Stale 'not on PyPI' references after the September publishes — #410, #282, C-50, #63, one report

[Polichinel]

Three platform packages this repo depends on were published to PyPI between 28 July and 15 September,
and a set of documents and register entries still describe them as unpublished. None of it changes
what runs; all of it misleads the next reader. One ticket, one PR.

What is now true (verified against PyPI 2026-09-15)

package	on PyPI	since	models here
views-baseline	1.0.2	1.0.0 on 2026-07-28; 1.0.2 fixes #445	37
views-hydranet	0.1.0	2026-09-15 (views-hydranet#340)	8, all pinned ~=0.1.0 (#462)
views-postprocessing	1.2.0	—	FAO launcher (still installs from git+…@main, separately)

All 8 HydraNet models were verified to install from real PyPI in a clean venv and import
HydranetManager on 2026-09-15.

What still says otherwise

Living documents — fix these:

1. PR Release: development → main (442 commits) — G5 of the runbook, HOLD at G4b #410 (the development → main release), section "What this release explicitly does not do":

   resolve views-hydranet or views-postprocessing not being on PyPI (which is why 8 models and the
   FAO launcher still cannot pin to released packages)

   Both are published. Also lists "fix the views-r2darts2 three-way split (C-115)" — done in fix(#317): one satisfiable views-r2darts2 spec across all 31 models #461.

2. Issue views-models release runbook — G0 at 5/6, blocked by #336 #282 (the release runbook):

   • §0 publish table: views-postprocessing — not on PyPI, views-hydranet — not on PyPI
   • pin table (~line 107): views-hydranet >=0.1.0,<1.0.0 (8 models) ❌ not on PyPI — the pin is
     now ~=0.1.0 and it resolves; views-postprocessing ❌ not on PyPI
   • pin table: views-r2darts2 — three specs across 31 models — see C-115 — one spec since fix(#317): one satisfiable views-r2darts2 spec across all 31 models #461
   • §0 table: views-baseline 1.0.1 — 1.0.2 now, and 1.0.1 was the broken one
   • "Two blockers remain for 'pin to released versions'" — hydranet is resolved; only the FAO
     launcher's git+…@main install remains, and "not on PyPI" is no longer the reason for it
   • "does not do" list (~line 186): "Resolve views-hydranet or views-postprocessing publication"

3. reports/technical_risk_register.md C-50 — "views-baseline not published to PyPI", Status
   Open. Stale since 2026-07-28. Should be Resolved with the publish date and a pointer to fix(#459): adopt views-baseline 1.0.2 — raise the floor ×37, and make the smoke test able to fail #460.

4. Issue chore: verify C-42 resolution and track views-baseline PyPI publish (C-50) #63 — tracks C-42 / C-50, the baseline publish. Closable once C-50 is marked.

Historical record — mark, don't rewrite:

5. reports/conda_to_uv_migration_investigation.md — lines 20, 307, 348, 469, 510, 870–871, 1536
   say views-baseline / views-hydranet are not on PyPI and propose git+ dependencies for them. It is a
   dated investigation and line 1536 already hedges "as of this writing … check PyPI". One line at the
   top noting the publishes, with dates, is enough; do not edit the body.

Not in scope

• The FAO launcher's git+…@main install of views-postprocessing. Real, separate, and a moving
  branch pointer is the problem there — not PyPI availability. Belongs with views-pipeline-core#222.
• CI installing views-hydranet so test_roster_configs_load.py stops skipping. That is a workflow
  change, not a doc change; its own issue.

Acceptance

• Release: development → main (442 commits) — G5 of the runbook, HOLD at G4b #410 and views-models release runbook — G0 at 5/6, blocked by #336 #282 no longer say views-hydranet or views-postprocessing are unpublished
• views-models release runbook — G0 at 5/6, blocked by #336 #282's pin table matches models/*/requirements.txt on development
• C-50 Resolved; chore: verify C-42 resolution and track views-baseline PyPI publish (C-50) #63 closed
• grep -rn "not on PyPI" docs/ reports/ README.md returns only lines that are explicitly dated history

[Polichinel]

Item 1 (#410) and item 2 (#282) done — bodies edited in place 2026-09-15, both packages. Six hunks total; before/after saved. Remaining: C-50 (register), #63 (close), and the one-line note on the uv investigation report. Those three are a small PR.

[Polichinel]

All five items done. #410 and #282 edited in place (2026-09-15); C-50 and C-42 Resolved and the uv report annotated in #470, merged as c1e4a83. #63 closes with this.