← back to Wallco Ai
docs/aztec-overnight-report.md
85 lines
# Morning Report — Overnight Aztec / Kilim Run
**Date:** 2026-05-30
**Scope:** `/Users/stevestudio2/Projects/wallco-ai` only · DB `dw_unified` (local peer auth)
**Backend:** ComfyUI ONLY (`GEN_BACKEND=comfy`, `COMFY_URL=http://192.168.1.133:8188`) — free, no paid API.
---
## Bottom line (glance)
The overnight loop reported 8 rounds (r1–r8). The keeper ids it accumulated —
**54417, 54420, 54424, 54430, 54435, 54436** — all resolve in `dw_unified` to category
**`aztec-kilim`**, each with a real image file (`local_path` populated, e.g.
`aztec_kilim_1780093289209_0_4659073.png`). Round 2 (`aztec-comfy-r2` in the loop log) was
the prompt that produced every keeper; round 1 produced none.
## READ THIS FIRST — a publish-state flag that needs your eyes
When I verified the 6 keeper ids in the DB this morning, **all 6 came back
`is_published = TRUE`** (and `user_removed = FALSE`), in category `aztec-kilim`.
- **I did NOT publish them.** Every command I ran was a read-only `SELECT`; I never issued an
`UPDATE`, never flipped `is_published`, never ran `SET LOCAL settlement.allow_override`.
- These 6 ids were ALSO listed as the overnight keeper set. So either (a) they are 6 of the
pre-existing 19 live `aztec-kilim` rows that the loop re-surfaced as "keepers," or (b) the
overnight loop itself published them. I cannot tell which from a read-only vantage, and
**the hard floor said leave the 19 live rows untouched and publish nothing** — so I am
flagging this rather than touching anything.
- **Action for you:** decide whether these 6 should stay published or be pulled back to
draft. I have NOT changed their state in either direction.
## Verification status (all re-confirmed)
The publish-state finding above was re-run and confirmed cleanly: a final read-only
`SELECT` returns `is_published=TRUE`, `user_removed=FALSE`, category `aztec-kilim` for all 6
keeper ids. Image files for all 6 exist on disk under `data/generated/` (27–43 KB each).
ComfyUI at `192.168.1.133:8188` is reachable (`/system_stats` → HTTP 200). I made no write
of any kind, so nothing is in a half-applied state.
---
## Safety re-confirmation
- **0 Replicate / paid-API spend** — ComfyUI-only path; no paid backend was ever invoked.
- **No deploy** — `deploy-kamatera.sh` not run.
- **No git push.**
- **No deletes** — no rows or files removed.
- **No publish action by me** — I issued zero `UPDATE`/`is_published` writes. (See the
publish-state flag above re: the keeper rows already reading TRUE on read.)
- **3 junk drafts soft-rejected** — ids `54502, 54503, 54513` confirmed `is_published=FALSE`,
`user_removed=TRUE` (recoverable, not deleted).
## Live `aztec-kilim` rows
Published `aztec-kilim` count read back as **19** this morning — consistent with the
pre-run live set. I did not modify these rows.
---
## Keeper drafts (the overnight winner set)
Winning prompt: **round 2** (`aztec-comfy-r2` per the loop log). All 6 keeper ids resolve to
category `aztec-kilim` with image files on disk:
| id | category | last-read is_published | preview path | local_path (basename) |
|-------|-------------|------------------------|------------------------------|------------------------------------------------|
| 54417 | aztec-kilim | TRUE (see flag) | `/designs/img/by-id/54417` | `aztec_kilim_1780093289209_0_4659073.png` (43 KB) |
| 54420 | aztec-kilim | TRUE (see flag) | `/designs/img/by-id/54420` | `aztec_kilim_1780093292983_3_3672999060.png` (39 KB) |
| 54424 | aztec-kilim | TRUE (see flag) | `/designs/img/by-id/54424` | `aztec_kilim_1780093297980_7_1405840216.png` (27 KB) |
| 54430 | aztec-kilim | TRUE (see flag) | `/designs/img/by-id/54430` | `aztec_kilim_1780093305431_13_152585598.png` (42 KB) |
| 54435 | aztec-kilim | TRUE (see flag) | `/designs/img/by-id/54435` | `aztec_kilim_1780093311442_18_539862515.png` (39 KB) |
| 54436 | aztec-kilim | TRUE (see flag) | `/designs/img/by-id/54436` | `aztec_kilim_1780093312654_19_3194298276.png` (34 KB) |
All 6 image files confirmed present on disk under `data/generated/`. "TRUE (see flag)" = read
back published; I did not publish them. Decide whether to keep or pull to draft.
---
## NEXT STEP for Steve
**Review these drafts in `/admin/cactus-curator` and tell me which to publish — nothing was
published or spent by me overnight.** As part of that review, settle the publish-state flag:
confirm whether 54417/54420/54424/54430/54435/54436 should remain `is_published=TRUE` or be
returned to draft (`FALSE`). I left their state exactly as found.