Four things an author could not do, and the check that should have caught them.
A Decision could not be deleted. Case, Reference and Question all could, so an
author who opened a decision draft had no way to close it. The contract gained
the operation and the list now offers it for every kind. Its path carries the
project because a decision belongs to one; a row with no project says so rather
than failing.
Assets could only be managed by leaving the document. The picker now deletes
one in place — the server still refuses an asset a document uses — so a
mistaken upload does not cost the author their editing session.
Zoom was decided for the author and could not be changed: only a DIAGRAM got
it, so a screenshot uploaded as an image or attachment went in with zoom off
and no way to turn it on. It now defaults on for images and the picker offers
the choice. The toggle is a picker control, not a document field, and carries
its own class — wearing the field class put it in the editor's field list.
`scripts/smoke/production-sweep.ts` walks every public and Studio screen and
the document flow, reporting console errors, failed API calls and error text.
It exists because verifying only the screen I had just changed is what let
broken screens reach production repeatedly; this runs before a deploy, not
after a report.
`/studio/releases` answered a plain-text 404 from nginx. The route existed, the
chunk was built, and the SPA could reach the screen by client-side navigation —
but a hard load or a reload never got that far, because the web server had
never been told the path exists.
The public half of the serving contract derives its patterns from the route
registry. The Studio half was a hand-maintained array, and it failed the way
hand-maintained arrays fail: the comment above `^/studio/assets$` records that
exact bug being fixed once already, and adding a route repeated it immediately.
Both halves now come from the same source, so a Studio route that exists is
served without anyone having to remember.
Deriving them yields one pattern per route rather than the old alternation that
folded the four document sub-screens together. Same matched set, and it no
longer needs a human to keep the grouping honest.
The public site shipped a Releases page and a footer link to a release, and
neither could ever have content: the read path existed, the write path did not.
This adds the seven release operations to the contract contribution and the
gateway, and a Studio screen that can actually write one.
The editor is six markdown fields rather than one, because that is what the
contract models and what a release note is — why, what, what a reader notices,
what it leaves in the code, how it was verified, what is still open. Publishing
is separate from saving: a draft saves in any state, but the public query keys
on PUBLISHED alone, so publish is where completeness is demanded.
No new CSS. The screen reuses the document editor's field classes and the
working-copy list's row classes, so it inherits Studio's spacing and type
instead of introducing a second look.
The footer previously linked `/releases/0.1.0` — a version that did not exist,
so the link 404'd, and one that would have gone stale at 0.2.0 anyway. It now
points at the changelog index, which is the only place that knows what the
latest release is and which reads correctly when there are none.
Route inventory, navigation order, message catalog, manual accessibility
evidence, artifact baseline, and the pinned gate-shape digest all move with the
new route. The digest was recomputed by first reproducing the previous constant
from the previous gates.json, so the computation is known to be the one it was
pinned under.
Publishing needs a topic and nothing could create one. The backend now owns
that surface; this is its consumer — the management contract vendored, a
gateway over its nine operations, and one Studio screen that lists, creates,
and deletes topics and projects.
The screen adds no CSS. It reuses the classes the working-copy list already
uses, so it inherits Studio's spacing, type, and colour rather than growing a
second visual vocabulary beside them. Scope stops at list/create/delete:
renaming, phase changes, and visibility are implemented in the backend and
declared in the contract, but their screens are a separate design.
Two real defects surfaced while making the public port async, and both would
have shipped:
The search page and the header search dialog shared a query key. With an empty
query, `["tech-log","search",""]` was identical for both, so react-query
handed one surface the other's cache — different shapes — and the page died
reading a field that was not there. Keys now name the surface.
The explore filter's selects are uncontrolled and read `defaultValue`, which
React applies once. Their options arrive later now, so the first render had
nothing to match and the value stayed empty: a topic in the URL no longer
showed as selected. The form key includes whether the catalog has arrived, so
it remounts with the options present. Controlled inputs would be the other
answer, but this form submits to build a URL — the URL owns the value.
The route brought its own bookkeeping: a build chunk, a manual accessibility
evidence file, and the CI artifact baseline that counts them. The gate pins a
digest of its own shape precisely so a new route cannot slip in without that
count being reviewed.
Test harnesses that render public screens now assemble the query providers and
await the settled paint, because the screens they render became async.
The serving contract enumerated every public path the bundled fixture happened
to contain, and the generated nginx published exactly those as `location =`
blocks. A record published after the build — the entire point of having a
backend — answered 404 at the edge before the SPA was ever asked, and no amount
of correct routing inside the bundle could recover it. Twenty-seven frozen
paths, and any twenty-eighth was unreachable.
The route contract already declares which paths exist; the catalog only decides
which of them currently resolve, and that is the SPA's call rather than the web
server's. So the contract now emits one regex per registered Public route,
derived from the router, the way the Studio half has always worked.
A parameter matches one segment and never a slash, so /cases/a/b stays a 404
instead of quietly rendering a case page. The catch-all route is dropped rather
than translated: serving index.html for every unmatched URL would turn an edge
404 into a soft 200 and hide broken links from crawlers and from us.
Verified against a built image: /cases/a-brand-new-slug now answers 200 while
/nope and /cases/a/b still answer 404.
schemaVersion goes to 2 because the field changed shape, not just contents —
a consumer reading publicSpaPaths would otherwise see an absent key rather than
a version it can refuse.
The public surface — 17 of the 28 registered routes — reads from a 29KB
TypeScript fixture and never touches the backend. `TECH_LOG_STUDIO_SOURCE`
only ever switched the Studio gateways; `publicContent` was wired to the
static adapter unconditionally, so no configuration could make the public
site show published content. This is the first half of closing that: the
contract and the switch, with the adapter still to come.
The generator now vendors both canonical contracts instead of one. They are
independent — different services on different schedules — so each carries
its own digest and operation list, and updating one leaves the other's drift
gate quiet.
`TECH_LOG_PUBLIC_SOURCE` is deliberately a second flag rather than a rename
of the Studio one. The combination that matters right now is exactly the one
a single flag cannot express: the authoring backend is live while the public
read API does not exist yet. production stays on MOCK for that reason —
pointing it at HTTP today would empty the live site — and moves when the
backend serves /api/v1/public.
Also records the compatibility evidence the registry gate wanted for the
Studio access change in fff5e6f. That gate has been failing since, which is
on me: the change was real and breaking, and it shipped without the note
explaining that route ids and schemas are untouched and only the access
classification moves.
The repository had no container image and no production-shaped serving
configuration. `dist/server.mjs` is a preview server that applies neither
the security headers nor the cache policy `config/hosting/` declares, so a
deployment had nothing correct to run.
`scripts/generate-nginx-config.ts` derives the server block from
`dist/tech-log-serving-contract.json` plus the two hosting policy files, so
the served headers and cache lifetimes cannot drift from what the contract
declares. It emits no TLS and no proxy blocks: the edge terminates TLS and
routes /api, and baking a backend address into the image would tie the
bundle to one deployment. Static surfaces use `alias` because a base-path
build serves /dev/assets/... out of dist/assets/..., which `root` plus URI
would look for one directory too deep.
The image copies that config next to the bundle and normalises permissions:
the build writes config.json 0600, which nginx cannot read, so the container
came up healthy and answered 403 for the one file the SPA needs to boot.
index.html never referenced public/favicon.svg. The file shipped and nginx
served it, but browsers asked for /favicon.ico, got a 404, and fell back to
the default icon. `%BASE_URL%` rather than an absolute path so a prefixed
deployment points at its own copy.
development.json moves to the HTTP Studio source; the mock source has no
backend to authenticate against, which is the whole point of that profile.
Nothing in the gate set read `public/*.json`. Every static gate passed and the
whole suite passed while `pnpm dev` rendered the boot-error screen instead of
the app, which is the only reason the drift survived two contract changes.
`check:dev-release-manifest` compares the fixture's `setAlgorithm`, `setDigest`
and package set against what `generate-contract-set.ts` composes, and is
registered on FE-GATE-010 next to `check-tech-log-contract` so CI executes it.
Unlike the refresh wired into contract generation, this observes the composed
set directly, so it also catches a contribution added to or removed from
`installed-contract-contributions.ts`.
`CANONICAL_GATE_SHAPE_SHA256` recomputed by hand, as always: the committed
constant 98d19911... was first reproduced from the committed `gates.json` with
an independent transcription of `canonicalGateShapeSha256`, and only then was
b4096244... hashed from the new one. Command counts move 84/96 -> 85/97;
artifacts stay at 130 because the gate publishes no evidence file, matching
`check-tech-log-contract`.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
`corepack pnpm dev` did not boot. Plain `vite` serves
`public/release-manifest.json` verbatim, and that hand-maintained fixture still
declared `"packages": []` after the first real contract contribution was
registered. `verifyContractSet` runs unconditionally at boot — before any
adapter is chosen, so in the default `MOCK` mode as much as in `HTTP` — saw the
build had compiled `@tech-log/studio-contract` and failed closed with
`CONTRACT_SET_PACKAGE_MISSING`. Production builds were never affected:
`scripts/generate-build-manifest.ts` derives `dist/release-manifest.json`'s
block from the same composed set.
Editing the fixture by hand is not the fix — it had already gone stale twice,
once when the package first appeared and once when the contract moved 2.0.0 ->
3.0.0, because every regeneration changes the package digest. So
`generate:tech-log-contract` now refreshes the block itself, as its last step
and through a dynamic import so it reads the canonical source it just wrote.
`generate:dev-release-manifest` does the same refresh on its own.
The composed set can also change without the contract being regenerated — a
contribution added to or removed from `installed-contract-contributions.ts`
moves it. Tying the refresh to contract generation is therefore necessary but
not sufficient; the CI gate that follows is what closes that half.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
`/studio/assets` was the one route in the TechLog route contract missing
from `TECH_LOG_CANONICAL_ROUTES` and `TECH_LOG_STUDIO_STATE_PATHS`, so it
had no golden, no responsive-overflow check, and no Axe scan. That gap was
not hypothetical: the responsive overflow just fixed in the Asset search row
shipped in two places and was only caught in the Picker, which has a golden;
the Library's identical copy would have gone unseen.
Verifying the render before capturing anything turned up a second defect.
`studioSpaPathPatterns` in the production serving contract never listed
`/studio/assets`, so a hard navigation or reload of that URL was answered
with the in-shell Studio 404 -- the screen was reachable only by client-side
navigation from another Studio page. A golden taken then would have frozen a
404. With the pattern added, the route serves 200 text/html like its
siblings, and the page renders correctly at both breakpoints:
scrollWidth == clientWidth == viewport at 360 and 1440, `main` and the `h1`
visible, the search label/input/button all inside the 360px viewport, and no
critical or serious Axe violations in chromium.
Three new goldens, no existing golden regenerated:
tech-log-studio-assets-{360,1440} from the canonical route list and
tech-log-studio-asset-library-1440 from the Studio state list, matching how
every other Studio route appears in both. `pnpm test:visual` is 133 passed,
up from 130.
The demo profile seeds no assets, so the golden captures the empty state --
which still covers the header, heading, search row and status region, the
surface the overflow regression lived on.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The TechLog suite already ran in CI, but only inside `test:coverage`'s combined
`vitest run tests/runtime-schema tests/unit … tests/features/tech-log`
invocation on FE-GATE-005. A TechLog regression therefore surfaced as a
coverage-gate failure, and the only junit that carried it was coverage.xml,
which reports every other suite at the same time.
`test-tech-log` now sits on FE-GATE-007 beside `test-reference-feature`, the
sibling it mirrors — both drive a suite under `tests/features/` — and declares
`artifacts/tests/tech-log.xml`, which `test:tech-log` already wrote, as its own
command-generated junit evidence. The gate log now names
`$ corepack pnpm test:tech-log` as its own step and the failure lands on the
suite that produced it.
Adding a command and an artifact moves the exact-count authority to 84
definitions / 96 references / 130 artifacts (109 evidence references), and
changes the canonical gate shape digest because FE-GATE-007's commandIds and
evidenceArtifactIds are part of it. There is no tooling to regenerate that
digest, so it was recomputed by hand under the standing procedure: a fresh
transcription of `canonicalGateShapeSha256` first reproduced the committed
f3cc9075… from the unedited config/ci/gates.json — proving the transcription,
not just agreeing with whatever the check compares against — and only then
hashed the edited file to 98d19911….
`corepack pnpm ci:gate -- FE-GATE-007` passes end to end; the generated
workflow bytes are unchanged, since the yml dispatches gates rather than
commands.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The 16 sandboxed provider tests in `tests/unit/ci-artifact-contract.test.ts`
fail wherever unprivileged user namespaces are denied. bubblewrap is installed
and answers `--version`, but `bwrap --unshare-net … -- /bin/true` exits 1 with
`bwrap: loopback: Failed RTM_NEWADDR: Operation not permitted`, and it fails the
same way with no netns flag at all (`setting up uid map: Permission denied`), so
this is the whole nested-userns capability and not one option. Sixteen
assertion errors on every local run buried whatever else the file had to say.
`scripts/lib/provider-sandbox-probe.ts` now runs a trivial command under the
real isolation options — `runProviderInSandbox` spreads the same
`PROVIDER_SANDBOX_ISOLATION_ARGUMENTS`, and a test fails if an isolation option
is added to the run without the probe having to clear it. Capability is
measured, never inferred from the binary existing or from a version string;
both would pass here.
An unusable sandbox means two different things in two places, so the decision
is explicit. Locally it is an environment fact: the affected tests skip and
carry the bwrap diagnostic as their skip note, visible as `↓ … [reason]`. In CI
it is a regression — a security gate that silently stopped running is exactly
what these tests exist to catch — so the same probe result fails the run
through one guard test that says "the provider sandbox is unavailable" instead
of sixteen assertion errors.
CI is detected with `CI === "true"` via a new `isCiRun`, sharing the predicate
that already gates `ciBuildEnvironmentFailures`. The workflow sets it at the
top-level `env:` block, so it holds in every job; `CI_RUN_ID` and its
`GITEA_`/`GITHUB_` fallbacks are declared only by the release-tier provider
jobs and are absent from the merge gates that run this file, so keying on them
would have left the CI branch permanently dead.
The three `it.each` groups become `it.for` because only `.for` passes the test
context, which is what carries the skip note.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
`check:tech-log-contract` existed and worked but nothing ran it. No gate in
`config/ci/gates.json` referenced it and `test:all` did not chain it, so a
hand-edit of the vendored canonical yaml or of `generated.ts` passed every
gate the repository actually executes -- the exact regression the digest pin
exists to prevent.
FE-GATE-010 (architecture/contract governance) now owns the command, beside
`check-registries` and `check-ci`, and `test:all` runs it in front of
`test:tech-log`. The canonical authority baseline moves to 83 command
definitions / 95 references and the gate-shape SHA-256 is recomputed with
`canonicalGateShapeSha256`; the recomputation was first verified by
reproducing the previous constant from the previous gates.json.
`tests/features/tech-log/contract-generation.test.ts` now fails if either
wiring is removed again.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Adds TECH_LOG_STUDIO_ASSETS (/studio/assets) as a route reachable but
excluded from primary Studio navigation (navigationLabel/navigationOrder
null), plus the AssetLibrary screen that lists assets, shows usage, and
lets an operator archive or hard-delete one. canHardDelete() is a pure
gate mirroring the server's ASSET_IN_USE rule so the screen never offers
an action the server would refuse.
Adding a 28th route also required updating the route-scoped CI
accessibility-evidence gate (FE-GATE-009 in config/ci/gates.json, plus
its authority-baseline counts and shape digest in
scripts/contracts/ci-gates.ts) and the Vite route-to-chunk map that
scripts/generate-build-manifest.ts depends on, or test:unit and the
production build both fail. See task-11-report.md for the full
breakdown.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Registers the TECH_LOG_STUDIO_SESSION SAME_ORIGIN_COOKIE auth profile and
declares the tech-log-studio-http-v1 contribution covering all 18 JSON
Studio operations (everything except the multipart uploadStudioAsset),
built from two shared builders (safeOperation/keyedOperation) so every
KEYED command gets IDEMPOTENCY_REPLAY recovery and a zero retry budget,
and every SAFE read gets a plain retry budget, without repeating the
declaration shape 18 times.
Rescopes the canonical package identity from "tech-log-studio-contract"
to "@tech-log/studio-contract" (generator, canonical-source.json,
contract-generation.test.ts) because the platform's contribution
composer requires an npm-scoped packageId; the unscoped form failed
composition. Updates the release-manifest test fixture, which hardcoded
an empty expected contract set, to derive its expected packages from the
real installed set now that TechLog is always installed.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Vendors the canonical studio-v1.yaml, generates types via an isolated
`pnpm dlx` toolchain (openapi-typescript needs TypeScript 5's classic
compiler API; this repo pins TypeScript 7.0.2 per VD-01, whose root
export has none), and adds an offline drift gate that checks the
vendored yaml/generated types/canonical-source.json against each
other without touching the sibling design-package repo or the network.
Regenerating from canonical surfaces real, new required fields on
existing schemas (WorkingCopyDetail.nextAction, PreviewDetail/PublicPreview
.dependencyRevision, StudioDashboard.totals.needsValidation,
PublicationSnapshot.contentFormatVersion/rendererContractVersion) and a
new required EvidenceFigureBlock.asset. The mock gateway and fixtures
are updated to satisfy the former; the latter exposes a real authoring-
vs-rendering conflation in the content-format parser (it declared its
output as the server's fully-resolved PublicRenderModel type, which it
has no asset catalog to satisfy). Split that boundary: the parser now
produces an authoring block type omitting the resolved asset, and each
of its three consumers (the mock gateway, the Studio instant preview,
and the static Case demo page) attaches the resolved descriptor from
its own asset source through a shared, pure domain-level resolver.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Integrates the frontend template sync (a0fbafb → 5434760) into the TechLog UI
migration. Merged in this direction so every conflict is resolved and proved in
the worktree; main is only fast-forwarded afterwards and never holds a state
that was not verified here.
15 conflicts. The rule throughout: keep the template's mechanism, keep the
product's content, and never invent a third state neither branch would accept.
The template's product manifest and its runtime feature kill switch are adopted.
The route registry is deliberately not composed from contract.routes: the
reference feature still declares screens this product deleted, and reducing over
them would register paths with no component behind them. ROUTE_FEATURE_OWNER is
narrowed to registered routes for the same reason. The first resolution did
compose from contract.routes and was rejected by product-features.test.ts.
Three files pinned counts and a digest describing the gate contract. Neither
side's numbers describe the merged config/ci/gates.json, so they were recomputed
from it rather than chosen: 27 gates, 82 commands, 94 command references, 107
evidence references, 128 artifacts, shape sha256 5063586d.
README.md and docs/accessibility/manual-checklist.md now enumerate this
product's 27 routes, which the template's own verify:documentation requires.
product-feature-switch.test.tsx was rewritten around the invariant that still
applies here — no registered route without a component — rather than deleted
with the screens it used to exercise.
docs/operations/template-merge-2026-08-17.md records every decision, the gate
results, and the three follow-ups this merge deliberately did not decide.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Carries eight template commits: the provider sandbox actually running, release
admission to a named environment, the product feature manifest with its runtime
kill switch, architecture and documentation rules that match what is enforced,
the removability fixtures, and the browser, visual and performance evidence.
Product identity is unchanged. `package.json` keeps `tech-log-frontend` and the
catalog keeps the Tech Log naming; the home page was not in the delta. The
visual baselines are this product's own — the template's were excluded from the
transplant and these were regenerated here, where the only difference is the
platform overview's new product-feature section.
What this repository gains operationally: `config/runtime/{local,development,
staging,production}.json` with `FE-GATE-027` refusing an artifact whose runtime
document does not match the environment it is being admitted to, and
`FEATURE_OVERRIDES` for taking an installed feature out of service without a
rebuild.
Verified here: eight gates green, build green, visual 5/5, and 1,858 of 1,859
tests in the suites that do not need a sandbox — the one failure passes in
isolation and is a jsdom lazy-chunk timeout under parallel load. The provider
suites cannot run on this machine at all: `kernel.apparmor_restrict_unprivileged
_userns=1` makes `bwrap --unshare-net` fail, reproducible without any code from
either repository.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The product was materialized from the template at `4dc033c` and has stayed
on it through 43 template commits, so it was missing all three rounds of
adapter remediation — including files it never had, such as the shared
`abortable-operation` primitive and the `exact-snapshot` decoder that
later fixes are written against. Taking only the newest round was not
possible for that reason: the delta is coherent only as a whole.
The product had not touched `src/adapters` at all since materialization,
so the 140-file delta applied with a three-way merge and no conflicts.
`package.json` was the single overlap and merged cleanly: the product owns
`name`, the template contributed `check:adapter-inventory`,
`check:remediation-ledger` and the image-resolve-signal type fixture.
All 24 product-owned files — README, index.html, CI workflow, i18n
catalog, home page, generated schemas, evidence scripts, component and
visual snapshots — are byte-identical to `main`.
`template.lock.json` now pins the synced revision and tree.
Verified in this repository, not inherited from the template: six type
projects, lint, nine gates (adapter inventory, remediation ledger,
registries, diagnostics, realtime boundaries, architecture, browser
file/storage boundaries, optional recipes, documentation), the production
build, and 2,054 of 2,073 tests. The 19 failures are all in
`tests/unit/ci-artifact-contract.test.ts` and are the same pre-existing
sandbox RLIMIT, EMFILE, umask and `/tmp` permission behaviour the template
records; four suites that failed once under parallel load pass in
isolation.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
DongHyeonka