feat: add Markdown versions of public pages

[celeroncoder]

What changed

• serve Markdown versions at /.md, /index.md, /blog.md, and /blog/.md
• convert Payload Lexical blog content to Markdown, including fenced code blocks
• advertise each Markdown page with a text/markdown alternate link
• keep homepage Markdown in sync through shared portfolio content
• return Markdown 404 and 500 responses with the correct content type

Verification

• ./node_modules/.bin/tsc --noEmit
• ./node_modules/.bin/next build
• HTTP check confirmed .md rewrites return text/markdown and Markdown error bodies

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
celeroncoder-web	Ready	Preview	Aug 25, 2026 7:27am

[github-actions]

React Doctor found no new issues. 🎉

_{Reviewed by React Doctor for commit f54d742.}

[celeroncoder]

Hermes reviewer is reviewing this PR and will post shortly.

[celeroncoder]

FOLLOW-UP · P1: 0 · P2: 1 · P3: 1

P2 · lib/markdown-pages.ts:131-134 (renderBlogPostMarkdown) — Trigger: a post has a heroImage whose url is server-relative (Payload's vercel-blob/media adapter returns paths like /api/media/file/..., and the code already treats non-object heroImages as absent, so both shapes flow through here). Broken behavior: the image is emitted as ![alt](/api/media/file/...) inside a document served from /blog/<slug>.md. Any consumer resolving that link against the document URL (standard Markdown resolution, curl-based scrapers, most LLM fetch pipelines) requests /blog/api/media/file/..., which 404s. Consequence: every post with a hero image ships a broken image reference in exactly the machine-readable surface this PR exists to provide. Fix: run heroImage.url through absoluteUrl(origin, ...) like the other links.

P3 · app/(frontend)/llm-markdown/[[...path]]/route.ts:75-82 — Trigger: a legacy post row with an empty slug (the column is optional in collections/Posts.ts; the auto-slug hook only fills it on new saves). Broken behavior: postMarkdownUrl builds /blog/.md into the home and blog-index listings. Consequence: dead alternate link pointing at a Markdown 404 until the post is re-saved in admin. Low probability given slugs are unique/admin-managed, hence P3.

What else got checked: proxy matcher correctly skips api, admin, _next/*; /.md, /index.md, /blog.md, /blog/<slug>.md all map cleanly through the optional catch-all including trailing-slash and unknown-path cases returning proper 404 Markdown; error responses get no-store plus X-Robots-Tag: noindex; Lexical conversion guards malformed rich text and falls back to a typed 500; blog-index limit of 50 matches the HTML page so no drift between mirrors; component refactor to site-content.ts preserves all copy verbatim (footer GitHub repo URL still points at celeroncoder.tech, pre-existing); new Date().getFullYear() in footer was already suppressHydrationWarning.

Checks: GitGuardian pass; react-doctor and Vercel pending at review time.

verify:curl -s https://<preview>/blog/<slug-with-hero-image>.md | grep '^!\[' — confirm the image URL is absolute, then fetch it and expect 200 instead of 404.

[celeroncoder]

What changed

• serve Markdown versions at /.md, /index.md, /blog.md, and /blog/.md
• convert Payload Lexical blog content to Markdown, including fenced code blocks
• advertise each Markdown page with a text/markdown alternate link
• keep homepage Markdown in sync through shared portfolio content
• return Markdown 404 and 500 responses with the correct content type

Verification

• ./node_modules/.bin/tsc --noEmit
• ./node_modules/.bin/next build
• HTTP check confirmed .md rewrites return text/markdown and Markdown error bodies

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
celeroncoder-web	Ready	Preview	Aug 25, 2026 7:27am

[github-actions]

React Doctor found no new issues. 🎉

_{Reviewed by React Doctor for commit f54d742.}

[celeroncoder]

Hermes reviewer is reviewing this PR and will post shortly.

[celeroncoder]

FOLLOW-UP · P1: 0 · P2: 1 · P3: 1

P2 · lib/markdown-pages.ts:131-134 (renderBlogPostMarkdown) — Trigger: a post has a heroImage whose url is server-relative (Payload's vercel-blob/media adapter returns paths like /api/media/file/..., and the code already treats non-object heroImages as absent, so both shapes flow through here). Broken behavior: the image is emitted as ![alt](/api/media/file/...) inside a document served from /blog/<slug>.md. Any consumer resolving that link against the document URL (standard Markdown resolution, curl-based scrapers, most LLM fetch pipelines) requests /blog/api/media/file/..., which 404s. Consequence: every post with a hero image ships a broken image reference in exactly the machine-readable surface this PR exists to provide. Fix: run heroImage.url through absoluteUrl(origin, ...) like the other links.

P3 · app/(frontend)/llm-markdown/[[...path]]/route.ts:75-82 — Trigger: a legacy post row with an empty slug (the column is optional in collections/Posts.ts; the auto-slug hook only fills it on new saves). Broken behavior: postMarkdownUrl builds /blog/.md into the home and blog-index listings. Consequence: dead alternate link pointing at a Markdown 404 until the post is re-saved in admin. Low probability given slugs are unique/admin-managed, hence P3.

What else got checked: proxy matcher correctly skips api, admin, _next/*; /.md, /index.md, /blog.md, /blog/<slug>.md all map cleanly through the optional catch-all including trailing-slash and unknown-path cases returning proper 404 Markdown; error responses get no-store plus X-Robots-Tag: noindex; Lexical conversion guards malformed rich text and falls back to a typed 500; blog-index limit of 50 matches the HTML page so no drift between mirrors; component refactor to site-content.ts preserves all copy verbatim (footer GitHub repo URL still points at celeroncoder.tech, pre-existing); new Date().getFullYear() in footer was already suppressHydrationWarning.

Checks: GitGuardian pass; react-doctor and Vercel pending at review time.

verify:curl -s https://<preview>/blog/<slug-with-hero-image>.md | grep '^!\[' — confirm the image URL is absolute, then fetch it and expect 200 instead of 404.

[celeroncoder]

What changed

• serve Markdown versions at /.md, /index.md, /blog.md, and /blog/.md
• convert Payload Lexical blog content to Markdown, including fenced code blocks
• advertise each Markdown page with a text/markdown alternate link
• keep homepage Markdown in sync through shared portfolio content
• return Markdown 404 and 500 responses with the correct content type

Verification

• ./node_modules/.bin/tsc --noEmit
• ./node_modules/.bin/next build
• HTTP check confirmed .md rewrites return text/markdown and Markdown error bodies

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
celeroncoder-web	Ready	Preview	Aug 25, 2026 7:27am

[github-actions]

React Doctor found no new issues. 🎉

_{Reviewed by React Doctor for commit f54d742.}

[celeroncoder]

Hermes reviewer is reviewing this PR and will post shortly.

[celeroncoder]

FOLLOW-UP · P1: 0 · P2: 1 · P3: 1

P2 · lib/markdown-pages.ts:131-134 (renderBlogPostMarkdown) — Trigger: a post has a heroImage whose url is server-relative (Payload's vercel-blob/media adapter returns paths like /api/media/file/..., and the code already treats non-object heroImages as absent, so both shapes flow through here). Broken behavior: the image is emitted as ![alt](/api/media/file/...) inside a document served from /blog/<slug>.md. Any consumer resolving that link against the document URL (standard Markdown resolution, curl-based scrapers, most LLM fetch pipelines) requests /blog/api/media/file/..., which 404s. Consequence: every post with a hero image ships a broken image reference in exactly the machine-readable surface this PR exists to provide. Fix: run heroImage.url through absoluteUrl(origin, ...) like the other links.

P3 · app/(frontend)/llm-markdown/[[...path]]/route.ts:75-82 — Trigger: a legacy post row with an empty slug (the column is optional in collections/Posts.ts; the auto-slug hook only fills it on new saves). Broken behavior: postMarkdownUrl builds /blog/.md into the home and blog-index listings. Consequence: dead alternate link pointing at a Markdown 404 until the post is re-saved in admin. Low probability given slugs are unique/admin-managed, hence P3.

What else got checked: proxy matcher correctly skips api, admin, _next/*; /.md, /index.md, /blog.md, /blog/<slug>.md all map cleanly through the optional catch-all including trailing-slash and unknown-path cases returning proper 404 Markdown; error responses get no-store plus X-Robots-Tag: noindex; Lexical conversion guards malformed rich text and falls back to a typed 500; blog-index limit of 50 matches the HTML page so no drift between mirrors; component refactor to site-content.ts preserves all copy verbatim (footer GitHub repo URL still points at celeroncoder.tech, pre-existing); new Date().getFullYear() in footer was already suppressHydrationWarning.

Checks: GitGuardian pass; react-doctor and Vercel pending at review time.

verify:curl -s https://<preview>/blog/<slug-with-hero-image>.md | grep '^!\[' — confirm the image URL is absolute, then fetch it and expect 200 instead of 404.

[celeroncoder]

What changed

• serve Markdown versions at /.md, /index.md, /blog.md, and /blog/.md
• convert Payload Lexical blog content to Markdown, including fenced code blocks
• advertise each Markdown page with a text/markdown alternate link
• keep homepage Markdown in sync through shared portfolio content
• return Markdown 404 and 500 responses with the correct content type

Verification

• ./node_modules/.bin/tsc --noEmit
• ./node_modules/.bin/next build
• HTTP check confirmed .md rewrites return text/markdown and Markdown error bodies

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
celeroncoder-web	Ready	Preview	Aug 25, 2026 7:27am

[github-actions]

React Doctor found no new issues. 🎉

_{Reviewed by React Doctor for commit f54d742.}

[celeroncoder]

Hermes reviewer is reviewing this PR and will post shortly.

[celeroncoder]

FOLLOW-UP · P1: 0 · P2: 1 · P3: 1

P2 · lib/markdown-pages.ts:131-134 (renderBlogPostMarkdown) — Trigger: a post has a heroImage whose url is server-relative (Payload's vercel-blob/media adapter returns paths like /api/media/file/..., and the code already treats non-object heroImages as absent, so both shapes flow through here). Broken behavior: the image is emitted as ![alt](/api/media/file/...) inside a document served from /blog/<slug>.md. Any consumer resolving that link against the document URL (standard Markdown resolution, curl-based scrapers, most LLM fetch pipelines) requests /blog/api/media/file/..., which 404s. Consequence: every post with a hero image ships a broken image reference in exactly the machine-readable surface this PR exists to provide. Fix: run heroImage.url through absoluteUrl(origin, ...) like the other links.

P3 · app/(frontend)/llm-markdown/[[...path]]/route.ts:75-82 — Trigger: a legacy post row with an empty slug (the column is optional in collections/Posts.ts; the auto-slug hook only fills it on new saves). Broken behavior: postMarkdownUrl builds /blog/.md into the home and blog-index listings. Consequence: dead alternate link pointing at a Markdown 404 until the post is re-saved in admin. Low probability given slugs are unique/admin-managed, hence P3.

What else got checked: proxy matcher correctly skips api, admin, _next/*; /.md, /index.md, /blog.md, /blog/<slug>.md all map cleanly through the optional catch-all including trailing-slash and unknown-path cases returning proper 404 Markdown; error responses get no-store plus X-Robots-Tag: noindex; Lexical conversion guards malformed rich text and falls back to a typed 500; blog-index limit of 50 matches the HTML page so no drift between mirrors; component refactor to site-content.ts preserves all copy verbatim (footer GitHub repo URL still points at celeroncoder.tech, pre-existing); new Date().getFullYear() in footer was already suppressHydrationWarning.

Checks: GitGuardian pass; react-doctor and Vercel pending at review time.

verify:curl -s https://<preview>/blog/<slug-with-hero-image>.md | grep '^!\[' — confirm the image URL is absolute, then fetch it and expect 200 instead of 404.

[celeroncoder]

What changed

• serve Markdown versions at /.md, /index.md, /blog.md, and /blog/.md
• convert Payload Lexical blog content to Markdown, including fenced code blocks
• advertise each Markdown page with a text/markdown alternate link
• keep homepage Markdown in sync through shared portfolio content
• return Markdown 404 and 500 responses with the correct content type

Verification

• ./node_modules/.bin/tsc --noEmit
• ./node_modules/.bin/next build
• HTTP check confirmed .md rewrites return text/markdown and Markdown error bodies

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
celeroncoder-web	Ready	Preview	Aug 25, 2026 7:27am

[github-actions]

React Doctor found no new issues. 🎉

_{Reviewed by React Doctor for commit f54d742.}

[celeroncoder]

Hermes reviewer is reviewing this PR and will post shortly.

[celeroncoder]

FOLLOW-UP · P1: 0 · P2: 1 · P3: 1

P2 · lib/markdown-pages.ts:131-134 (renderBlogPostMarkdown) — Trigger: a post has a heroImage whose url is server-relative (Payload's vercel-blob/media adapter returns paths like /api/media/file/..., and the code already treats non-object heroImages as absent, so both shapes flow through here). Broken behavior: the image is emitted as ![alt](/api/media/file/...) inside a document served from /blog/<slug>.md. Any consumer resolving that link against the document URL (standard Markdown resolution, curl-based scrapers, most LLM fetch pipelines) requests /blog/api/media/file/..., which 404s. Consequence: every post with a hero image ships a broken image reference in exactly the machine-readable surface this PR exists to provide. Fix: run heroImage.url through absoluteUrl(origin, ...) like the other links.

P3 · app/(frontend)/llm-markdown/[[...path]]/route.ts:75-82 — Trigger: a legacy post row with an empty slug (the column is optional in collections/Posts.ts; the auto-slug hook only fills it on new saves). Broken behavior: postMarkdownUrl builds /blog/.md into the home and blog-index listings. Consequence: dead alternate link pointing at a Markdown 404 until the post is re-saved in admin. Low probability given slugs are unique/admin-managed, hence P3.

What else got checked: proxy matcher correctly skips api, admin, _next/*; /.md, /index.md, /blog.md, /blog/<slug>.md all map cleanly through the optional catch-all including trailing-slash and unknown-path cases returning proper 404 Markdown; error responses get no-store plus X-Robots-Tag: noindex; Lexical conversion guards malformed rich text and falls back to a typed 500; blog-index limit of 50 matches the HTML page so no drift between mirrors; component refactor to site-content.ts preserves all copy verbatim (footer GitHub repo URL still points at celeroncoder.tech, pre-existing); new Date().getFullYear() in footer was already suppressHydrationWarning.

Checks: GitGuardian pass; react-doctor and Vercel pending at review time.

verify:curl -s https://<preview>/blog/<slug-with-hero-image>.md | grep '^!\[' — confirm the image URL is absolute, then fetch it and expect 200 instead of 404.

[celeroncoder]

What changed

• serve Markdown versions at /.md, /index.md, /blog.md, and /blog/.md
• convert Payload Lexical blog content to Markdown, including fenced code blocks
• advertise each Markdown page with a text/markdown alternate link
• keep homepage Markdown in sync through shared portfolio content
• return Markdown 404 and 500 responses with the correct content type

Verification

• ./node_modules/.bin/tsc --noEmit
• ./node_modules/.bin/next build
• HTTP check confirmed .md rewrites return text/markdown and Markdown error bodies

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
celeroncoder-web	Ready	Preview	Aug 25, 2026 7:27am

[github-actions]

React Doctor found no new issues. 🎉

_{Reviewed by React Doctor for commit f54d742.}

[celeroncoder]

Hermes reviewer is reviewing this PR and will post shortly.

[celeroncoder]

FOLLOW-UP · P1: 0 · P2: 1 · P3: 1

P2 · lib/markdown-pages.ts:131-134 (renderBlogPostMarkdown) — Trigger: a post has a heroImage whose url is server-relative (Payload's vercel-blob/media adapter returns paths like /api/media/file/..., and the code already treats non-object heroImages as absent, so both shapes flow through here). Broken behavior: the image is emitted as ![alt](/api/media/file/...) inside a document served from /blog/<slug>.md. Any consumer resolving that link against the document URL (standard Markdown resolution, curl-based scrapers, most LLM fetch pipelines) requests /blog/api/media/file/..., which 404s. Consequence: every post with a hero image ships a broken image reference in exactly the machine-readable surface this PR exists to provide. Fix: run heroImage.url through absoluteUrl(origin, ...) like the other links.

P3 · app/(frontend)/llm-markdown/[[...path]]/route.ts:75-82 — Trigger: a legacy post row with an empty slug (the column is optional in collections/Posts.ts; the auto-slug hook only fills it on new saves). Broken behavior: postMarkdownUrl builds /blog/.md into the home and blog-index listings. Consequence: dead alternate link pointing at a Markdown 404 until the post is re-saved in admin. Low probability given slugs are unique/admin-managed, hence P3.

What else got checked: proxy matcher correctly skips api, admin, _next/*; /.md, /index.md, /blog.md, /blog/<slug>.md all map cleanly through the optional catch-all including trailing-slash and unknown-path cases returning proper 404 Markdown; error responses get no-store plus X-Robots-Tag: noindex; Lexical conversion guards malformed rich text and falls back to a typed 500; blog-index limit of 50 matches the HTML page so no drift between mirrors; component refactor to site-content.ts preserves all copy verbatim (footer GitHub repo URL still points at celeroncoder.tech, pre-existing); new Date().getFullYear() in footer was already suppressHydrationWarning.

Checks: GitGuardian pass; react-doctor and Vercel pending at review time.

verify:curl -s https://<preview>/blog/<slug-with-hero-image>.md | grep '^!\[' — confirm the image URL is absolute, then fetch it and expect 200 instead of 404.

[celeroncoder]

What changed

• serve Markdown versions at /.md, /index.md, /blog.md, and /blog/.md
• convert Payload Lexical blog content to Markdown, including fenced code blocks
• advertise each Markdown page with a text/markdown alternate link
• keep homepage Markdown in sync through shared portfolio content
• return Markdown 404 and 500 responses with the correct content type

Verification

• ./node_modules/.bin/tsc --noEmit
• ./node_modules/.bin/next build
• HTTP check confirmed .md rewrites return text/markdown and Markdown error bodies

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
celeroncoder-web	Ready	Preview	Aug 25, 2026 7:27am

[github-actions]

React Doctor found no new issues. 🎉

_{Reviewed by React Doctor for commit f54d742.}

[celeroncoder]

Hermes reviewer is reviewing this PR and will post shortly.

[celeroncoder]

FOLLOW-UP · P1: 0 · P2: 1 · P3: 1

P2 · lib/markdown-pages.ts:131-134 (renderBlogPostMarkdown) — Trigger: a post has a heroImage whose url is server-relative (Payload's vercel-blob/media adapter returns paths like /api/media/file/..., and the code already treats non-object heroImages as absent, so both shapes flow through here). Broken behavior: the image is emitted as ![alt](/api/media/file/...) inside a document served from /blog/<slug>.md. Any consumer resolving that link against the document URL (standard Markdown resolution, curl-based scrapers, most LLM fetch pipelines) requests /blog/api/media/file/..., which 404s. Consequence: every post with a hero image ships a broken image reference in exactly the machine-readable surface this PR exists to provide. Fix: run heroImage.url through absoluteUrl(origin, ...) like the other links.

P3 · app/(frontend)/llm-markdown/[[...path]]/route.ts:75-82 — Trigger: a legacy post row with an empty slug (the column is optional in collections/Posts.ts; the auto-slug hook only fills it on new saves). Broken behavior: postMarkdownUrl builds /blog/.md into the home and blog-index listings. Consequence: dead alternate link pointing at a Markdown 404 until the post is re-saved in admin. Low probability given slugs are unique/admin-managed, hence P3.

What else got checked: proxy matcher correctly skips api, admin, _next/*; /.md, /index.md, /blog.md, /blog/<slug>.md all map cleanly through the optional catch-all including trailing-slash and unknown-path cases returning proper 404 Markdown; error responses get no-store plus X-Robots-Tag: noindex; Lexical conversion guards malformed rich text and falls back to a typed 500; blog-index limit of 50 matches the HTML page so no drift between mirrors; component refactor to site-content.ts preserves all copy verbatim (footer GitHub repo URL still points at celeroncoder.tech, pre-existing); new Date().getFullYear() in footer was already suppressHydrationWarning.

Checks: GitGuardian pass; react-doctor and Vercel pending at review time.

verify:curl -s https://<preview>/blog/<slug-with-hero-image>.md | grep '^!\[' — confirm the image URL is absolute, then fetch it and expect 200 instead of 404.

[celeroncoder]

What changed

• serve Markdown versions at /.md, /index.md, /blog.md, and /blog/.md
• convert Payload Lexical blog content to Markdown, including fenced code blocks
• advertise each Markdown page with a text/markdown alternate link
• keep homepage Markdown in sync through shared portfolio content
• return Markdown 404 and 500 responses with the correct content type

Verification

• ./node_modules/.bin/tsc --noEmit
• ./node_modules/.bin/next build
• HTTP check confirmed .md rewrites return text/markdown and Markdown error bodies

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
celeroncoder-web	Ready	Preview	Aug 25, 2026 7:27am

[github-actions]

React Doctor found no new issues. 🎉

_{Reviewed by React Doctor for commit f54d742.}

[celeroncoder]

Hermes reviewer is reviewing this PR and will post shortly.

[celeroncoder]

FOLLOW-UP · P1: 0 · P2: 1 · P3: 1

P2 · lib/markdown-pages.ts:131-134 (renderBlogPostMarkdown) — Trigger: a post has a heroImage whose url is server-relative (Payload's vercel-blob/media adapter returns paths like /api/media/file/..., and the code already treats non-object heroImages as absent, so both shapes flow through here). Broken behavior: the image is emitted as ![alt](/api/media/file/...) inside a document served from /blog/<slug>.md. Any consumer resolving that link against the document URL (standard Markdown resolution, curl-based scrapers, most LLM fetch pipelines) requests /blog/api/media/file/..., which 404s. Consequence: every post with a hero image ships a broken image reference in exactly the machine-readable surface this PR exists to provide. Fix: run heroImage.url through absoluteUrl(origin, ...) like the other links.

P3 · app/(frontend)/llm-markdown/[[...path]]/route.ts:75-82 — Trigger: a legacy post row with an empty slug (the column is optional in collections/Posts.ts; the auto-slug hook only fills it on new saves). Broken behavior: postMarkdownUrl builds /blog/.md into the home and blog-index listings. Consequence: dead alternate link pointing at a Markdown 404 until the post is re-saved in admin. Low probability given slugs are unique/admin-managed, hence P3.

What else got checked: proxy matcher correctly skips api, admin, _next/*; /.md, /index.md, /blog.md, /blog/<slug>.md all map cleanly through the optional catch-all including trailing-slash and unknown-path cases returning proper 404 Markdown; error responses get no-store plus X-Robots-Tag: noindex; Lexical conversion guards malformed rich text and falls back to a typed 500; blog-index limit of 50 matches the HTML page so no drift between mirrors; component refactor to site-content.ts preserves all copy verbatim (footer GitHub repo URL still points at celeroncoder.tech, pre-existing); new Date().getFullYear() in footer was already suppressHydrationWarning.

Checks: GitGuardian pass; react-doctor and Vercel pending at review time.

verify:curl -s https://<preview>/blog/<slug-with-hero-image>.md | grep '^!\[' — confirm the image URL is absolute, then fetch it and expect 200 instead of 404.