feat: ai@2026-07-03-beta schema (affected category) + power-flow prototype - #314
Open
pichot wants to merge 20 commits into
Open
feat: ai@2026-07-03-beta schema (affected category) + power-flow prototype#314pichot wants to merge 20 commits into
pichot wants to merge 20 commits into
Conversation
Add the "People affected" category (canvas label "Used on") — the counterpart to `accountable` that names who a system is used on or impacted by. Required, with a `relationship` element_context discriminator (subject / bystander / community) and a generic `people` seed element so `required: true` is satisfiable. Shift downstream category `order` values to open the slot, and gate the new beta in CI. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Capture prior art (Bovens accountability theory, W3C PROV/ODRL, Net-Map/RACI power-mapping, graph-vs-list comprehension evidence) for reshaping a DTPR datachain into a typed relationship graph — the grounding for the "affected" category and the power-flow prototype.
Standalone HTML explorations of a "power flow" graph view that renders who does what to whom over a DTPR datachain (index, canvas, boardgame).
|
Too many files changed for review. ( Bypass the limit by tagging |
Add canvas-affected disclosure-board prototypes (incl. v5 with a neutral, non-alarming "Used on" seat), a boardgame variant, and a boardgame.io-based power-flow game with content, logic, UI, and tests. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Fork canvas-affected-v5 into v6: replace the three uncohering per-seat
value treatments (title / instance / qualifier tags) with one generative
composition model. Every board piece renders from a typed source
{ icon, type, instance, classification, facts } through a per-category
GRAMMAR registry, at two densities — compact stack (A) and composed
sentence (C).
- Marker/colour primitives: mk() highlighter wraps every deployment fill;
colour spent only on the six published schema classifications (PII +
autonomy). Autonomy underline retired; the system-sentence autonomy
phrase now renders as a coloured marker.
- A-stack renderer: headline (lead slot) / type·classification / facts;
data/people/org lead with the specific, processing degrades to
headline+type when it has no classification.
- C-sentence renderer: per-category, per-locale templates; instance in
parens, classification em-dash set-off, facts joined via Intl.ListFormat;
A-only categories (risk) skip C.
- Risk grammar: harm type leads, required narrative, optional mitigation as
a paired response; missing mitigation raises the one earned alarm; list of
pairs; seat omitted when empty.
- EN/FR toggle: four systems re-authored per-locale; Intl number/list
formatting; no runtime inflection.
Verified in-browser: no console errors, AE1-AE5 all pass (fr data C matches
'Biométrie (votre visage) — données identifiables — conservées 24 h et sur
le cloud d'un fournisseur'; label localizes to 'Atténuation').
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Every DTPR icon (purpose, org, data-flow, processing, risk, people, mode, and rights icons) now carries a localized data-tip and shows a title popup instantly on hover — no native-title delay. A single delegated mouseover/mouseout handler on document positions a shared tooltip above (or below, near the top edge) the icon and survives every locale re-render. Tips localize with the EN/FR toggle. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add facilitation canvas/workshop prototypes (v1-v4) and their plan docs. Split the risks_mitigation element_variables into separate required `risk` and optional `mitigation` fields. Move the affected-seat relationship onto a PII-style badge in canvas-affected-v5. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
New root-level canvas/ app for the DTPR AI register + feedback layer. Mirrors dtpr-ai/docs-site: cloudflare-module Nitro preset under $production, wrangler.jsonc with a FEEDBACK_DB D1 binding (distinct from @nuxt/content's DB), @nuxt/ui + Tailwind v4, @nuxtjs/i18n en/fr. Registered in pnpm-workspace.yaml with root dev/build/deploy/test scripts. Boots in dev (200) and builds for Cloudflare (.output emitted). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Lift v6's SYSTEMS/PII/REL into typed, versioned, bilingual in-repo modules keyed by (system, variant, version) with a live flag and stable seat keys. Registry (SYSTEMS) plus resolvers: resolveCanvas defaults to the live variant + current version, liveCanvases surfaces only live variants for the register. Registry integrity + resolution covered by vitest; canvas-data typechecks under strict. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Port v6's grammar/mk/clsMarker/renderPiece/sentence into a pure grammar.ts (structured marks + segments, not HTML strings) plus Vue components (CanvasBoard, PieceStack, Marker) and canvas.css. Preserves the v6 marker/colour policy — colour only on PII + autonomy, neutral for relationships/harm — and the immediate hover tooltip. Every seat wrapper carries a stable data-seat key for per-piece feedback anchoring (U6). Grammar characterized by vitest (colours, tiers, degradation, sentence); verified live rendering the face-gates canvas with correct markers. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
D1 schema (respondent + feedback_response, indexed on the canvas keys) and two Nitro routes: POST /api/feedback (validate + persist) and GET /api/feedback/summary (segmented aggregate feeding compare). The data layer (server/utils/db.ts) is pure and D1-typed so vitest exercises the real SQL through a better-sqlite3 D1 shim — covering seat vs canvas scope (AE4), public/professional never merged (AE3), self-tag set once, and 400 on invalid input. Routes reach FEEDBACK_DB via the Cloudflare env; bundle cleanly into .output. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
useRespondent composable: a localStorage respondent_id generated once and stable across reloads, a one-time public/professional self-tag, and optional contact (never required). Pure storage helpers are vitest- covered for AE1 — first visit creates+persists the id and leaves the tag unset, a return visit reuses it, the tag is asked exactly once, contact stays optional. SelfTag.vue presents the one-time choice + optional contact (wired into the feedback flow in U6). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
CanvasFeedbackLayer wraps CanvasBoard and delegates seat clicks to a SeatReact popover (clear/confusing/unsure + optional note); ClarityRating covers canvas-level overall clarity. useFeedback posts the full tag (system, variant, version, seat?, scope, respondent) and gates the first submit from an untagged respondent behind the one-time SelfTag. Verified live against workerd + local D1: seat and canvas POSTs persist, an invalid reaction is 400, and the summary keeps public/professional segmented (AE3) with seat-vs-canvas scope (AE4). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Register index lists live variants only (R16), each linking to its SSR canvas page. The s/[system]/[[variant]]/[[version]] route resolves keys to a canvas (defaulting to the live variant + current version), renders CanvasBoard + the feedback layer, and emits OG/Twitter/canonical meta via useHead so LinkedIn/newsletter deep links preview (R15). Unknown keys 404. A default layout adds the EN/FR toggle driving both chrome and canvas content. Verified live: register lists 4 systems, deep link shows OG tags + 10 feedback seats, specific variant/version resolves, unknown system/version 404. share-meta is pure + vitest-covered. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Per-system compare: fetches /api/feedback/summary and aggregates into (variant, version) columns via a pure, tested aggregator. Each column shows overall clarity segmented public/professional (never merged, AE3), per-seat confusion ranked by where it clusters, and a thin-sample flag below the response threshold; a restyle appears as its own column while the prior version persists (AE2). VariantColumn + SeatConfusion components; a compare link is surfaced on each canvas page. Verified live against local D1 with segmented feedback; unknown system 404s. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Post-implementation simplify (3-reviewer pass): drop unused PEOPLE_HEX_PATH/PEOPLE_SYMBOL exports (CanvasBoard inlines the SVG) and the unreferenced relationshipLegend/relationshipLabel + REL tail/legend fields; extract the duplicated D1-binding lookup + 503 guard into a shared requireFeedbackDb server util; memoize per-risk riskView() once per render instead of 4× per binding; trim unused composable returns. Behavior preserved — 46 tests pass, build clean, live D1 roundtrip and risk rendering re-verified. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…, coverage Correctness + security + testing findings from review: - validateFeedback enforces per-field length caps (keys/id <=128 hard reject; note<=2000, contact<=256 truncated) so an anonymous public POST can't persist multi-megabyte values into D1 (storage-abuse guard). - feedback.post wraps insertFeedback so a D1 error returns a shaped 500 instead of leaking internals. - respondent contact is now fill-once (COALESCE(existing, new)) — a return visit fills an empty contact but never overwrites one, since the id is a client-supplied key. - register list keys by full (system,variant,version) coords, not systemKey alone, so a system with >1 live variant renders correctly. - Added coverage: length caps + contact fill-once (real SQL), useFeedback self-tag gating/flush (R12), reactive useRespondent forSubmit, resolve- layer currentVersion/defaultVariant fallbacks (AE2/R16), multi-mode sentence; removed a trivially-true assertion. 55 tests pass; hardening re-verified live against local D1. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Replace per-system compare page with a single CanvasMatrix compare view, add Tag component and shared rightActions util, and extend grammar with system-data updates and expanded test coverage. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…-canvas # Conflicts: # pnpm-lock.yaml
Deploying with
|
| Status | Name | Latest Commit | Updated (UTC) |
|---|---|---|---|
| ✅ Deployment successful! View logs |
dtpr-ai | 0fe6014 | Jul 22 2026, 01:21 PM |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Introduces the
ai@2026-07-03-betaschema version, whose headline change is a new required "affected" category ("People affected", canvas label "Used on") — the counterpart toaccountablethat names who a system is used on or impacted by, carrying arelationshipdiscriminator (subject / bystander / community), a genericpeopleseed element sorequired: trueis satisfiable, and shifted downstream categoryordervalues to open the slot. CI now also validates the new beta. Alongside the schema, this branch lands the research grounding for the graph/"power flow" model (Bovens accountability theory, W3C PROV/ODRL, Net-Map/RACI power-mapping, graph-vs-list comprehension evidence) and three standalone HTML power-flow graph-canvas prototypes (index,canvas,boardgame).Validation
pnpm --filter ./api schema:validate ai@2026-07-03-betapasses (12 categories, 109 elements);schema:buildstamps a realcontent_hashand composes thepeopleicon across all three discriminator variants; a minimal instance placingpeopleinaffectedvalidatesok: true(noREQUIRED_CATEGORY_MISSING); api tests (444+64) and typecheck are green.