode/NeoPi identity cutover across CLI guidance, provider settings, usage presentation, shared display fallbacks and provider diagnostics. The harness is named NeoPi, not NeoPi/OMP. Existing explicitly named instances remain authoritative, including names saved by older versions.

Project-schema guidance names NeoCode without changing t3.json fields or its compatibility URL. Triage uses this fork's repository and nightly playbook; upstream attribution remains explicit.

Why

Addresses #59. NeoCode is NeoPi's graphical frontend, not an upstream T3 Code instance with an optional provider. Earlier changes already established application branding, README guidance, NeoPi-first ordering and fresh-settings defaults; this removes the remaining bundled-app naming gaps.

A direct read of this fork's current main playbook showed upstream instructions. Refreshing from main would therefore undo the new fork routing before promotion. The playbook now refreshes from nightly and uses nightly as the fallback source branch.

No command names, provider IDs, data directories, saved choices or protocol identifiers change. OMP executable compatibility and other harnesses remain supported. Actual upstream services, legal attribution, quoted testimonials and upstream distribution guidance are not relabeled as fork services. CLI release provenance is separately repaired by merged #107; its stable publisher follow-up is #106.

Testing

• Twelve focused files initially ran 121 tests: 120 passed, one failed because a resume-validation assertion pinned the prior product wording. The corrected assertion checks the typed validation error; its isolated rerun passed 1 test, 8 skipped. The other 120 results are retained, not rerun.
• Focused files: server bin.test.ts; CLI app.test.ts, pair.test.ts, service.test.ts, triagePrompt.test.ts; provider NeoPiAdapter.test.ts, neopiUsageLimits.test.ts, mappers/ui.test.ts, NeoPiDriver.test.ts; shared/contracts t3ProjectFile.test.ts; client-runtime providerInstanceDisplay.test.ts.
• After correcting the playbook branch, the argv-safety regression passed 1 test. A throwaway check verified canonical/embedded text agreement, the fork nightly refresh URL and nightly fallback. No permanent prompt-copy test remains.
• Targeted formatting passed for all 41 changed files and the three follow-up files; the commit hook formatted staged files.
• Installed and visually verified: desktop batch c9f91e00a43096e5f94e050627738b26bc7a0c9b, version 0.0.42-nightly.20261020.c9f91e00, includes this PR and fix(releases): download CLI runtimes from the NeoCode fork #107. Real launch and Settings → Providers passed in private Xvfb :10, with normal shell hydration. NeoPi is first, enabled/authenticated, and the saved npi-nightly path survived. The six other harnesses remain disabled in this fixture. Before/after screenshots were saved and viewed; no owner database/wallet was used. Installed receipt.
• Public installed-app proof: before · after · structured receipt. The exact locally installed AppImage and its checksum are attached to the release whose tag resolves directly to c9f91e00a43096e5f94e050627738b26bc7a0c9b. The separate CI AppImage is not claimed byte-identical.
• Installed server CLI --help also identifies NeoCode. No full suite or package typecheck; no Windows/native mobile execution or repeated session-restoration/model-turn acceptance run claimed.

---

• vp check and vpr typecheck pass (CI runs both; locally run only the checks for the touched scope, per AGENTS.md › Verifying) — Focused regressions and formatting only.
• Tested locally
• Before/after images (video for motion) uploaded to GitHub for UI changes, not committed (if user-facing)

Integration: openai-codex/gpt-6-astra. Implementation: openai-codex/gpt-6.1-sol. Focused verification: openai-codex/gpt-6-luna. Harness: NeoPi.