Let a user opt a connector token into logging. The settings card gains a
write toggle; the mint endpoint accepts a closed scope choice (read, or
read_write) and builds the permission array explicitly — it can mint only
[health:read] or [health:read, health:write], never a wildcard or any other
grant. The OAuth flow grants health:write only when a client requests it, and
the consent screen states in plain language that the app will be able to read
AND log/write health data when write is requested; the token endpoint grants
exactly the approved scopes and never widens. Read-only behaviour is
unchanged. Add the English and German strings for the new toggle and warning.
Register two write tools — log_measurement and log_mood — only for a
health:write-scoped session; a read-only session never sees them advertised.
Each write mirrors the proven Telegram capture path (same range guard,
canonical unit, cache invalidation and rollup recompute) pinned to the MCP
source, and is idempotent on a hashed externalId derived from a client key, so
a retried call is a no-op rather than a duplicate.
Because the remote wire is stateless (no elicitation), the tools carry a
protocol-level confirm gate: without confirm:true they write nothing and
return the exact record they would write plus an instruction to confirm the
value with the user and re-call with the same idempotency key. Commits run
under a tighter per-credential write budget on top of the per-binding request
limiter.
Harden the resource-server audience binding: a token whose scopes are all in
{health:read, health:write} is bound to the MCP surface and refused on every
REST write/delete. A health:write token therefore admits the in-process MCP
write tools only and can never become a general REST write credential; admin
stays unreachable (cookie-only). Refuse, by construction, any medication, lab,
delete/update, or data-export tool on this surface.
Append an MCP value to MeasurementSource (migration 0196, additive enum
extension) so a row logged through the remote MCP endpoint is distinguishable
from MANUAL/TELEGRAM provenance and gets an isolated idempotency namespace
(userId, type, MCP, externalId). The value joins the read enum so existing
read/response shapes decode the rows; the REST batch and single-POST write
surfaces still reject it (server-resolved only). Regenerate the OpenAPI
contract for the additive enum.
Expose four read-only relational tools and one prompt over the MCP wire,
each a thin re-export of an existing server-authoritative engine — no new
statistics, no new business logic:
- get_correlation: the FDR-controlled, lag-aware association between two
named metrics, selected out of the existing correlation discovery the
Coach and the insight route already run. Descriptive, never causal.
- compare_metric: two metrics over the same trailing window, or one metric
across two trailing windows, computed over the WMY rollup readers with a
delta only when both sides share a unit.
- get_metric_baseline: where today's reading sits against the user's own
usual range (median ± robust deviation) plus the strongest lagged driver,
reusing the metric-page baseline builder. Honest-null below the learning
floor — never a fabricated range.
- detect_changepoints: level shifts over the rollup tier via a small in-repo
CUSUM binary segmentation with a high firing bar.
- doctor_visit_summary prompt: a structured, grounded visit-prep summary
assembled from the doctor-report data path so the numbers match the
exported PDF, with the grounding framing injected once — data and context
only, no diagnosis or verdict.
Every tool returns values, units, reference bands and provenance, declares
read-only / non-destructive / closed-world annotations plus an output schema,
and reports { present: false } for absence. All reads stay session-narrowed
and read-only; the existing security posture is unchanged.
Harden the remote MCP endpoint and the in-repo OAuth bridge against the
adversarial review of the milestone branch. The two ship-blockers made the
"read-only, revocable" promise false; the rest tightened the surface.
Audience binding at the resource server: a token whose only grant is the new
`health:read` scope is the MCP read credential. The REST edge now accepts it on
safe (read) methods and refuses it on writes/deletes with 403, so a connector
token can read over `/mcp` and GET routes but can never POST/DELETE the record
despite a "read-only" consent. The `/mcp` resolver is unaffected.
Revocable, theft-detectable refresh tokens: refresh artifacts now anchor to a
persistent connection row. Revoking a connection stops every future refresh and
kills its access tokens; a replayed (already-rotated) refresh is treated as
theft and revokes the whole connection. Connections surface in settings as the
unit a user can disconnect. The old access token is revoked on rotation.
Also: fail closed on the OAuth/`/mcp` surface when no canonical origin is
configured and mark the discovery documents non-cacheable; honour the global API
kill-switch on `/mcp`, `/token`, `/authorize`, and `/register`; cap the CIMD and
registration body reads while streaming instead of after; show the verified
client origin on consent; validate CIMD redirect-URI schemes; assert same-origin
on the consent POST; and add a daily prune for expired connector tokens.
New migration 0195 adds the connection anchor and links access tokens to it.
A dedicated Settings card co-locates the off-by-default MCP module toggle
with token management for connecting an external assistant:
- Mint / list / revoke a dedicated `health:read`-scoped token (shown once,
copy to clipboard) via new `/api/mcp/tokens` routes — never the
`medication:ingest` scope, never the wildcard.
- Surface the connector URL and the enable switch in one place, wired to
the existing module-preference endpoint.
- New section wired into the Settings shell next to API & Tokens, with
EN + DE strings (and the other locales) under `settings.mcp` and
`settings.sections.mcp`.
- Add the `search` + `fetch` retrieval pair with the exact ChatGPT wire
shapes (`{id,title,url}[]` and `{id,title,text,url,metadata}`), backed by
the same server-authoritative reads and returning real deep-link URLs into
the web app for citations. Without these the server is invisible in
default-mode ChatGPT.
- Annotate every read tool `readOnlyHint`/`destructiveHint:false`/
`openWorldHint:false`; structured tools also declare an `outputSchema` and
return `structuredContent`. Omitting the hints is a validation error in the
connectors.
- Reject a cross-origin browser request to `/mcp` with 403 (DNS-rebinding
defense) and point the 401 `WWW-Authenticate` at the resource metadata.
Let a remote MCP client (Claude.ai, ChatGPT) self-register and connect to
the `/mcp` endpoint over OAuth 2.1 + PKCE, bridged onto the existing `hlk_`
Bearer model — no parallel identity store.
- Discovery: RFC 9728 Protected Resource Metadata and RFC 8414 Authorization
Server Metadata at `/.well-known/oauth-protected-resource` and
`/.well-known/oauth-authorization-server` (added to the proxy allowlist).
- `/api/mcp/oauth/authorize` — consent gate backed by the existing session;
a code is only issued on an authenticated POST decision, never on a GET.
- `/api/mcp/oauth/token` — authorization-code and refresh-token grants with
mandatory PKCE S256, single-use codes, and rotating refresh tokens. The
access token is an ordinary short-lived `hlk_` token scoped `health:read`.
- `/api/mcp/oauth/register` — RFC 7591 DCR fallback; CIMD (SEP-991) is the
preferred path and resolves client metadata through `safeFetch`.
- Audience binding (RFC 8707): every token is bound to the canonical `/mcp`
resource URI; a mismatched `resource` is rejected at authorize and token.
- The grant artifacts (code, refresh, DCR client id) are HMAC-signed,
self-describing, and domain-separated; the central log redactors scrub
them plus the PKCE verifier.
Bridge the existing transport-agnostic MCP server factory to a Next.js
route at /mcp using the SDK's Web-standard Streamable HTTP transport, so a
remote client can connect with a manually-minted hlk_ Bearer token. The
same tool/resource layer the local stdio transport serves is reused
verbatim — no duplicated tool logic.
The route owns four gates in front of the transport:
- Auth: Authorization: Bearer hlk_… resolved by the same validator the
HTTP edge uses, so revoked/expired/scope semantics never drift. A cookie
is never consulted, so the cookie-only admin boundary stays unreachable
here regardless of token scope.
- Rate limit: the Postgres limiter keyed by the <userId>:<tokenId> binding
(mcp:<binding>), so a single shared credential cannot drain the account.
- Module gate: a new opt-in mcp module, OFF by default. With it off the
endpoint answers 404 and reveals nothing — a self-hoster who never opts
in gets zero new exposure. The Modules settings hub renders the toggle.
- Transport: a fresh stateless server + transport per request.
Read-only by construction (only read tools are registered) and same-origin
by construction (no CORS headers). The proxy lets /mcp skip the page cookie
redirect since it authenticates by header, mirroring the webhook surfaces.
Introduce a read-only Model Context Protocol surface built on the official
@modelcontextprotocol/sdk so a user can query their own HealthLog data from an
MCP-capable assistant over local stdio with a pasted, scoped Bearer token — no
extra infrastructure.
The tool and resource layer in src/lib/mcp is transport-agnostic: five read
tools — list_metrics, get_metric_series, get_medication_compliance, get_labs,
get_correlations — each a thin wrapper over the existing server-authoritative
retrieval paths, returning structured values, units, reference bands, and
provenance with an explicit { present: false } for absence and no prose
verdicts. Two resources expose the user's profile (data-minimised) and
medication list. A stdio entrypoint (src/cli/mcp-stdio.ts, pnpm mcp:stdio)
resolves the token through the shared Bearer path and binds the session to
<user>:<token>; with no token it refuses to start.
Read-only and admin-unreachable by construction: only read tools are
registered, and requireAdmin stays cookie-only so the Bearer/MCP wire can never
elevate. Covered by unit tests for every tool and resource, the auth/scope
model, and the cookie-only admin boundary. The server ships in the existing
container; docker compose up is unchanged.
Move the canonical Bearer-token validation out of the api-handler request
wrapper into src/lib/auth/bearer.ts so it can run outside a Next.js request
scope without dragging next/server, next/headers, or the Wide-Event machinery
into the import graph. requireAuth's Bearer path now delegates to it and keeps
the existing auth.bearer.* audit trail and Wide-Event annotation, so the
revoked / expired / scope / lastUsedAt semantics live in one place and cannot
drift between callers.
Covered by a focused unit test pinning the hash-only lookup and the
revoked / expired / insufficient-permission rejections; the existing
require-auth Bearer suite stays green.
The dedicated conversation-history page gates on the operator master
flag and the per-user opt-out like the main Coach route, so add it to
both grep-based gate-site allowlists.
Add a Settings entry to the composer's "+" actions menu and remove the
page-toolbar gear; the drawer header keeps its own gear. Let the seeded
"worth a look" opener fill the composer column and give it a per-day
dismiss control. Route the Conversations action to /coach/conversations
and drop the in-page history drawer.
Replace the left slide-in conversations drawer with a standalone
/coach/conversations page: a search field above a recency-grouped list
(Today / Yesterday / This week / Earlier) built on the existing
conversations query. Selecting a row routes to /coach?c=<id>, reusing the
deep-link open mechanism — no new API, no new state. The drawer's
Conversations affordance now navigates to the new route, and the Coach
page drops its ?view=conversations in-page drawer handling.
Mirror the about-me management routes: the PATCH (confirm/update) and DELETE
on /api/coach/plans/[id] now apply a per-user mutation limit, closing the
consistency gap with the rest of the Coach management stack. Owner-scoped and
cheap, so the limit only caps a runaway client loop.
Drop the unused extractCoachClaims helper (+ its CoachClaim types) and the
unused CoachPlanStatus type — the deterministic grader uses the boolean claim
helpers directly, and the plan status enum is consumed via the const array.
Update the dashboard-snapshot suite for the locale-keyed cell: stub the i18n
context and add getQueriesData to the fake query client so the locale-aware
preload probe is exercised.
Hotfix: the Coach and daily briefing failed for ChatGPT-provider accounts —
a provider tool-call rejection surfaced as a 500. Provider failures now
degrade gracefully, the Codex tool-call wire format is corrected, and the
provider error body is captured for diagnosis. Also: the thinking bubble
shows only the animated dots.
The pre-stream thinking bubble dropped the visible "Thinking…" word and now
renders just the three bouncing dots. The label stays in an sr-only span so
the thinking state is still announced to screen readers.
A Codex backend 400 on the richer Coach tool-call request (and the daily
briefing) propagated out of the provider chain runner unwrapped: a 4xx other
than 401/403/429 was not classified as a hard provider failure, so it bypassed
AllProvidersFailedError and rethrew from the chat route as an HTTP 500. Plain
status-card completions, which never send a tools array, were unaffected.
Harden the failure path so a provider error can never surface as a 500:
- Treat any 4xx a provider wire returns as a hard failure so it cascades and is
wrapped in AllProvidersFailedError. InsightSchemaError (the strict-JSON 422
surface) stays the only non-hard class, so that diagnostic is preserved.
- The chat route now also classifies a tagged provider error that reaches its
catch unwrapped and emits the graceful coach.provider.* SSE frame instead of
rethrowing. Only an untagged error keeps the 500 + error-reporter path.
Make the Codex 400 diagnosable: fold the redacted upstream response body into
the thrown error message and the chain wide-event, so the field the backend
rejects is visible in logs (it was previously dropped).
Fix the Codex request shape behind the tool-call 400:
- Send the required strict flag on each Responses-API function tool.
- Replay prior assistant turns as output_text, not input_text, per the
Responses API multi-turn contract — the violation only surfaced on round two
of the tool loop, which is the path that broke.
Codex-only; the OpenAI, Anthropic, local and mock providers are untouched.
A deterministic golden-set + red-team floor that gates every Coach change
offline, plus an opt-in live-judge layer that scores warmth and groundedness
with a pinned model only when COACH_EVAL_API_KEY is set.
- 42 golden cases across grounding, cross-metric, data-honesty, provider-parity
and own-baseline taxonomies, graded by behaviour-anchored binary criteria.
- Structured-claim grounding helpers extend the numeric verifier from numbers
to threshold / own-baseline / confident-verdict claims; the numeric API is
unchanged so the existing grounding tests stay green.
- Red-team battery asserts the existing inbound refusal, outbound dose screen,
acute red-flag escalation and learn-link guards fire.
- Nightly + manual workflow runs the deterministic floor always; the live judge
step runs only when the secret is present and is non-blocking.
Add a CoachPlan model: durable, user-scoped if-then implementation plans
the Coach proposes and the user confirms. The background memory-refresh
pass extracts plans the user agreed to and writes them as "proposed"; only
the user-facing PATCH on /api/coach/plans/[id] activates one. The free-text
cue / action / target are encrypted at rest with the shared Bytes codec;
metric, status and dates stay plain so the injection picker ranks without a
per-row decrypt. Active plans fold into the Coach snapshot's memory block
and survive the prompt-budget drop alongside facts. List, confirm and
soft-delete routes are Coach-gated and narrow the owner from the session.
Key the dashboard-snapshot query cell by locale so a locale switch reads
freshly localised tile copy on its own cell instead of serving the prior
locale's prose until the slot goes stale. The zero-arg key stays
byte-identical, so every prefix-based invalidation keeps reaching the
locale-keyed cells; the offline-banner and preloader match on the prefix.
Migration 0194 (additive CREATE TABLE). Encryption-rotation registry and
script cover the three new Bytes columns.
The A1–A3 source files carried v1.22.0 docstrings from before the work was
re-cut as the v1.21.2 point release. Comment-only sweep so blame and the
in-file version tags read consistently with the shipped version.
The "the Coach is already on …" pill resolved its metric label through the
brand-free English domain vocabulary, so it read in English on every
non-English locale. It now prefers the existing localised measurements.type*
name for each measurement-backed source, keeping the English phrase only for
the gait/audio long tail that rarely becomes a scoped launch. Also rewords the
English return-to-baseline line so the day count reads as the in-range run
length rather than the time since.
The builder is server-only by construction — only the metric-page route
imports it and it reads Prisma — and no other module in the tree depends on
the `server-only` package, so the import was both redundant and an undeclared
dependency. Removing it keeps the dead-code gate clean without weakening the
boundary.
The metric-page "Coach read" strip and the unscoped-hero seeded opener are
Coach surfaces, so both routes now call requireAssistantSurface("coach")
alongside the insights module gate. With the Coach turned off they return
the same refusal every other Coach route does, instead of lingering as a
"Coach read" line while the assistant is disabled. Restores the coach
route-gate inventory invariant.
The hero score card and the daily briefing accept tension, returnToBand, and
memory props, but nothing populated them server-side, so the lines never rendered.
Resolve all three server-authoritatively on the dashboard snapshot DTO (iOS rides
the same payload):
- tension — fires when the readiness composite's contributors disagree (a strongly
favourable one alongside a strongly unfavourable one), reusing the readiness
components the wellness ring grades and the shared disagreement detector.
Suppressed under a clinical red-flag (a coincident-deviation that no illness
explains) so the red-flag path is never reconciled into a calm verdict.
- returnToBand — runs the streak detector across the salient vitals and surfaces
at most one metric that has come back inside its own personal range after a
genuine prior out-of-band run.
- memory — maps the machine-derived coach memory block to the briefing's recall +
forward-look.
The blocks are locale-agnostic (contributor keys + a metric type); the insights
hero localises them through the existing readiness-contributor and metric-name
labels before handing the cards their resolved strings. Each value omits rather
than guesses. The snapshot cache key now carries the locale for the recall prose.
Add a deterministic narrative layer the Coach narrates rather than invents:
- Briefing callbacks + forward-looks: the daily briefing recalls the prior
period and points ahead, anchored to a figure, one callback per surface.
- Tension Verdict: when the readiness composite's contributors disagree
(good sleep but a rising resting pulse), surface a brief honest internal
read on the health-score card. A real clinical red-flag is never reconciled
away into a calm verdict.
- Streaks + return-to-baseline: a pure detector over a DAY-bucket series
closes a worry once a metric returns inside the personal MAD band after a
genuine out-of-band run; a gap day breaks a streak.
- Coach prompt: add develop-discrepancy, roll-with-resistance, and an
anti-persuasion clause; ratchet PROMPT_VERSION.
The tension flag and streak event are computed server-side; the card lines
render resolved fields and the model narrates them.
Replace the hidden composer prefill with a visible scope affordance: when
the Coach is launched narrowed to a metric, the hero shows a scope pill
naming that metric plus a tappable seed question, and the drawer sources
rail surfaces the same scope line. When the Coach opens unscoped, resolve
today's most notable derived signal server-side and offer it as a tappable
opener; fall back to the neutral greeting when no signal crosses the gate,
so the blank box is never cold and never fabricated.
Add an internal GET /api/insights/coach/seeded-question that reuses the
briefing's derived-signal detector to select the opener server-side; the
client renders the resolved DTO.
Render a compact two-line "Coach read" above each metric sub-page chart:
an own-baseline line ("Your usual range is X–Y; today's Z sits within /
above / below") and one top lagged-association line from discovered
correlations, framed descriptively. It makes a metric read seen-in-context
rather than isolated.
The strip is server-authoritative: a new GET /api/insights/coach-read
resolves the band via the median ± k·MAD baseline engine (with its 7-day
history floor — below it the strip says "still learning your range" rather
than inventing a range) and the strongest on-metric driver from the
existing FDR-controlled, effect-size-floored correlation discovery (no
lowered floor; the line is omitted when nothing clears it). The client
component only renders the resolved DTO, so web and iOS read identical
numbers. DTO shapes and the pure selection logic live in a client-safe
sibling shared by the component and the unit test.
Wired through the HealthKit metric scaffold (every generic metric page) and
the bespoke weight, blood-pressure, and pulse pages.
The post-hoc numeric verifier only graded the tool path, so local /
no-tools providers got no grounding net. Grade the no-tools prose
against the snapshot the model was shown — its structured sections,
which already carry the correlations-snapshot block — and flag any
cited number that traces to no snapshot leaf. Non-blocking: the count
is annotated under coach.prose.number_unverified and the unverified
token soft-stripped, never a stream-killer. Only runs when the full
figures were delivered this turn; a cheap follow-up has no fresh set
to grade and no-ops. Tool-path behaviour and tolerance unchanged.
The desktop branch of ResponsiveSheet rendered DialogContent as a single
overflow-y-auto grid, so a tall form — or one made tall by browser zoom,
display scaling, or a longer-language label — scrolled its header, body,
and footer together and could push the primary button below the fold. The
branch now runs as a flex column capped at max-h-[calc(100dvh-2rem)], holds
the header and footer shrink-0, and scrolls only the body, matching the
bottom-sheet branch that already pinned its footer. Every form mounted
through the shared surface inherits the fix.
Add a desktop end-to-end guard at a short viewport asserting the body
scrolls and the footer's bottom edge stays within the viewport.
The three intake dialogs (log, edit, dose-history add) still used a raw
datetime-local input, so they ignored the user's date-format preference.
Route them through DateTimeField like every other date surface. Also
correct a stale diastolic-floor comment in the dashboard verdict.
The correlation discovery matrix gained two non-MeasurementType channels
(medication compliance, symptom severity), but the Coach get_correlations
reader still filtered its measurement query by hand and never built the new
channels. The channel keys leaked into the Postgres enum cast and threw for any
user with medication or illness data.
Extract the route's two channel-series fetchers into a shared module
(correlation-channel-series.ts) so both consumers build the channels identically,
then wire the reader to derive its type filter from discoveryMeasurementTypes and
fold the fetched compliance + symptom series into the discovery input. The reader
now humanises the channel keys to "medication adherence" / "symptom severity".
Render daily medication-compliance rate and illness symptom-severity as
first-class series in the FDR discovery engine, so the adherence-dip →
symptom-flare link (and compliance → vital drift) can finally surface.
Compliance pools every active medication's dose-history ledger into a per-day
taken/(taken+missed) rate, re-keyed to the user's display timezone. Symptom
severity reads the illness day-log functional impact, zero-filling healthy days
only across real episode spans so a user with no episodes yields an empty
series rather than a constant. Both flow through the existing n >= 20, BH-FDR,
effect-size, and shrinkage gates unchanged; a sparse series degrades to absent.
Each channel forms its own metric family (no self-lag tautology) and keeps the
association-not-causation framing. Add discoveryMeasurementTypes() so every
caller drops the non-measurement channels before a Prisma type filter.
The labs block already surfaces value + unit + range + date, and the Coach
prompt instructs it to quote them verbatim, so labs are cited end-to-end — no
gap. Pin that with a tool-level guard asserting value and unit survive
get_labs, so a future change to the labs pass-through can't quietly strip the
number the Coach needs to cite.
The Coach may point at a published /learn guide, but only a slug in the
catalog. That was a prompt instruction with no enforcement — a model could
hallucinate /learn/lower-your-cortisol and ship a dead link. Add a deterministic
post-filter on the assembled reply that keeps a published slug verbatim and
strips any reference whose slug is not in the catalog, tidying the prose around
the removed link. Makes the catalog's impossible-by-construction claim an
enforced guarantee.
The illness-recovery tool and the illness context block carried only labels,
lifecycle, and dates plus the recovery composite — never the recovery-gap, the
metric that drove it, the nadir, the pre-onset deviations, or the red flags the
illness card renders. So a recovery question got the composite, not the gap the
user sees, and a sustained-fever or low-SpO2 escalation was invisible
in-conversation.
Run the existing episode-correlation read-layer (the same engine the card and
the red-flag notifier use) for the most relevant episode — active first, else
the most-recently-resolved — and attach the computed scores to get_illness_recovery,
coverage-gated so a thin signal yields nothing rather than a fabricated number.
Document the illnessScores shape in the Coach prompt so it restates those
numbers verbatim and escalates a red flag rather than reassuring.
Each retrieval tool re-scoped the snapshot to a single source, so its cache key
never matched the inventory's full-source build — a four-tool turn paid four
extra full snapshot builds on top of the inventory's. Thread the inventory's
probe scope to every tool so a per-tool read lands the one cache entry the
inventory already primed, collapsing the turn to a single build (a window
override still gets its own correct build). Raise the loop's round cap from two
to three so a sequential cross-metric why-chain isn't starved; the per-round
token budget is unchanged and the absolute ceiling tracks one above so the
final round still forces prose.
The tool-mode path dropped the launch scope's sources: the inventory probes the
full source set, so a Coach opened from a metric page saw the whole inventory
and could roam, while only the no-tools path honoured the narrowing. Inject a
one-line FOCUS hint naming the launched domain(s) into the tool-mode user turn
so the model prioritises that metric and fetches its figures first, with every
other domain still reachable if the question leads there. Empty on a generic
open, so the prompt prefix is unchanged on that path.
The per-bucket OLS accumulators stored x on the raw epoch-day axis (x ~ 20540),
so sum_xx accumulated past ~1e10 and the square shed ~10 of a double's
significant digits before the value was ever stored. The sub-day x detail was
lost at accumulation time, which no read-side identity can recover; an
ill-conditioned window's composed slope/r2 then drifted from the live REGR_*
probe past the parity gauge.
Rebase x to a fixed origin (epoch-days of 2020-01-01) at write time so the
squared terms stay O(1e7) and exact. Slope, r2, and population sd are invariant
under the affine x-shift, so the live raw-epoch probe still parity-matches the
cross-bucket compose; composeRegression needs no formula change. The per-bucket
slope/r2 columns stay on the raw axis (shift-invariant).
Migration 0193 recomputes sum_x/sum_xy/sum_xx from measurements on the rebased
basis, one set-based update per granularity. Recompute-from-source is idempotent
and rerun-safe; it touches only rows that already carry the accumulators and
leaves pre-migration NULLs for the boot re-fold.
Un-quarantine the DST-boundary parity test: it now matches live REGR_* within
the existing 1e-9 relative bound, with the WEIGHT case still bit-identical.
Marc-André Bombeck