Skip to content

CLAUDE.md

Latest commit

 

History

History
221 lines (159 loc) · 27.1 KB

CLAUDE.md

File metadata and controls

221 lines (159 loc) · 27.1 KB
Raw

CLAUDE.md

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

Project

Personal portfolio + MDX blog at ahnafnafee.dev. Next.js 16 (App Router) + React 19 + TypeScript + Tailwind CSS v4. Deployed to Vercel; also supports a stripped-down static export (used for a university-hosted mirror).

Common Commands

Yarn is the package manager — yarn.lock is committed. Don't use npm install.

yarn dev                # dev server (Turbopack, env from .env.development.local)
yarn dev:webpack        # dev server forced to webpack (use if Turbopack misbehaves with mermaid/MDX)
yarn build              # production build (Vercel)
yarn export             # static export — runs export:prepare → next build → export:cleanup
yarn export:university  # static export with .env.development.static.local (BASE_PATH-aware)
yarn lint               # eslint . (flat config in eslint.config.mjs)
yarn lint:fix           # eslint . --fix
yarn type-check         # tsc --noEmit
yarn test               # vitest run (unit + integration suite)
yarn test:watch         # vitest in watch mode
yarn test:coverage      # vitest with v8 coverage report
yarn format             # prettier write across js/ts/tsx/md/mdx/json
yarn analyze            # bundle analyzer (cross-env ANALYZE=true next build)
yarn validate:json-ld   # walk built HTML, verify every <script type="application/ld+json">
yarn validate:sitemap   # verify public/sitemap.xml exists, parses, and uses canonical URLs
yarn audit:alt-text     # scan MDX for weak <ContentImage>/<img> alt attributes
yarn generate:sitemap   # regenerate public/sitemap.xml via custom-next-sitemap.js (chained into build)
yarn generate:llms      # regenerate public/llms.txt + public/llms-full.txt (chained into build)
npx tsx indexing/sendIndexingRequest.ts   # batch-submit URLs to Google Indexing API

Tests live under __tests__ directories co-located with the code they cover (e.g. src/libs/seo/__tests__/, src/services/content/__tests__/). The Vitest config (vitest.config.ts) uses happy-dom and resolves @/* paths via Vite's native resolve.tsconfigPaths: true. CI runs the suite on every push/PR via .github/workflows/ci.yml.

Content Pipeline (the core architecture)

All blog, portfolio, and research entries are MDX files on disk — there's no CMS.

  1. Source files live in src/data/blog/*.mdx, src/data/portfolio/*.mdx, and src/data/research/*.mdx. Filename (minus .mdx) becomes the URL slug.
  2. Frontmatter is parsed by gray-matter in src/services/content/. The shape is enforced by ambient types in src/types/index.d.ts (declare module 'me') — Blog, Portfolio, Research (plus its sub-types Author, Affiliation, Venue, ResearchLinks, ResearchIdentifiers), Snippet, etc. Add new content fields there first, then update the MDX files.
  3. Reading content:
    • getContents<T>('/blog') — list all entries (used by generateStaticParams and listing pages).
    • getContentBySlug<T>('/blog', slug) — single entry.
    • getContentHeaders<T>('/blog') — header-only (skips MDX body parsing) — use this whenever a page only needs frontmatter.
    • All three resolve paths through LOCATION_DIR in src/services/directory/location.ts (= src/data). Change content location there, not in the readers.
  4. Rendering uses next-mdx-remote/rsc's MDXRemote in src/app/blog/[slug]/page.tsx and the portfolio / research equivalents. Plugin set is centralized in src/libs/mdxConfig.ts (remark-gfm, remark-math, rehype-prism-plus, rehype-slug, rehype-katex).
  5. Custom MDX components live in src/components/content/mdx/ and are wired into the MDXComponents map in mdx/index.ts. Notable overrides:
    • Pre, Code — code blocks with prism syntax highlighting.
    • Mermaid — renders ```mermaid blocks; mermaid + stylis are transpiled and pre-bundled (see transpilePackages and experimental.optimizePackageImports in next.config.js).
    • ContentImage — explicit component for ImageKit-hosted images. The default img MDX mapping was intentionally removed so external images / SVGs in MDX don't break; use <ContentImage> (or a raw <img>) explicitly.
    • MDXLink, Blockquote, Table, Headings — styled overrides.
    • Interactive features register through mdx/interactive.ts. Compose the shared panel and choice controls in src/components/content/interactive/; keep feature data and behavior in its own content folder. See that folder's README for the addition steps.
  6. Research-specific components live in src/components/content/research/:
    • Detail-page: HeadingResearch (status chip, title, structured authors + affiliations, venue, action-button row, hr — author/affiliation <sup> markers are conditional + carry title tooltips, see Conventions), ResearchTeaser (full-bleed figure + caption, lives outside prose so it can breathe), ResearchAbstract (purple-accented not-prose card), ResearchBibTeX (anchored at #bibtex, copy-to-clipboard, mirrors Pre.tsx chrome).
    • Listing-page: ResearchOverview (intro paragraph), ResearchNews (date-column timeline; entries live in the NEWS array inside the file), ResearchAreas (colored chip row), ResearchItem + ResearchSections (listing card + section grouping). ComingSoonImage is a pastel-gradient placeholder used in place of the listing thumbnail when an entry has comingSoon: true.
    • Shared: SectionHeading is the bordered text-lg md:text-xl font-bold text-black dark:text-white h2 used by every section on the listing page (Overview / News / Research Areas / Sections groupers). Use it for any new sections to keep the rhythm consistent.
    • The detail page composition is in src/app/research/[slug]/page.tsx. The listing page is in src/app/research/page.tsx — both pass pt-6 md:pt-10 to AppLayoutPage to give the first section breathing room from the sticky header.
  7. SEO is generated per-page in src/app/blog/[slug]/page.tsx and the portfolio / research equivalents: generateMetadata builds Next.js Metadata (OG, Twitter, canonical, modifiedTime), and the page itself injects a single @graph JSON-LD block:
    • Blog: BlogPosting + WebPage + BreadcrumbList
    • Portfolio: SoftwareSourceCode + WebPage + BreadcrumbList
    • Research: ScholarlyArticle + WebPage + BreadcrumbList. The first author (matching SITE_AUTHOR.name) maps to {'@id': PERSON_ID}; coauthors emit inline Person nodes with affiliation and ORCID identifier. isPartOf resolves to Periodical for published/accepted/workshop venues and CreativeWork otherwise. DOI / arXiv / ResearchGate IDs from identifiers become PropertyValue entries on identifier. The Person @id (https://www.ahnafnafee.dev/#person) is referenced from author/publisher; the full Person node is emitted only on the home page. When you add a new content type, mirror this pattern.
  8. OG image fallback: if a post has no thumbnail, generateOgImage from src/libs/metapage constructs one via /api/og. For research listing cards, the resolution order is thumbnail > teaser > generated OG; ImageKit URLs without an explicit ?tr= get ?tr=w-600 appended so retina displays render sharp on the 176px max slot. The isImagekitUrl check parses new URL(...).hostname (CodeQL flagged the substring form as bypassable).

Path Aliases

tsconfig.json baseUrl is ./src. Use:

  • @/* → src/*

Imports are auto-sorted by @ianvs/prettier-plugin-sort-imports (order defined in .prettierrc.js): @/components → @/services → @/libs → @/... → relative → third-party → @/styles/*. Empty strings in importOrder produce blank-line separators between groups. Don't reshuffle by hand; let Prettier do it.

Component Layout

Two roots under src/components/:

  • ui/ — shadcn/ui primitives. Every file here is generated by npx shadcn@latest add <name> and follows shadcn's data-slot, variant, and class conventions. Don't edit by hand unless intentionally extending; instead use npx shadcn@latest add <name> --diff to merge upstream changes. Built on radix-ui (configured in components.json as style: radix-nova); radix-ui is not a competing UI system — it's shadcn's primitive layer.
  • site/ — bespoke composition components specific to this site (Header, Footer, Nav, AppLayoutPage, Hero, Searchbar, BackToTop, EmptyResult, MobileNav, ThemeMenu, links, image wrappers). They compose ui/ primitives + project state.

When you need a new primitive, search shadcn first (npx shadcn@latest search). Only fall back to a bespoke component in site/ (or a feature folder) if no primitive fits.

Theme System

Two layers feed Tailwind v4:

  1. tailwind.config.js keeps the legacy palette scales — primary: colors.blue, theme: colors.neutral, plus a custom gray-{0..900} ramp. Existing classes like bg-primary-500, text-theme-700, dark:bg-gray-900 resolve through this. Bridged into v4 via @config "../../tailwind.config.js" in globals.css.
  2. globals.css @theme inline + :root / .dark blocks define the shadcn semantic tokens (--background, --foreground, --primary, --muted, --border, --ring, etc.) and intentionally map them to the same OKLCH values as the legacy scales:
    • --primary ↔ blue-500 (light) / blue-400 (dark) — same as legacy primary-500/400.
    • --background ↔ neutral-50 (light) / custom gray-900 #111 (dark) — matches legacy body bg.
    • --foreground ↔ neutral-800/100, --muted ↔ neutral-100/800, --border ↔ neutral-200/700, --ring ↔ blue-500/400 (legacy focus ring).

Net result: bg-primary (shadcn) and bg-primary-500 (legacy) render the same color; both surfaces remain valid during the transition. When adding new code, prefer the semantic tokens. Don't write dark: overrides for shadcn classes — --background etc. switch automatically via .dark.

Fonts: --font-sans resolves to 'Google Sans', system-ui, -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, 'Helvetica Neue', Arial, 'Noto Sans', sans-serif, 'Apple Color Emoji', 'Segoe UI Emoji', 'Segoe UI Symbol', 'Noto Color Emoji'. Google Sans is loaded via Google Fonts CSS (<link> in layout.tsx head, with preconnect to fonts.gstatic.com — font-src in next.config.js CSP must include it). Local Inter @font-face stays in globals.css as a deeper fallback so the page never FOUCs. tailwind.config.js fontFamily.primary mirrors the same stack so the legacy font-primary utility picks up Google Sans too.

Two Build Modes

yarn build is the standard Vercel build. yarn export produces a static site — but Next.js can't static-export a tree with API routes, so the export pipeline temporarily moves src/app/api/ out:

  • scripts/export-prepare.js copies src/app/api/ → temp_api_backup/ and deletes the original.
  • next build runs with STATIC_EXPORT=true (read in next.config.js to flip output: 'export', images.unoptimized: true, and apply BASE_PATH if set).
  • scripts/export-cleanup.js restores the API directory from the backup.

If a build crashes mid-export, temp_api_backup/ may be left behind — restore it manually before the next run, otherwise the API routes are gone.

yarn export:university is the same flow with .env.development.static.local, which sets a BASE_PATH for hosting under a sub-path.

Revalidation

ISR endpoint at GET /api/revalidate?secret=<SECRET>&slug=/blog/<slug>. Secret is NEXT_PUBLIC_SECRET or SECRET_KEY. The slug parameter is the actual path (e.g. /blog/local-llm-pdf-ocr), not the route template.

Sitemap

custom-next-sitemap.js runs as the generate:sitemap step, chained explicitly at the end of build. Do not move it back to a postbuild lifecycle hook: Yarn (used by CI and Vercel) does not run npm-style pre/post scripts, so the sitemap silently never generates and /sitemap.xml serves the HTML 404 page. yarn validate:sitemap is the CI gate that catches this. The config allowlists AI/LLM crawlers (GPTBot, Claude-Web, anthropic-ai, etc.), assigns per-route priorities, and skips robots.txt generation under STATIC_EXPORT=true. Edit this file (not a separate sitemap config) when adding new top-level routes.

llms.txt

scripts/generate-llms-txt.js writes two files, chained into build as generate:llms:

  • public/llms.txt — curated index: site metadata, the API/RSS surface, and one entry per blog post, research entry, and portfolio project (title, URL, summary, topics).
  • public/llms-full.txt — every document's MDX body inlined verbatim under an H1 and a URL: line, so an assistant can answer from one fetch instead of crawling each page. MDX bodies carry no H1 of their own (the page template renders the frontmatter title), which is what keeps the concatenated heading hierarchy intact.

Both are tracked in git and regenerated at build time. The committed copies are the fallback; the build output is what actually ships. CI runs git diff --exit-code on them after the build, so a content edit without a regenerate fails the run. Run yarn generate:llms and commit the result whenever you add or edit an MDX entry.

next.config.js Quirks

  • Wrapped in both withAxiom (logs) and withPWA (@ducanh2912/next-pwa — the maintained next-pwa fork, Workbox 7, disabled in dev). Options nest under workboxOptions (skipWaiting, buildExcludes); extendDefaultRuntimeCaching: true preserves the legacy next-pwa@5 cache strategies. Note: @serwist/next was evaluated but doesn't generate sw.js under Next 16's Turbopack-default build (tracking serwist#54) — revisit when stable.
  • Strict CSP including *.youtube.com, giscus.app, *.googletagmanager.com, cdn.jsdelivr.net.
  • Image remotePatterns allowlist — when adding a new external image host, register it here or next/image will reject it.
  • transpilePackages: ['mermaid', 'stylis'] — leave it; mermaid ships ESM that Next can't consume otherwise.
  • Several legacy URL redirects (/projects → /portfolio, /cv → /AhnafAnNafeeResume.pdf, etc.) live in redirects().

Conventions

  • Prettier: no semicolons, 2-space indent, single quotes, 120-col print width, trailingComma: 'none'. Import sorting via @ianvs/prettier-plugin-sort-imports; tailwind class sorting via prettier-plugin-tailwindcss. ESLint extends next/core-web-vitals + @typescript-eslint/recommended; no-explicit-any is intentionally off.

  • shadcn rules: forms compose Field + FieldLabel + FieldDescription (no raw div + Label); inputs with leading/trailing icons use InputGroup + InputGroupAddon; icons inside Button get data-icon="inline-start" / inline-end (no size-* classes); use cn() for conditional classes; reach for <Empty>, <Skeleton>, <Alert>, <Badge>, <Separator> instead of styled divs. Toasts via sonner (toast() from sonner, <Toaster> from @/components/ui/sonner).

  • Commits go through commitizen with cz-conventional-changelog — run yarn commit for the prompt. lint-staged runs Prettier on staged js/jsx/ts/tsx/mdx/md/css files.

  • Frontmatter published for blog and research is MM/DD/YYYY (portfolio uses date). featured: true surfaces an entry on the home page. Optional updated: "MM/DD/YYYY" overrides published as the freshness signal — it flows into dateModified (BlogPosting / ScholarlyArticle JSON-LD), <meta property="article:modified_time">, and the sitemap lastmod. Set it whenever you make a substantive content edit; leave it off for typo fixes.

  • Authoring convention: open every blog post with a 60–120 word TL;DR paragraph immediately under the H1. LLM crawlers (ChatGPT, Perplexity, AI Overviews) draw ~44% of their citations from the first 30% of an article, so front-loading the answer compounds visibility.

  • Research authoring: use the YAML folded scalar (abstract: >-) — block scalar (|) preserves newlines and the abstract renders with broken lines. Authors use a structured array; affiliations indices are 1-based and reference the entry's affiliations array. section drives listing-page grouping (top-tier / conferences / journals / workshops / others). new: true renders a NEW badge inline with the title on the listing card. comingSoon: true swaps the listing thumbnail for the ComingSoonImage pastel placeholder (no thumbnail/teaser required). Authors matching SITE_AUTHOR.name are auto-bolded across HeadingResearch and ResearchItem.

  • Author / affiliation superscripts in HeadingResearch are conditional. Each marker only renders when it actually disambiguates something:

    • * (corresponding author) — corresponding: true on the author. Shows when 2+ authors AND at least one is corresponding.
    • † (equal contribution) — equalContribution: true on the author. Shows when 2+ authors AND at least 2 are flagged. Renders on every flagged author.
    • ‡ (principal investigator) — principalInvestigator: true on the author. Shows when 2+ authors AND at least one is PI.
    • ¹ ² ³ (affiliation indices) — derived from affiliations: [...] on each author. Shows when 2+ affiliations exist on the entry.

    Marker order on each author follows academic convention: *†‡ then numeric indices. The affiliation legend below the author line drops its leading <sup> under the same showAffSup rule. Single-author / single-affiliation entries collapse to clean text — no orphan markers. Each <sup> has cursor-help + a native title tooltip (affiliation index sups resolve to the full affiliation name; symbol sups resolve to their caption text). A combined caption line appears below the affiliation row when any symbol marker is shown — e.g. *Corresponding author · †Equal contribution · ‡Principal investigator — joined by ·.

Content Authoring Workflow

How to write a new blog / portfolio / research entry without re-deriving conventions from scratch. Read this before any "write me a blog post about X" task.

  1. File locations. Blog → src/data/blog/*.mdx. Portfolio → src/data/portfolio/*.mdx. Research → src/data/research/*.mdx. Filename (minus .mdx) becomes the URL slug — kebab-case, short, topical (3–4 words max, no date prefix).

  2. Tone anchors. First-person, conversational-technical, title-case headings, 850–1,400 words. Before drafting, read 1–2 thematically adjacent posts in full to calibrate voice. Reference posts: local-llm-pdf-ocr.mdx (deep technical, ~1,400w), autograder-architecture.mdx (architecture, ~850w), doi-paper-scraper.mdx (tooling), modelfile-syntax-extension.mdx (developer tool launch). Calibrate voice, not punctuation: the anchor posts lean heavily on em dashes, but new posts must NOT (see #11).

  3. TL;DR rule. Open every blog post with a 60–120 word TL;DR paragraph immediately under the frontmatter (no # Title markdown — the page template renders frontmatter title as H1). LLM crawlers pull ~44% of citations from the first 30%, so front-loading the punch is non-negotiable.

  4. Frontmatter shape (Blog). Mirror existing posts:

    title: 'Bold sentence-case title'
summary: "Single-sentence summary, doubles as meta description."
featured: true                          # surfaces on home page
author_name: 'Ahnaf An Nafee'
github_username: 'ahnafnafee'
published: 'MM/DD/YYYY'                 # today's date
topics: ['Topic 1', 'Topic 2', ...]     # 4–6 broad categories
keywords: ['kw1', 'kw2', ...]           # 30–40 lowercase entries
thumbnail: null                         # or ImageKit URL — see #7
related: []                             # other slugs to cross-link

    Research entries are richer (authors[], affiliations[], venue, links, identifiers, structured author flags) — see Research in src/types/index.d.ts and existing entries in src/data/research/. Research abstracts MUST use the YAML folded scalar abstract: >- (block scalar | preserves newlines and renders broken).

  5. Body structure. Title-case headings. One Mermaid diagram per major architectural section is on-brand (use ```mermaid fences — handled by Mermaid.tsx). Code blocks framed by prose, not by intro text ("Here's an example:" — avoid). Tables fine for reference data. Raw <img src=... className='w-full rounded-lg' /> for in-body images — the default <img> MDX mapping was intentionally removed.

  6. MDX components available. From src/components/content/mdx/: Pre, Code, Mermaid, ContentImage, MDXLink, Blockquote, Table, Headings. Plugin set: remark-gfm, remark-math + rehype-katex ($...$ and $$...$$), rehype-prism-plus, rehype-slug. Prism languages include js/ts/tsx/jsx/py/bash/json/yaml/css/html/sql/rust/go/java/c/cpp — unknown languages render as plain text.

  7. Thumbnail image generation (canonical theme). The thumbnail: field is either an ImageKit URL at https://ik.imagekit.io/8ieg70pvks/blog/<slug>.{jpg|jpeg|png} or null to fall back to /api/og. For AI-generated thumbnails (ChatGPT Images 2.0 / GPT Image / DALL·E 3), use the "Editorial Tech Illustration" theme:

    • Style anchor: Flat editorial tech illustration — Stripe blog / Vercel changelog / Linear docs feel. Geometric, clean, modern. No hand-drawn squiggles, no sticker / doodle aesthetic, no 3D renders, no photorealism, no whiteboard textures.

    • Background: Warm off-white #FAF9F6 with a barely-visible dotted grid for subtle texture.

    • Composition: Title top-left in bold sans-serif (#1F2937, ≤ 2 lines, 8% padding from edges); byline Ahnaf An Nafee directly below in regular weight (#6B7280, 70% title size); hero illustration occupying the center 60% horizontally centered; bottom 10% breathing room.

    • Palette (strict): only these colors —

      Hex Role
      #FAF9F6 Background
      #3B82F6 Blue primary (main brand accent)
      #8B5CF6 Purple accent
      #F59E0B Orange accent
      #10B981 Green accent (optional — "growth" / "success" themes)
      #EF4444 Warning red (sparingly — error / warning semantics only)
      #1F2937 Dark text
      #6B7280 Muted text / connector lines
      #FFFFFF Card surfaces
      #E5E7EB Card borders

    • Draw style: Flat shapes, no gradients, no glow, subtle drop shadows (max 8% opacity), 1.5–2px line weights, 8px rounded corners, generous whitespace.

    • Aspect ratio: 16:9 at 1200×630. The listing-card and post-detail thumbnail containers crop non-16:9 images on narrow viewports (mobile), so tighter aspect ratios like 4:3 / 3:2 lose top/bottom content. Note: 16:9 also matches the OG card spec, but thumbnails and OG cards are still separate concerns — when thumbnail: is null the OG image is generated programmatically via /api/og (see src/libs/metapage generateOgImage); when set, the same image is used for both the listing card and the OG card.

    • Universal template prompt (fill in {TITLE} and {HERO}):

      Create a 1200×630 wide thumbnail image (16:9 aspect ratio) for a developer blog post titled "{TITLE}" by Ahnaf An Nafee. Style: flat editorial tech illustration in the spirit of Stripe blog, Vercel changelog, or Linear docs — geometric, clean, modern. Background: warm off-white #FAF9F6 with a faint dotted-grid texture. Title text in bold sans-serif (#1F2937), top-left aligned with 8% padding, 2 lines maximum. Byline "Ahnaf An Nafee" directly below the title in regular-weight sans-serif (#6B7280), 70% the title size. Hero illustration occupying the center-right 60% of the canvas (16:9 is wide — use the horizontal space): {HERO}. Strict palette — only these colors: #FAF9F6 (bg), #3B82F6 (blue), #8B5CF6 (purple), #F59E0B (orange), #10B981 (green, only for growth themes), #EF4444 (red, only for warning semantics), #1F2937 (dark text), #6B7280 (muted/lines), #FFFFFF with #E5E7EB borders (cards). Flat shapes only — no gradients, no glow, subtle drop shadows (max 8% opacity), 1.5–2px line weights, 8px rounded corners, generous whitespace. Avoid: hand-drawn squiggles, sticker / doodle style, 3D depth, photorealism, real readable code, whiteboard textures, brand logos other than abstract representations.

    • Retroactive application: The earlier thumbnails (local-llm-pdf-ocr, doi-paper-scraper, autograder-architecture) are in an older hand-drawn whiteboard style and should be re-generated under this theme when bandwidth allows. Workflow: generate the new image → upload to ImageKit at blog/<slug>.{jpg|png} → update the post's thumbnail: field with the new URL (include ?updatedAt=<unix-ms> for cache-bust) → optionally set updated: 'MM/DD/YYYY' to bump the freshness signal in the sitemap and dateModified JSON-LD.

  8. Related posts. related: [] is the default. When adding a new post, optionally scan for natural backlinks and update older posts' related: arrays — polish step, don't gate on it.

  9. Pre-write checklist. Glob src/data/blog/*.mdx → read closest anchor post → confirm Blog type in src/types/index.d.ts → fetch repo README if applicable (gh api repos/<owner>/<repo>/readme --jq '.content' | base64 -d) → confirm today's date for published.

  10. Verify checklist. yarn type-check → yarn lint → yarn dev → visit /blog/<slug> → confirm TL;DR renders under H1, headings have anchors, Mermaid renders, code blocks colored → yarn build → yarn validate:json-ld → yarn audit:alt-text if any <img> / <ContentImage> → word count in 850–1,400 range → TL;DR is 60–120 words (count it).

  11. What NOT to do. No # Title markdown (page template handles H1). No "In this post, I'll explain..." filler — open with the punch. No keyword stuffing without body coverage. No fabricated stats / version numbers / model names — verify against linked sources. No paragraph-length comments inside code blocks — let prose do the work. No em dashes (—, U+2014) and no AI-writing tells. Avoid dramatic reveal-colons, "that's the point", "the clever bit", "refreshingly X", "isn't just X, it's Y", rule-of-three flourishes, and anthropomorphizing the code. Draft in plain punctuation (commas, periods, parentheses, the occasional functional colon), then scan the finished .mdx for the em-dash codepoint (U+2014) and rewrite any that slipped in. The anchor posts in #2 use em dashes heavily; do not copy that habit.

Indexing Helper

indexing/sendIndexingRequest.ts reads URLs from the live sitemap, checks Search Console for index status, and submits unindexed URLs to the Google Indexing API. Requires indexing/service_account.json (gitignored) with Indexing API + Search Console API enabled and the service account added as an Owner in Search Console. The indexing/ directory is excluded from the TypeScript build (tsconfig.json).

Memory

There is per-project memory at C:\Users\ahnaf\.claude\projects\D--GitHub-ahnafnafee-dev\memory\ — check MEMORY.md there at the start of a session for any user feedback or environment notes (e.g. gh CLI is installed and preferred over curl/WebFetch for any GitHub work).