# Chapter 2 · Three records, one word: completed · brief · 11 September 2026

The second chapter of the _Tara · practise_ track, rewritten in the Eve edition
on 11 September 2026. It follows the plan in `EVE-V1-REDESIGN-TODOS.md` and the
track brief in [tara-practise-track.md](tara-practise-track.md). The receipt is
under `verification/`.

## The job

Chapter one ended on a screen that said completed and a record that said
completed and meant something else. This chapter is the contract and the
evaluator behind those words.

Three records use the word. `RitualSession` in the contracts validates a
lifecycle: states, an event history that has to agree with itself, and an audio
floor of eighty per cent that no template can lower. The domain evaluator has
its own vocabulary — `completed`, `completed_with_adjustments`, `ended_early` —
decided from required step counts, a required ratio and an actual duration.
`SittingCompletionEvent` is player telemetry with a timing guard of its own. No
inspected code converts any one of them into another.

Beside them runs a percentage that answers none of the three. It divides
completed estimated minutes by required estimated minutes, and under the default
optional-step policy an optional step's minutes land in the numerator and not
the denominator — so the meter can read a hundred per cent while a required step
is still unfinished.

**Running example.** A ritual of three steps: a four-minute guided practice, a
three-minute required passage and a four-minute optional journal. The practice
and the journal are complete; the passage is skipped. The meter reads 100%. The
evaluator returns `ended_early`. Recovery retains the three-minute passage and
keeps the original seven-minute duration minimum, so completing it returns
`ended_early` again.

## Three learning promises (the cover)

1. **Three records say completed** and decide it three different ways.
2. **The meter is not the decision**, and under the default policy it can reach
   a hundred per cent with required work unfinished.
3. **Recovery is a proposal**, and the one it builds can be impossible to
   complete.

## Three things to remember (the close)

1. **Name the record before reading the word.** Contract, evaluator and player
   telemetry each own one meaning of completed.
2. **Read `shouldCountAsCompleted`, not the percentage.** They answer different
   questions and the source says so.
3. **Inspect what recovery returns.** The helper spreads the original estimate
   and completion policy into a shorter template, and its message is friendlier
   than its policy.

**Bridge.** Chapter three — Tara audio state and interruption: where the audio
percentage this chapter's floor depends on actually comes from.

## Sections

Four sections of five, six, five and five.

| Section                                    | Slides                                                                                     |
| ------------------------------------------ | ------------------------------------------------------------------------------------------ |
| Three records, one word                    | the case, the three records, the session record, the transition table, the lifecycle shape |
| What the session contract checks           | the event, the history, the audio floor, player telemetry, continuation, prerequisites     |
| The evaluator, and the meter beside it     | the decision, the hundred-per-cent case, the denominator, the skip, recovery               |
| What recovery proposes, and what it proves | the retained minimum, the completion event, the projections, the resume link, the close    |

## Treatment decisions

- **One table, kept deliberately.** `tara-session-transitions` is the contract's
  own `SESSION_STATE_TRANSITIONS` map, printed as its eight declared rows rather
  than the six the inherited slide grouped them into — `started` and `resumed`
  are separate keys in the source, and so are the two terminal states. The plan
  allows a table exactly where exact values must be compared, and this is that
  case. It is the chapter's only one, in twenty-seven slides.
- **The transition table and the state machine are a pair.** The table has the
  exact edges; `tara-session-return-routes` has the shape, with `started` and
  `resumed` drawn as one box and the three interruptible states as another, each
  naming the contract states it stands for in its `states` field.
- **Two plan-listed tables became card grids.** The player-event timing cases
  and the three optional-step policies are short enough to read as cards, and
  keeping them as tables would have put three in twenty-seven slides.
- **`tara-session-audio-floor` became a compare panel.** Its rows already asked
  one question twice, of `completed` against `partially-completed`.
- **The chapter's one capture rides on a case study.** This chapter has no
  surface: it is contracts and library functions. The composition contract has
  always allowed a capture beside a case, and no delivered slide had used it, so
  `tara-session-percent-case` is the first. It carries the practise track's own
  pinned player crop, because the only percentage a member currently sees is the
  player's elapsed-over-planned figure — a third percentage, and not the one the
  slide is about.

  Being first through that path found two things. The `.case-capture` image rule
  set a full width with a capped height, so the frame drew wider than the
  picture inside it; the width follows the height now and the border sits on the
  image. And a composition guard asserted that no delivered case study renders a
  capture, which is exactly the regression it was written to catch — it names
  this slide as the one deliberate exception now, and checks the rest both ways.

## What this chapter does not establish

No persistence. Every function here returns data; none of them saves, delivers
or notifies, and the chapter says so on each slide rather than once at the end.
No authorization: an actor field is a string, a memory scope is a declared
purpose, and a resume token is routing context rather than a credential. No
deployed consumer for the completion projections. And no clinical or personal
claim: the ritual, its steps and its durations are illustrative evaluator input.
