# DriLux — The Weekly Muse (AI Editorial House)

**Framing (2026-07-11, #638)**: DriLux is a **personal hobby**. The product is the workflow itself — the Studio pipeline, Hermes orchestration, and one excellent editorial a week at ~$50/month of generation. Monetization (Fanvue tiers, paid X analytics, growth targets) exists only as **optional dials** (see Ideas); nothing in it is a prerequisite or a goal.

**Register model (2026-07-11, #668)**: The Studio/API pipeline is **no-nudity** — the xAI API won't return nudes, so it's futile to try. Two prompt **registers** (a wording style, not an API unlock): `editorial` (couture magazine) and `sensual` (default — max-sexy *without* nudity: lingerie, sheer, cleavage, wet, boudoir). The clothed sensual sets are the **setup**; Lucas escalates to nudity himself in the **Grok Imagine app** using the copy-ready Imagine pack (identity + clothed prompts). The Studio never generates or publishes nudity. See `TIER_BOUNDARIES.md`. Supersedes the #652 "Hot lane produces nudity" direction (that half is reverted — the API can't).

**Platform decision (2026-07-11, #625)**: public teasers on **X** (suggestive register, labeled), the complete editorial + making-of behind **Fanvue** (AI-creator disclosure, no-nudity register per #668). Supersedes the #434-era IG/TikTok + Patreon channels; the weekly muse format, disclosure-first posture, and tier logic stand. Strategy: `15_FANVUE_STRATEGY.md`.

**Pivot (2026-07-07, #434)**: DriLux became an **openly AI-generated luxury editorial brand**. One AI muse + one editorial concept per week. (The register model is now #668's editorial/sensual split; premium = more/warmer of the same muse, never a shortcut around the hard rules.) Replaced the previous content direction entirely.

`DriLux-backup/` is a local untracked snapshot of the pre-pivot (#434) docs — gitignored and dockerignored; do not follow, cite, or commit it.

## Current focus
- [ ] Produce the first editorial end-to-end through the Studio
- [ ] Hook Hermes to the weekly batch endpoint (#434)

## Next up
- [ ] Post the first editorial's teasers to X (AI-disclosed, sensitive-media labeled) when it feels ready — no growth targets
- [ ] Wire performance logging into the winning-patterns loop
- [ ] DriLux — add package.json + test script; prompts builders + xai extractJson (#723)
- [ ] adriana-recipes: Meta* Instagram export recipe scanner (#620)

## Last touched
- 2026-07-12 — Frames are now distinct looks with different scenario, clothing and pose; 3 per tier to start, and a regenerate tweak replaces instead of accumulating (#727)
- 2026-07-11 — Restart week stays enabled and force-deletes past the guard, so a hung media job with stuck mediaRunning can no longer trap the editor (#697)
- 2026-07-11 — Generation falls back to in-process when the Cloud Run Job worker can't launch, so a missing IAM grant no longer bricks the Studio (#690)
- 2026-07-11 — Switched to no-nudity with editorial and sensual prompt registers defaulting to max-sexy clothed, and the Imagine pack is now the clothed app handoff (#668)
- 2026-07-11 — Dual-lane model: concept prompts allow body-forward editorial and the Studio builds a copy-ready Grok Imagine hot pack per muse (#652)
- 2026-07-11 — Regenerate keeps the previous image with a spinner and one-click undo until curation completes; Invite and remove moved onto the photo (#651, #650)
- 2026-07-11 — Reframed as a personal hobby: the workflow is the product, monetization parked as optional dials (#638)
- 2026-07-11 — Fanvue direction decided and landed: boundaries rewritten for X public + Fanvue premium, Fanvue strategy doc, and Studio captions now target X and Fanvue (#625)
- 2026-07-11 — Chat-handoff deliverables landed: reality-check + service-strategy docs, X analytics collector with weekly digest, six-role Hermes ops skill, and the public /muse email-capture lifeboat with admin export (#625)
- 2026-07-10 — Cloud Run now keeps the instance up and unthrottled through long generation batches, so headless Sunday runs finish without a Studio tab open (#588)

## Blockers
- (none)

## Open questions
- One Invite teaser from up to 3 Invite photos vs a separate teaser per Invite photo?

## Ideas
- **Monetization dials (off by default, #638)**: Fanvue account + tiers (captions already generate for it; pre-launch checklist in 15_FANVUE_STRATEGY includes reading primary policy texts and pricing validation); X analytics collector (ready in analytics/, needs the paid X API tier ~$200/mo — only worth it past ~$300/mo revenue); ACC enrollment if the X feed ever crosses into adult-classified
- Member concept votes wired into the concepting stage
- Quarterly season-retrospective PDF for members

## Notes
- Parked work (2026-07-10): the in-flight **#589 durable-job-worker** changes (`app/jobs.mjs`, `jobs/run.mjs`, big `mount.mjs`/`pipeline.mjs`/`xai.mjs` rewrites, `lib/gcs-persist.mjs`, `deploy-cloud-run.yml`) live on branch **`wip/pre-pull-snapshot`** (commit `5f730996`, also pushed to origin). Do **not** delete or prune that branch until #589 ships — recover with `git checkout wip/pre-pull-snapshot -- <paths>`, then delete local + remote copies.

## Infrastructure
- Studio at `/drilux` (admin-gated), state GCS-persisted; xAI key in repo `.env`
- Video generation: xAI `/v1/videos/generations` async (grok-imagine-video, $0.05/s); knobs: `DRILUX_VIDEO_SECONDS` (6), `DRILUX_VIDEO_CONCURRENCY` (2), `XAI_VIDEO_MODEL`