Verification, not bookkeeping: the read an agent reaches for when asking "did that refresh index what I moved" gives the wrong answer, and I walked into it tonight.
Measured 2026-11-03
Read
Answer
GET /Items?IncludeItemTypes=Movie&Recursive=true
2,101 items
GET /Items?ParentId=<Movies ItemId>
2,548 children — one per folder on disk
Eight folders the recursive list does not return — the three Mobile Suit Gundam 00 Special Edition films and the five Sailor Moon Memorial films — each answer as Type: Movie through SearchTerm and appear in the folder listing, at the paths the journal moved them to. So the shortfall is grouping, not missing media.
The consequence
Verifying a move against the recursive type list produces a false "not indexed" verdict. The next action that verdict invites is another full-library scan, which costs CPU on a host serving users and changes nothing. The page now states which read proves an item, and demotes the recursive count to a trend line.
Diff
One law paragraph in serverAppConfigs-docs/Dankserver/Jellyfin.md, placed after the coverage check (the two reads an agent runs in the same sitting). No script, unit, or config touched.
./tests/run.sh — 822 seams, all pass. The vault's own link-integrity and OKF seams cover the edited page.
The procedural copy of this law lives in the operator's managed skill jellyfin-index-completeness-census, which now links here as the owner; the DB-side sibling trap (BaseItems.Type stores CLR class names, so Type='Movie' counts zero) is recorded there rather than duplicating this page.
Verification, not bookkeeping: the read an agent reaches for when asking "did that refresh index what I moved" gives the wrong answer, and I walked into it tonight.
## Measured 2026-11-03
| Read | Answer |
|---|---|
| `GET /Items?IncludeItemTypes=Movie&Recursive=true` | 2,101 items |
| `GET /Items?ParentId=<Movies ItemId>` | 2,548 children — one per folder on disk |
Eight folders the recursive list does not return — the three *Mobile Suit Gundam 00 Special Edition* films and the five *Sailor Moon Memorial* films — each answer as `Type: Movie` through `SearchTerm` and appear in the folder listing, at the paths the journal moved them to. So the shortfall is **grouping, not missing media**.
## The consequence
Verifying a move against the recursive type list produces a false "not indexed" verdict. The next action that verdict invites is another full-library scan, which costs CPU on a host serving users and changes nothing. The page now states which read proves an item, and demotes the recursive count to a trend line.
## Diff
One law paragraph in `serverAppConfigs-docs/Dankserver/Jellyfin.md`, placed after the coverage check (the two reads an agent runs in the same sitting). No script, unit, or config touched.
`./tests/run.sh` — **822 seams, all pass**. The vault's own link-integrity and OKF seams cover the edited page.
The procedural copy of this law lives in the operator's managed skill `jellyfin-index-completeness-census`, which now links here as the owner; the DB-side sibling trap (`BaseItems.Type` stores CLR class names, so `Type='Movie'` counts zero) is recorded there rather than duplicating this page.
---
cwd: /home/blank/.omp/wt/wt-20261030-163847-6d1fc1f · session: 01a18f2e-18b1-7573-bcd5-ac99636f8bcb
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Verification, not bookkeeping: the read an agent reaches for when asking "did that refresh index what I moved" gives the wrong answer, and I walked into it tonight.
Measured 2026-11-03
GET /Items?IncludeItemTypes=Movie&Recursive=trueGET /Items?ParentId=<Movies ItemId>Eight folders the recursive list does not return — the three Mobile Suit Gundam 00 Special Edition films and the five Sailor Moon Memorial films — each answer as
Type: MoviethroughSearchTermand appear in the folder listing, at the paths the journal moved them to. So the shortfall is grouping, not missing media.The consequence
Verifying a move against the recursive type list produces a false "not indexed" verdict. The next action that verdict invites is another full-library scan, which costs CPU on a host serving users and changes nothing. The page now states which read proves an item, and demotes the recursive count to a trend line.
Diff
One law paragraph in
serverAppConfigs-docs/Dankserver/Jellyfin.md, placed after the coverage check (the two reads an agent runs in the same sitting). No script, unit, or config touched../tests/run.sh— 822 seams, all pass. The vault's own link-integrity and OKF seams cover the edited page.The procedural copy of this law lives in the operator's managed skill
jellyfin-index-completeness-census, which now links here as the owner; the DB-side sibling trap (BaseItems.Typestores CLR class names, soType='Movie'counts zero) is recorded there rather than duplicating this page.cwd: /home/blank/.omp/wt/wt-20261030-163847-6d1fc1f · session: 01a18f2e-18b1-7573-bcd5-ac99636f8bcb