# Update Log ## 2026-10-31 (first release builds failed: pinned the amd64 manifest) * **Fix**: [Dockerfile](../Dockerfile) and [Dockerfile.dev](../Dockerfile.dev) - the base digest `sha256:18f03e4b...` was the **linux/amd64 platform manifest**, so `docker build --platform linux/arm64` on the arm64 runner failed at `Build image` one second after the step started (run 157, v1.5.1; run 153, v1.5.0). Both now pin the multi-platform OCI **index** digest `sha256:e4db4c91...`, verified with `docker buildx imagetools inspect` to contain linux/amd64 and linux/arm64/v8. See [Production image shape](decisions/production-image.md). * **Detail**: checkout, Docker CLI install, `Check daemon access`, registry login, the immutability guard and version resolution all passed - the runner and registry setup were correct; only the base image pin was wrong. ## 2026-10-31 (release-please tag: empty package-name is the verified fix) * **Correction**: removing `package-name` in #36 was not enough. Run 183 shows the node strategy falling back to `package.json` `name`, so the release phase still failed with `PR component: undefined does not match configured component: nuxt-app`. * **Fix**: [release-please-config.json](../release-please-config.json) sets `"package-name": ""`. An empty string is not nullish, so `getBranchComponent()` stays empty and matches the legacy `release-please--branches--main` branch. This is the only setting that lets release-please accept the already-merged #35. * **Evidence**: the action's bundled release-please 16.15.6 was run locally against the live Gitea API with writes intercepted. It produced `POST /repos/CV_Website/CV_Site_Nuxt/releases` with `{"name":"v1.5.0","tag_name":"v1.5.0","target_commitish":"a80c9ebf..."}`, then swapped #35 to `autorelease: tagged`. The official default `release-type: node` mode was tested the same way and still fails for #35 with `configured component: nuxt-app`, so it cannot retroactively tag the legacy PR. * **Expected result**: the run after this merge creates `v1.5.0` at `a80c9eb` through the release-please action itself. No manual `git tag`. ## 2026-10-31 (release-please component mismatch: no v1.5.0 tag) * **Fix**: [release-please-config.json](../release-please-config.json) - removed `package-name: CV_Site_Frontend`. With `include-component-in-tag: false` the strategy component and the release PR body component are empty, but `getBranchComponent()` still returned `package-name`; the release phase logged `PR component: undefined does not match configured component: CV_Site_Frontend` and returned before creating the tag. The component now matches, so the next run should tag #35. * **Evidence**: run 181 (task 578, immediately after #35 merged) - the reconcile step correctly left #35 for the action because its branch survived, then the action found #35, hit the component mismatch and aborted with "There are untagged, merged release PRs outstanding". The earlier label/branch fixes were correct; this was the remaining check. * **Expected result**: the run after this merge creates `v1.5.0` at the #35 merge commit (`a80c9eb`), marks #35 `autorelease: tagged`, and stops opening duplicate release PRs. ## 2026-10-31 (release-please reconcile step needs a checkout) * **Fix**: [release-please.yml](../.gitea/workflows/release-please.yml) - added `actions/checkout@v4` before the reconcile step. The job had no checkout because the release-please action clones the repository into its own temp directory; the new step runs `scripts/reconcile-release-please.mjs` from the workspace, so the first post-merge run failed with `Cannot find module '/workspace/CV_Website/CV_Site_Nuxt/scripts/reconcile-release-please.mjs'` (task 571). The checkout supplies the file; the action still uses its own temp clone for the release branch. ## 2026-10-31 (PR #33 second review: release tag validated, pending PRs queried by label) * **Fix**: [reconcile-release-please.mjs](../scripts/reconcile-release-please.mjs) - both PR-Agent findings were confirmed. A tag at the merge commit now counts only when it is exactly `v`; a foreign CI/annotation tag is logged and ignored. Pending release PRs are fetched through the `autorelease: pending` label filter instead of paging every closed PR, so a concurrent close cannot shift a page and hide one. * **Guard**: a merged PR without `merge_commit_sha`, or a manifest without a version, now fails loudly instead of producing `vundefined`. * **Verified locally**: dry-run against the live repo still matches #22 to `v1.4.0`; `pnpm lint`, `pnpm typecheck`, `pnpm build` and `pnpm audit --prod --audit-level moderate` pass on the rebased tree. * **Known issue**: PR-Agent's `/describe` posts `Failed to update PR description`; the bot token has `write:issue` but not the PR-edit scope. Recorded with the fix in [PR-Agent operations](pr-agent.md). ## 2026-10-31 (registry host corrected: Gitea's built-in registry on git.noelsabia.dev) * **Correction**: `registry.noelsabia.dev` is gone from the documentation and the workflow. Gitea's built-in container registry is served on the Gitea hostname itself, so the image is **`git.noelsabia.dev/cv_website/nuxt:`** ([docs.gitea.com/usage/packages/container](https://docs.gitea.com/usage/packages/container/)). `registry.noelsabia.dev` was a separate hostname set up before that was known; the server-side setup is being reverted. No separate registry, no extra hostname, no extra certificate - Gitea already provides this by default. * **Update**: [publish-image.yml](../.gitea/workflows/publish-image.yml) - `IMAGE`, `docker login` and the v2 `manifests/` immutability check all point at `git.noelsabia.dev`. [Publishing images to the own registry](decisions/registry-publishing.md) records the corrected host and the reason. ## 2026-10-31 (registry confirmed: Gitea's built-in container registry) * **Confirmation**: Logged in and pushed by hand - push access works. The registry is **Gitea's built-in container registry** ([docs.gitea.com/usage/packages/container](https://docs.gitea.com/usage/packages/container/)), not a separately deployed component: it is served on the Gitea host, and the `/cv_website/` path segment is the organisation namespace. This is the default - no Registry/Harbor to stand up, nothing to reinvent. Host corrected in the entry above; see [Publishing images to the own registry](decisions/registry-publishing.md). ## 2026-10-30 (PR-Agent moved into the Gitea compose stack) * **Refactor**: the `pr-agent` service now lives in `/home/noel/self-hosted/gitea/docker-compose.yml` with `/home/noel/self-hosted/gitea/pr-agent.env`. The separate `/home/noel/self-hosted/pr-agent` stack and the unused `/opt` operator-access scripts were removed; the service is managed with the Gitea stack. * **Docs**: the operator README moved to [PR-Agent operations](pr-agent.md). The decision, index and update log point at the new topology. * **Evidence**: after the move, the container is healthy, a signed webhook from the Gitea container returns 200, and a direct `glm-5.3-flash` call returns `OK`. ## 2026-10-30 (PR #28: lockfile and PR-Agent routing repaired) * **Fix**: `pnpm-lock.yaml` - merging `main` left the lockfile at `@nuxt/icon` 2.6.0 while `package.json` asked for 2.6.2, so `pnpm install --frozen-lockfile` aborted in CI. The lockfile now resolves 2.6.2. * **Fix**: `pnpm-workspace.yaml` - `seroval` 1.6.9 was the two moderate audit findings (GHSA-f86p-h4j2-7qgj, GHSA-9wx9-5pv9-3349). The patched 1.6.11 is forced through `overrides`; the blocking audit now reports only two low-severity advisories and passes the moderate gate. * **Fix**: PR-Agent model routing - OpenCode Go rejects requests without `x-opencode-session`; `LITELLM__EXTRA_HEADERS` now supplies it, and the fallback keeps the `openai/` prefix. Auto commands are `/describe` and `/review`; `/improve` is on demand. A dummy-key call reaches the provider's auth check instead of failing on the missing header. * **Evidence**: the key was wired into the live `.env`; a direct `glm-5.3-flash` call returned `OK`, and `/review` on PR #28 posted a PR Reviewer Guide (no findings). The failed-review comment was removed. * **State**: the key is set only in `gitea/pr-agent.env`; the first `/review` posted on PR #28 with no findings. ## 2026-10-30 (PR-Agent replaces the opencode review job) * **Decision**: [PR-Agent for Gitea reviews](decisions/pr-agent.md) - a pinned `pragent/pr-agent:0.49.0-gitea_app` container on `server0` reviews PRs through the Gitea bot user `pr-agent`; the endpoint stays on the `gitea_gitea` Docker network. * **Update**: [pr-checker.yml](../.gitea/workflows/pr-checker.yml) - removed the `review` job, the opencode installer and the verdict-posting step. PR-Agent now runs `/describe` and `/review` on PR open/reopen and answers slash commands. * **Creation**: `deploy/pr-agent/` (later removed; the service now lives in the Gitea compose stack) - compose service, example environment and operator README. The real `.env` lived only at `/home/noel/self-hosted/pr-agent/.env` (mode 600) on `server0`. * **State**: the container is running and the webhook signature path is verified. Gitea gained `[webhook] ALLOWED_HOST_LIST = private:3000` (one restart, healthy) and the repository has a signed webhook. The OpenCode Go API key was still missing at commit time, so no model call or real review has run yet. ## 2026-10-31 (manual publish rehearsal without a tag) * **Creation**: [publish-image.yml](../.gitea/workflows/publish-image.yml) gained `workflow_dispatch` with `version`, `push` and `latest` inputs. Default is **build + smoke test only** (`push=false`): no tag, no package version, immutability guard skipped, `latest` never moved unless explicitly requested. Use it to verify the `publish` runner (socket + `automount`) before the first release. Manual runs do not replace the tag path - a published image still always maps to a release tag. ## 2026-10-31 (audit green again: nuxt 4.6.3 and dependency overrides) * **Fix**: `pnpm audit --prod --audit-level moderate` failed on every branch (#30, #31, #33 and this PR) - new advisories plus a Nuxt patch release. `nuxt` 4.6.2 -> **4.6.3** clears GHSA-9v7c-q64q-vv33 (dot-segment middleware bypass), GHSA-3xf4-9g3c-4h4w and GHSA-x287-wcfc-q46c (dev-server file read/runtime config). * **Fix**: [pnpm-workspace.yaml](../pnpm-workspace.yaml) - overrides for `node-forge: 1.4.1` (the old GHSA-86w9-cpqp-85rv ignore is removed - a patch exists now) and `glob@>=9.0.0 <=14.0.1: 14.0.2` (GHSA-3g3m-2fjr-6pw4; glob 7 is kept for tailwind-config-viewer). * **Decision**: minimatch GHSA-g3hv-75wg-m249 stays a reviewed ignore. Its patch (11.0.1) dropped `makeRe`, which `eslint-plugin-import-x` imports - forcing it makes `pnpm lint` fail with "does not provide an export named 'makeRe'" (measured). All affected paths are build/lint-time. See [Vulnerability audit policy](decisions/vulnerability-audit-policy.md). * **Verified locally**: `pnpm lint`, `pnpm typecheck`, `pnpm build` and the exact audit command all pass on the merged tree. ## 2026-10-27 (release-please still never tagged: the deleted head branch) * **Fix**: [release-please-config.json](../release-please-config.json) gained `"changelog-host": "https://git.noelsabia.dev"` - the changelog compare links were generated against `github.com/CV_Website/CV_Site_Nuxt`, a host this repository does not live on. Root cause of the missing tags: **Gitea rewrites `head.ref` to `refs/pull//head` when the merged PR's branch is deleted**, and release-please parses that field as the release branch, so `BranchName.parse` fails and the release is dropped - silently, exit code 0. Evidence: run 143 (task 475) logs `✖ Bad branch name: refs/pull/22/head` right after `Found pull request #22`, then `Expected 1 releases, only found 0`; #22 did carry `autorelease: pending`. The repository deleted every merged branch (`default_delete_branch_after_merge: true`); it is `false` now. Four versions (1.1.0-1.4.0) were consumed by the manifest with no tag ever created, so the release baseline has to be restored by hand. See [Release engineering](decisions/release-engineering.md). ## 2026-10-24 (release-please never tagged: the missing labels) * **Fix**: [release-please.yml](../.gitea/workflows/release-please.yml) - the workflow now creates `autorelease: pending` and `autorelease: tagged` before the action runs, and gained `workflow_dispatch` so a missed release can be replayed. Root cause: release-please finds merged release PRs **by label**, and Gitea resolves label names with `GetLabelIDsInRepoByNames`, a SQL `IN` filter that silently drops unknown names. The repository had no labels at all, so PRs #20 and #21 were merged unlabeled, every run found zero merged release PRs, created zero tags and opened the next release PR instead - reporting success each time. Evidence: run 137 green, 0 tags, 0 releases, `labels: []` on #20/#21/#22, and the repo's label list empty. See [Release engineering](decisions/release-engineering.md). ## 2026-10-24 (release publishes an image) * **Creation**: [publish-image.yml](../.gitea/workflows/publish-image.yml) - a `vX.Y.Z` tag push builds `Dockerfile` on the arm64 runner and pushes `git.noelsabia.dev/cv_website/nuxt` as the version and `sha-<12>`, plus `latest` when that tag is the newest release. Main pushes deliberately do not build: the registry holds releases, not commits. The job uses the runner's own Docker daemon (client only in the `node:22-bookworm` container), so the act_runner needs `container.docker_host: automount`; the first step fails with that cause if it is missing. That socket is root-equivalent, so it may only sit on a label PR jobs cannot use - recorded under the trust boundary in [Publishing images to the own registry](decisions/registry-publishing.md). Repo secrets: `REGISTRY_USERNAME`, `REGISTRY_PASSWORD`. * **Review follow-up** (PR #19, 1 BLOCKER + 4 WARN + 2 NIT): applied the strict `vMAJOR.MINOR.PATCH` check, the newest-release rule for `latest`, a smoke test that starts the image and requires 2xx on `/`, and `timeout-minutes: 20`. Rejected with evidence: the `.dockerignore` leak (`.env*`, `*.pem`, `.git` are excluded) and the Docker-client/daemon API mismatch (20.10.24 client against a 29.9.2 daemon built successfully over the socket). * **Second review follow-up** (PR #19): the job now runs on the dedicated `publish` label instead of `ubuntu-latest`, so the host socket it needs can never be shared with PR jobs - the earlier version required a runner configuration its own trust boundary forbade. Added the immutability guard (registry v2 `manifests/`: `200` aborts, `404` continues) after measuring that `docker manifest inspect` fails on OCI indexes with the pinned 20.10 client; the smoke test now also requires a rendered body (10 kB floor) rather than a bare 2xx. `.github/actionlint.yaml` declares the custom label, and the hook passes it with `-config-file` because this repo keeps workflows in `.gitea/`. ## 2026-10-23 (profile data in the terminal) * **Creation**: [Profile data](profile.md) - imported the LinkedIn export `Profile.pdf` into `app/data/chapters.ts`: PreZero active; Defifort, Alignerr, Outlier, freelance and 42 finished. Each row carries a context label (Enterprise, Web3 startup, AI Developer, Freelance, Education). The terminal now shows 1 active / 5 sealed / 14 rebounds; the export has no entry for `dreambit`, `ongoing-studies` or `apprenticeship`, so those rows are gone. `uptime` and `rebounds` remain mock. * **Update**: [backend endpoints](backend-endpoints.md) - examples and totals now match the real six-chapter list; the source frontmatter points at the export. ## 2026-10-23 (homepage metrics contract) * **Update**: [backend endpoints](backend-endpoints.md) - added `GET /api/metrics`: the chapter rows plus the terminal footer totals (`active`, `sealed`, `rebounds`) in one response. The homepage terminal is meant to show measurements, not mockup copy, so the gateway should own them; `/api/chapters` stays the row list for the life map. The old note that the counters are computed in the browser is gone. ## 2026-10-23 (PR #16 review checked, three findings applied) * **Update**: [pr-checker.yml](../.gitea/workflows/pr-checker.yml) - the review prompt now states that `pnpm-lock.yaml` is excluded from `.pr-review-diff.patch` on purpose. The reviewer read the exclusion as an uncommitted lockfile and raised a false BLOCKER; the lockfile is in the PR and matches `package.json`. * **Fix**: `scripts/generate-favicon.mjs` - `BACKGROUND = null` now writes alpha 0 instead of opaque black, matching the transparency the comment promises. The committed icons are unchanged (regenerated byte-identical). * **Update**: deleted `public/_robots.txt` - unreferenced leftover from the rename; `@nuxtjs/seo` writes `robots.txt` itself. ## 2026-10-23 (Nuxt SEO configured, static OG card) * **Decision**: Recorded [SEO stack and site metadata](decisions/seo-and-metadata.md) - `@nuxtjs/seo` 6.0.1 registered; `site` and `schemaOrg.identity` (Person) read `site.config.mjs`; canonical, sitemap, robots and social tags need no per-page setup. `ogImage: false` and `skewProtection: false` because the card is a committed PNG and the site deploys as static files. * **Update**: `public/og-image.png` (1200x630) and `og-image.alt.txt` generated from the pixel avatar and the site metadata by `pnpm og-image` (`scripts/generate-og-image.mjs`; needs `woff2_decompress` + `rsvg-convert`, output committed). SEO Utils emits `og:image`, `twitter:image` and their dimensions/alt automatically. * **Detail**: `app/pages/index.vue` sets the title to `Noel Sabia — ` and `twitter:card: summary_large_image`; `useHead({ titleTemplate: '%s' })` removes the `| Noel Sabia` suffix SEO Utils would add. `theme-color` is `#031428`. * **Detail**: `engines.node` is `^22.22.3 || ^24.15.0 || >=26.0.0`, the floor `@nuxtjs/seo` 6 enforces. Verified `nuxt build` and `nuxt generate`: canonical is `https://noelsabia.dev/`, sitemap has one URL, robots allows crawling and names the sitemap, schema graph carries the Person node, and all SEO assets return 200. ## 2026-10-23 (generated pixel favicon) * **Update**: `public/favicon.svg` and `public/favicon.ico` are generated from the header's pixel avatar by `pnpm favicon` (`scripts/generate-favicon.mjs`). No image library: the ICO is an uncompressed 32bpp BMP in a one-image ICO container (4.2 kB) and the SVG merges same-coloured cells into per-row runs (3.7 kB); a 256px render of both matches. * **Update**: `app/utils/avatar.js` now holds the palette and the 32x32 sprite as the single source, imported by `AppLogo.vue` and by the generator (plain JS, so Node can import it without a loader). `app.head.link` declares the SVG first and the ICO as the `32x32` fallback. * **Detail**: the icon keeps the header chip colour `#1b2b3f` as its background so the portrait stays readable on a dark tab bar; `BACKGROUND = null` in the generator makes it transparent. ## 2026-10-23 (icons from Nuxt Icon, icon webfont removed) * **Decision**: Recorded [Icons via Nuxt Icon](decisions/icons-via-nuxt-icon.md) - all icons come from `@nuxt/icon` with `@iconify-json/material-symbols` installed locally, `mode: 'svg'`, `serverBundle: { collections: [] }`, `clientBundle: { scan: true }`. Deleted `public/fonts/material-symbols-outlined-var.woff2` (3.9 MiB), its `@font-face` and the `.material-symbols-outlined` class in `app/assets/css/main.css`. Build reports 3 icons, 1.23 kB; SSR HTML carries the inline ``. * **Update**: The module's defaults would have called `api.iconify.design` at runtime (`mode: css`, `fallbackToApi: true`), which the local collection avoids; `serverBundle: 'auto'` would have bundled the whole set as an 8 MB server chunk. ## 2026-10-20 (cheaper review model, cached installs, merged checks) * **Update**: [pr-checker.yml](../.gitea/workflows/pr-checker.yml) - the review model is now `opencode-go/claude-haiku-5-5#medium`: cheap per token and, unlike `deepseek-v4.1-flash`, it has a real `#medium` effort level (flash exposes only low/high/max). ESLint and typecheck are merged into one `checks` job with a single `pnpm install`, and the pnpm store plus the opencode binary are restored with `actions/cache@v4.1.0` (v1 cache API, so it also works on a v1-only runner). Cache keys are built from `sha256sum pnpm-lock.yaml`, the pnpm store layout (`v11`) and the opencode version in shell steps, because `hashFiles()` evaluates empty in this act runner. See [AI review setup](decisions/ai-review-setup.md) and [PR pipeline layering](decisions/pr-pipeline-layering.md). ## 2026-10-20 (first flash review validated, two findings applied) * **Update**: PR #14 review (flash model, `VERDICT: APPROVE`) was fact-checked finding by finding. Applied: `` restored in `app.vue` (this PR's diff had dropped it with the NuxtWelcome swap), and the review job now requires `github.event_name == 'pull_request'` (manual dispatch - the new snyk trigger - would otherwise launch a paid review with no PR to post to). Rejected with evidence: the reviewer's `postcss-selector-parser` 6.1.5 override suggestion - the 6.1.5 backport exists on npm, but the blocking audit still flags it (`<7.1.6` covers 6.x), verified empirically; 7.1.6 stays. Deferred as design decisions: mobile nav, aria-current via NuxtLink, sprite-to-SVG, font fallback stacks, `borderRadius.full` token, the fall-endpoint budget in the API contract. ## 2026-10-20 (snyk pause made actionlint-clean) * **Update**: pr-checker.yml - the paused snyk job's `if: false` (flagged by actionlint as a constant condition once this PR touched the workflow file) became `if: github.event_name == 'workflow_dispatch'` plus a `workflow_dispatch:` trigger. Same pause semantics ([snyk paused](decisions/snyk-paused-and-secrets-policy.md)) - pushes/PRs burn no credits; the job is now runnable manually instead of dead code. ## 2026-10-20 (review model switched to flash after endpoint outage) * **Update**: [open questions](open-questions.md) - the review-model question is decided: `opencode-go/deepseek-v4.1-flash#max` in [pr-checker.yml](../.gitea/workflows/pr-checker.yml). The `deepseek-v4-pro#max` endpoint answered two consecutive runs with `Upstream request failed: Endpoint is unavailable` (runs 67 and 68 on PR #14, ~10 min apart, model still listed active) - the recorded default ("switch to flash") was applied. ## 2026-10-20 (blocking audit catches the tailwind dep, first override) * **Update**: [vulnerability audit policy](decisions/vulnerability-audit-policy.md) - the first PR run after adding `@nuxtjs/tailwindcss` failed the blocking audit on GHSA-rj75-hqrm-r3gf (moderate: `postcss-selector-parser` < 7.1.6, quadratic flat-selector parsing, pulled in by `tailwindcss` 3.4.19 → `postcss-nested` 6.2.0). Fixed per policy with the first `overrides` entry in `pnpm-workspace.yaml`, pinning 7.1.6 (already in the tree via `postcss-nesting`, build verified, audit green except the two reviewed highs). Same run: AI review job failed transiently (`Upstream request failed: Endpoint is unavailable` on `opencode-go/deepseek-v4-pro#max`; the model is still listed active, so retry on next push - the model-switch question stays open in [open questions](open-questions.md)). ## 2026-10-20 (frontend API contract) * **Creation**: Added [backend endpoints](backend-endpoints.md) - the smallest endpoint set the UI needs: `GET /api/health` (footer dot, `SYS_STABLE`), `GET /api/chapters` (terminal rows; counters computed client-side), and the later `POST /api/fall/{chapter}` and SSE `GET /api/events`, with polling-first as the deliberately simple start. ## 2026-10-19 (pipeline improvement backlog) * **Creation**: Added [future updates](future-updates.md) - five candidate improvements from Anthropic's code-review post (build job, PR intent input, no-repeat re-review, PR template, reviewer eval set), with effort estimates, batching advice for the review budget, and the list of scale-dependent machinery deliberately not adopted. ## 2026-10-19 (stale anonymous volume aborts pnpm) * **Update**: [Dev container shape](decisions/dev-container-shape.md) — the anonymous-volume remedy `docker compose up --build -V` moved from "not verified" to evidence (build, volume renewal, server ready in 2.62s). New failure mode recorded: after the mounted manifests drift from the volume's snapshot, pnpm 11's pre-run check wants to purge `node_modules` and aborts without a TTY (`ERR_PNPM_ABORTED_REMOVE_MODULES_DIR_NO_TTY`). Fix: `CI: "true"` in the compose service. ## 2026-10-18 (evening — release PR unblocked) * **Decision**: Recorded the [production image shape](decisions/production-image.md) - multi-stage, digest-pinned `node:26-slim`, non-root, healthchecked runnable Node server. Built and served locally: HTTP 200, uid 1000, healthcheck healthy. This decides the question that had kept `Dockerfile` empty (and failing the droast job with `DF071`). * **Decision**: Recorded the [vulnerability audit policy](decisions/vulnerability-audit-policy.md) - blocking audit with two reviewed ignores (`node-forge`, `braces`: no patched release exists; both build-time-only paths under nitropack). Learned en route: pnpm 11 ignores the `pnpm` field in `package.json` - settings live in `pnpm-workspace.yaml`. * **Update**: Added both decisions to the [decisions index](decisions/index.md). ## 2026-10-18 (later — CI/CD and releases) * **Creation**: Recorded the [PR pipeline layering](decisions/pr-pipeline-layering.md) — three layers (pre-commit, Gitea Actions, AI review), the cost rule that assigns checks to layers, and the failure modes the pipeline was designed against. * **Creation**: Recorded the [AI review setup](decisions/ai-review-setup.md): the `opencode-go` provider prefix (the wrong one failed every run with "Insufficient account funds"), the v2 installer (v1 cannot resolve model variants), env-only auth after the prompt-injection finding, the read-only permission set, and the verdict line as the posting gate. * **Creation**: Recorded [release engineering](decisions/release-engineering.md): release-please in manifest mode, plain `v`-tags, the `version` field requirement, and why conventional commits are enforced at commit time. * **Creation**: Recorded [Snyk paused and secrets policy](decisions/snyk-paused-and-secrets-policy.md): quota-paused job kept in the workflow, and the `write:issue` scope requirement for all workflow tokens (learned from a 403 that named its missing scope). * **Update**: Rewrote [index](/index.md) to group Docker-era and CI-era decisions and added a one-paragraph history. ## 2026-10-18 * **Creation**: Established the Nuxt bundle [index](/index.md) and the [decisions](/decisions/index.md) index. * **Decision**: Recorded the [dev container shape](/decisions/dev-container-shape.md) — source bind-mounted, dependencies in the image, run as the image's `node` user. * **Decision**: Recorded [build layer order](/decisions/build-layer-order.md) and the [build context filter](/decisions/build-context-filter.md), including the measured failure when a `COPY` target is excluded. * **Decision**: Recorded the [repository and compose layout](/decisions/repo-and-compose-layout.md) and the [mockup layout](/decisions/mockup-layout.md). * **Creation**: Added [open questions](/open-questions.md), including the shape of the production image.