rbrooks/Quest-Board
Watch
1
0
Fork
You've already forked Quest-Board
 0
Code Issues 58 Pull requests 7 Projects Releases 76 Packages Wiki Activity Actions

Lane 3 — the session, start to finish (#659) #686

Closed
claude-bot wants to merge 0 commits from redesign/lane3-sessions into main
pull from: redesign/lane3-sessions
merge into: rbrooks:main
Conversation 1 Commits - Files changed -
claude-bot commented 2026-09-17 22:18:44 +00:00
Collaborator
Copy link

Refs #659. At /feedback/prototype/lane3 once deployed.

SESS-1 to SESS-7 — 166 features, drawn as one page through five lives rather than six unrelated screens: proposed while people vote on a time, confirmed, cancelled, and afterwards the Read and Review tabs, plus the notes panel.

This lane was lane 2's. #659 scoped it as "This week, the session lifecycle, Members, Settings". Settings grew to 109 features, forced the 2/2b split, and the session pages were dropped — 2b closed out Settings rather than the lane, so nothing flagged it.

The six decisions, taken before drawing

# Decision
D27 GM review is a tab, not a layer over the reading view
D28 the post-session checklist is the GM's — SE-095–SE-102 are catalogued member, so today every player reads "Recap email: Sent"
D29 the voting grid transposes on a phone
D30 one notes panel with an audience switch (provisional — drawn for you to judge)
D31 pending event corrections are persisted server-side (IS-317)
D32 the session Talk tab defers to lane 6, which owns the thread model

Plus IS-316 for the notes audience switch.

And two decisions that were invisible. D17 and D20 are recorded in the decisions table as governing SESS-1/2/3 and TABLE-1, but no Decided line appeared on any of those surfaces, unlike every other decision. Anyone reading the surface block — which is what you read when you draw it — would have missed both.

Rendering it caught six faults, two of them in the shared kit

  1. Four of five rows of the voting grid were empty. .dot has a width and height and no display, so a bare inline span collapses to nothing. Every existing caller happens to put one inside a flex row; a table cell is not one. Fixed in kit.css — this was about to bite lanes 4–8 too.
  2. A Session 14 page said "Session 13" in its own Contents. shell.js hardcoded the sub-list heading whenever the active key was session13, which was true for look.html because it only ever drew one session.
  3. Four of five rows read as colour alone — the reader's own vote said "Yes", everyone else's was a bare dot. IS-295's exact failure, with the roster directly beneath already doing it correctly.
  4. Two slots tied on seven points and only one carried the best-slot star.
  5. Ade voted on every slot while the roster beneath said she had not answered — lane 2's fault again, a count disagreeing with its own list.
  6. "Who was there" showed RSVPs on a session already played — "Maybe" for Nix, "Out" for Sister Quill — directly beneath a banner saying four voices had been matched.

Three things in the code the drawing had to work around

  • A reload silently discards a GM's corrections. Strikes, rewordings and relabels live in local state until Regenerate; SE-154's leave-page guard is the only net. D31/IS-317.
  • The recording panel has no lifecycle gate at all (SE-157) — Start and Stop stay live on a completed and a cancelled session. Drawing SESS-5 from D9 rather than today's behaviour is what makes that visible.
  • An in-progress session has no badge colour (STATUS_CLASSES has no entry), so every state is drawn with a word.

Verification

  • Catalogue rebuilt: 1,586 rows on 102 surfaces.
  • Route in app.py, line in tests/test_app.py. 89 passed.
  • Every section rendered in headless Edge and read, in both themes.

What's left of the original lane 3

Table mode and recording — TABLE-1 to TABLE-5, 153 features, of which TABLE-1 alone is 101 and is the largest surface anywhere. Its biggest question is already visible: six header controls hidden below the phone breakpoint with no replacement (TB-070).

🤖 Generated with Claude Code

Refs #659. At `/feedback/prototype/lane3` once deployed. `SESS-1` to `SESS-7` — **166 features**, drawn as **one page through five lives** rather than six unrelated screens: proposed while people vote on a time, confirmed, cancelled, and afterwards the Read and Review tabs, plus the notes panel. **This lane was lane 2's.** #659 scoped it as *"This week, the session lifecycle, Members, Settings"*. Settings grew to 109 features, forced the 2/2b split, and the session pages were dropped — 2b closed out Settings rather than the lane, so nothing flagged it. ## The six decisions, taken before drawing | # | Decision | |---|---| | **D27** | GM review is a **tab**, not a layer over the reading view | | **D28** | the post-session checklist is the **GM's** — `SE-095`–`SE-102` are catalogued `member`, so today every player reads "Recap email: Sent" | | **D29** | the voting grid **transposes** on a phone | | **D30** | one notes panel with an **audience switch** (*provisional — drawn for you to judge*) | | **D31** | pending event corrections are **persisted server-side** (`IS-317`) | | **D32** | the session Talk tab **defers to lane 6**, which owns the thread model | Plus `IS-316` for the notes audience switch. **And two decisions that were invisible.** D17 and D20 are recorded in the decisions table as governing `SESS-1/2/3` and `TABLE-1`, but **no `Decided` line appeared on any of those surfaces**, unlike every other decision. Anyone reading the surface block — which is what you read when you draw it — would have missed both. ## Rendering it caught six faults, two of them in the shared kit 1. **Four of five rows of the voting grid were empty.** `.dot` has a width and height and *no `display`*, so a bare inline span collapses to nothing. Every existing caller happens to put one inside a flex row; a table cell is not one. Fixed in `kit.css` — this was about to bite lanes 4–8 too. 2. **A Session 14 page said "Session 13" in its own Contents.** `shell.js` hardcoded the sub-list heading whenever the active key was `session13`, which was true for `look.html` because it only ever drew one session. 3. **Four of five rows read as colour alone** — the reader's own vote said "Yes", everyone else's was a bare dot. `IS-295`'s exact failure, with the roster directly beneath already doing it correctly. 4. **Two slots tied on seven points** and only one carried the best-slot star. 5. **Ade voted on every slot while the roster beneath said she had not answered** — lane 2's fault again, a count disagreeing with its own list. 6. **"Who was there" showed RSVPs on a session already played** — "Maybe" for Nix, "Out" for Sister Quill — directly beneath a banner saying four voices had been matched. ## Three things in the code the drawing had to work around - **A reload silently discards a GM's corrections.** Strikes, rewordings and relabels live in local state until Regenerate; `SE-154`'s leave-page guard is the only net. D31/`IS-317`. - **The recording panel has no lifecycle gate at all** (`SE-157`) — Start and Stop stay live on a completed *and* a cancelled session. Drawing `SESS-5` from D9 rather than today's behaviour is what makes that visible. - **An in-progress session has no badge colour** (`STATUS_CLASSES` has no entry), so every state is drawn with a word. ## Verification - Catalogue rebuilt: **1,586 rows on 102 surfaces**. - Route in `app.py`, line in `tests/test_app.py`. **89 passed.** - Every section rendered in headless Edge and read, in both themes. ## What's left of the original lane 3 Table mode and recording — `TABLE-1` to `TABLE-5`, **153 features**, of which `TABLE-1` alone is 101 and is the largest surface anywhere. Its biggest question is already visible: six header controls hidden below the phone breakpoint with no replacement (`TB-070`). 🤖 Generated with [Claude Code](https://claude.com/claude-code)
Catalogue prep for lane 3, the session lifecycle -- which #659 scoped into
lane 2 ("This week, the session lifecycle, Members, Settings") and which lane
2 never drew: Settings turned out to be 109 rows, forced the 2/2b split, and
the session pages were quietly dropped. SESS-1 to SESS-7 is 165 rows and no
mockup touches any of it.

Six decisions from the owner, recorded in the intro and on the surfaces they
govern:

  D27  GM review is a tab, not a layer over the reading view
  D28  the post-session checklist is the GM's, not every member's
  D29  the voting grid transposes on a phone
  D30  one notes panel with an audience switch (provisional)
  D31  pending event corrections are persisted server-side
  D32  the session Talk tab defers to lane 6, which owns the thread model

D28 is the D14 shape again: SE-095 to SE-102 are all catalogued `member`, so
every player reads "Recap email: Sent" and "Journal entry: Queued". Nobody
chose that; it is what the page does.

Two of them create real scope, so they get rows rather than only a drawing:

  IS-316  one notes panel with an audience switch (SESS-7), using #593's
          single audience model instead of a fourth bespoke one
  IS-317  pending event corrections survive a reload (SESS-4) -- today they
          are local state only, and a GM correcting a long transcript can
          lose the lot to a stray refresh

## Two decisions were invisible to anyone drawing these surfaces

D17 (one time, the reader's own) and D20 (no explanatory callouts) are
recorded in the intro table as governing SESS-1/2/3 and TABLE-1, but no
"Decided" line appeared on any of those headers -- unlike every other
decision, which does carry one. Someone reading the surface block, which is
what you read when you draw it, would have missed both. Added.

Catalogue rebuilt: 1,586 rows on 102 surfaces.

Refs #659.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
feat(feedback-site): lane 3 — the session, start to finish (#659)
All checks were successful
CI / Bot/backend version sync (pull_request) Successful in 59s
Details
CI / Backend lint (ruff) (pull_request) Successful in 1m2s
Details
CI / Feedback site tests (pull_request) Successful in 1m14s
Details
CI / Summarisation accuracy eval harness (stub provider) (pull_request) Successful in 1m33s
Details
CI / Bot tests and audit (pull_request) Successful in 2m39s
Details
CI / Frontend tests, audit, and build (pull_request) Successful in 3m7s
Details
CI / Docker image build (pull_request) Successful in 5m4s
Details
CI / Backend migration, tests, and audit (pull_request) Successful in 12m20s
Details
CI / Synthetic session harness (no GPU, no LLM) (pull_request) Successful in 13m44s
Details
24239e2de5 
SESS-1 to SESS-7, drawn as one page through five lives rather than six
unrelated screens: proposed while people vote on a time, confirmed,
cancelled, and afterwards the Read and Review tabs, plus notes. 166 features.

Applies the six decisions taken before drawing (D27-D32) and the two that
were recorded in the catalogue but never written onto the surfaces they
govern (D17, D20).

## Rendering it caught six faults, two of them in the shared kit

1. Four of five rows of the voting grid were empty. `.dot` has a width and a
   height and no `display`, so a bare inline span collapses to nothing --
   every existing caller happens to put one inside a flex row, and a table
   cell is not one. Fixed in kit.css, so the next lane that reaches for a dot
   outside a flex parent does not lose an evening to it.
2. A Session 14 page said "Session 13" in its own Contents: shell.js
   hardcoded the sub-list heading whenever the active key was `session13`,
   which was true for look.html because it only ever drew one session. It now
   prefers what the caller passes.
3. Four of five rows read as colour alone -- the reader's own vote said
   "Yes", everyone else's was a bare coloured dot. IS-295's exact failure,
   with the roster directly beneath already doing it correctly.
4. Two slots tied on seven points and only one carried the best-slot star.
5. Ade voted on every slot while the roster beneath said she had not
   answered -- lane 2's fault again, a count disagreeing with its own list.
6. "Who was there" showed RSVPs on a session already played -- "Maybe" for
   Nix, "Out" for Sister Quill -- directly under a banner saying four voices
   had been matched. roster() answers who said they would come, which is the
   right question on Sunday and the wrong one on Tuesday; the after views now
   use an attendance list.

## Three things in the code the drawing had to work around

- A reload silently discards a GM's event corrections (D31/IS-317).
- The recording panel has no lifecycle gate at all (SE-157): Start and Stop
  stay live on a completed and on a cancelled session. D10 settles it, and
  drawing SESS-5 from D9 rather than from today's behaviour is what makes it
  visible.
- An in-progress session has no badge colour (STATUS_CLASSES has no entry),
  so every state is drawn with a word, which IS-295 wanted anyway.

SESS-6 is deliberately not drawn (D32): it reuses CAMP-19/20's thread model,
which lane 6 owns and has not designed.

Route in app.py, a line in tests/test_app.py, both run by CI. 89 passed.
Rendered and read in both themes.

Refs #659.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
claude-bot commented 2026-09-17 23:49:25 +00:00
Author
Collaborator
Copy link

Merged to main in 6334b4f, deployed, and reviewed with no changes asked for.

Closing by hand; the API merge returns 422 on this repo even with CI complete and mergeable: true.

D30 is no longer provisional — the notes panel with the audience switch was drawn to be judged and was accepted on review. Settled in #687 so it doesn't reach the rebuild as an open question that had already been closed.

Merged to `main` in 6334b4f, deployed, and **reviewed with no changes asked for**. Closing by hand; the API merge returns 422 on this repo even with CI complete and `mergeable: true`. **D30 is no longer provisional** — the notes panel with the audience switch was drawn to be judged and was accepted on review. Settled in #687 so it doesn't reach the rebuild as an open question that had already been closed.
claude-bot closed this pull request 2026-09-17 23:49:28 +00:00
All checks were successful
CI / Bot/backend version sync (pull_request) Successful in 59s
Required
Details
CI / Backend lint (ruff) (pull_request) Successful in 1m2s
Required
Details
CI / Feedback site tests (pull_request) Successful in 1m14s
Required
Details
CI / Summarisation accuracy eval harness (stub provider) (pull_request) Successful in 1m33s
Required
Details
CI / Bot tests and audit (pull_request) Successful in 2m39s
Required
Details
CI / Frontend tests, audit, and build (pull_request) Successful in 3m7s
Required
Details
CI / Docker image build (pull_request) Successful in 5m4s
Required
Details
CI / Backend migration, tests, and audit (pull_request) Successful in 12m20s
Required
Details
CI / Synthetic session harness (no GPU, no LLM) (pull_request) Successful in 13m44s
Required
Details

Pull request closed

This pull request cannot be reopened because the branch was deleted.
Sign in to join this conversation.
Reviewers
No reviewers
Labels
Clear labels   
backend
Issue related to the backend

  
bot
Issue related to the bot

  
bug
Something is not working

  
duplicate
This issue or pull request already exists

  
enhancement
New feature

  
frontend
Issue related to the frontend

  
help wanted
Need some help

  
invalid
Something is wrong

  
question
More information is needed

  
recording
Issue related to the recording system

  
security
Security hardening / vulnerability work

  
wontfix
This won't be fixed

No labels
backend
bot
bug
duplicate
enhancement
frontend
help wanted
invalid
question
recording
security
wontfix
Milestone
Clear milestone
No items
No milestone
Projects
Clear projects
No items
No project
Assignees
Clear assignees
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
rbrooks/Quest-Board!686
No description provided.