This package addresses the V1 member-room depth in
the full V1 audit and TODOS.md sections 7 and 10. Its
audience is product, design and engineering readers who have read tara. The
outcome is to distinguish session history, step completion, player telemetry,
continuation and an actual saved return. The new guide is
tara-ritual-sessions, immediately after tara. The existing taxonomy and
continuation overview notes have the precise corrections described below.
Implementation baseline: 4d893c29bb7a24f7c1b1dde051ae4d075db0d6ae. Primary
source revision: 1ceae52c7896a59ac74a4429d410f3218f0b91c6. The reviewed Tara
contracts and domain files are byte-identical between those revisions. Read
complete relevant source units before making reviewed teaching assignments. No
full Tara, player, scheduling, mentor or workbench acceptance is claimed by this
package.
Visual thesis: a calm, precise field guide to an interrupted practice, using cream paper, readable type and restrained terracotta emphasis. The visual anchors are actual state relationships, record anatomy, decision gates and worked numbers. Avoid decorative meditation imagery when the reader needs to understand which record changed.
Content plan: begin with a three-step practice interrupted before a required passage; separate the records; explain lifecycle and event validation; compare completion thresholds; inspect continuation; calculate domain completion; follow skip and recovery; then distinguish projection and device-link evidence from persistence. Close by resolving the opening case. Briefs travel with every canonical slide. Tables use meaningful headings and no positional emphasis.
Interaction thesis: preserve existing reduced-motion-aware entrance, source and transcript reveal, navigation and audio controls. Every relationship remains complete in static HTML and PDF. No new shared UI behavior is needed.
Source reconciliation#
libs/contracts/src/tara/index.ts: session state, template/step/completion schemas, continuation, event records, history validation and sitting telemetry. The feature page says completion is per-template; the session schema also imposes an 80% floor on both completed states, with partial below 100%. A separate player event requires 85% when planned duration is supplied.- History validation checks adjacent states, legal transitions, event-type consequences and final state. It seeds from the first supplied prior state; it does not require a complete history beginning at not-started or timestamp sorting. It checks completed/partial consequence flags but does not recompute stateChanged. A source fixture caught that distinction in the initial draft; the explanation and positive characterization were corrected. Parsing does not itself persist anything.
- A blocked continuation recommendation must name actual unsatisfied entries. The satisfied branch only requires an empty blocked list; it does not compute readiness from the prerequisites array. Governance metadata is not an access decision. The overview correction removes both overclaims.
- Lineage schema consistency is not proof that a consuming recommendation engine prevents every undisclosed blend. The taxonomy note now makes this explicit.
ritual-model.ts: completion uses required count AND ratio AND actual duration, plus the adapted-skip policy. Display percentage uses estimated completed-step minutes with ignored/included/bonus optional treatment. A 100% display can coexist with an incomplete required step.ritual-operations.ts: skip checks allowSkip and returns changed execution data; required skips open recovery. Recovery selects incomplete required steps, but spreads the original duration/policy before rebuilding. It does not necessarily shorten the duration threshold or the displayed estimate.ritual-completion-events.ts: completion event identity uses subject, ritual and occurredAt; projections use that object key. Search carries subject-only audience and authorization-filter version. Stable IDs and fields require enforcing stores and consumers; these pure functions do not deliver or index.practice-plan-continuity.tsanddeep-links.ts: explicit starter-plan input becomes a continuation card and links; currentDevice changes can set the readiness flag. The token is deterministic routing context, not authorization, synchronization or measured cross-device recovery.- Whole-tree
git grepat the implementation baseline finds the ritual session schema in the contract persistence registry. The named domain evaluator, scheduler, skip/recovery and completion-event builder have no non-test app call sites in that search. This is discovery corroboration, not proof of every possible dynamic integration. Teach the inspected functions as library behavior. V1/architecture/customer-domains.mdlines 107–249 were also read. Its session discussion correctly names the 80% floor, while its taxonomy and consent-aware-by-construction prose needs the same consumer qualifications. The checkpoint/recent-window details remain part of the pending player review; do not assign that whole architecture session unit from this guide alone.
Full scheduling, recurrence/quiet-hours, player/checkpoint, practice selection, teachers, companions, mentor presence and content-workbench review remain in the full V1 ledger. The scheduler was read for the recovery call boundary; its complete teaching belongs in a separate follow-up guide.
The four complete assigned feature units cover state, drift, event consequences and continuation. The feature's partial-completion unit remains unassigned: its 5% checkpoint and 12-hour recent-completion constants still need the player caller explanation. Coverage is not inferred from a page appearing in citations.
Candidate findings and resolution#
- Initial authoring used the wrong table input key (
headersrather thancolumns); the builder rejected it. The corrected 21-slide draft passed standalone desktop/compact/mobile and PDF layout checks. - Source tests now resolve the existing web Zod dependency through its package. Companion and navigation boundaries are explicit fixtures. Sixteen source characterization tests passed after setup and source reconciliation.
- A test disproved the initial stateChanged consistency assumption. Direct source inspection confirmed the missing check. The note now states the actual behavior, with an explicit accepted-edge characterization.
- The first assignment check rejected a skip slide missing the inventoried feature citation. The citation and assembled fingerprint were corrected.
- The pre-narration full guide review passed its completed layout checks, then timed out at playback because the new last track had not been generated. This is retained as incomplete. The settled capture-only review preceded narration; final delivery requires complete browser and center checks.
- The overview now uses record anatomy for peer continuation fields and unnumbered taxonomy columns. A stranded hyphen in the large recovery title was removed by rewriting the claim. Final optical review also shortened three record labels, followed by a fresh affected browser/PDF review.
Delivery and acceptance#
Author a draft, review the rendered candidate, then publish canonical chapter, catalog, HTML, PDF, narration, reviewed source assignments and a retained receipt. Correct only the two identified overview explanations; preserve other guides and recordings. Add source fixtures for contradictory completion, history, prerequisite and recovery cases. Exercise the existing full affected-guide Playwright and center checks, including mobile, source/transcript, keyboard, axe, playback and PDF; inspect every new final slide and PDF page optically.
Use the repository guide-review configuration and source-map interfaces, supervised sequential generation and the Python 3.11 narration environment. Run builder/coverage freshness and relevant presentation contracts once the manuscript settles. Retain failed attempts and rerun only invalidated scope. Normal-speed human listening and representative-reader acceptance remain open; automated audio decoding and model visual review cannot substitute for them.
Keep partial work in drafts/; preserve valid audio on interrupted generation.
Stop only owned processes. After scoped checks pass, commit with normal hooks,
push to origin/main, integrate remote changes if needed, and verify the branch
tip is published. This increment cannot close the full V1 goal.
Delivered checkpoint#
The frozen receipt
records the final 21-slide guide, two corrected overview slides, 32.2 minutes of
new speech, two revised recordings, both PDFs and four complete reviewed feature
units. All 223 final browser checks and fourteen focused center checks pass;
source fixtures and relevant presentation tests pass. Every new desktop slide,
the two overview changes and both complete PDF sequences were optically
reviewed. All 618 other recordings and 61 other guide HTML files are preserved.
Upstream 336013f511c0b4dcebfe47b50f2002b4f0f62798 was integrated without
changing the inspected presentation or Tara source inputs. Scheduling and
practice-surface discovery now have separate records for the next work.
The publication push then encountered the Eve dependency assignment at
0ec272ddef60f53096a8ad99584dab28c82cfb66. All three incoming slides and tracks
are integrated. Tara HTML, manuscripts and all 641 pre-integration recordings
remain exact bytes of the inspected candidate. All 31 PDF page renders also
remain identical after the final export. The receipt explicitly inherits the
unchanged Tara layout matrix and records fresh merged center verification,
scoped incoming-track integrity and 34 passing preservation and assignment
tests. Current totals are 644 slides and 80 reviewed source units; full V1
coverage remains incomplete.