devshipit.com

Source for devshipit.com, the portfolio and product site of Dev Agarwalla.

Plain HTML and CSS in public/, served by a small Cloudflare Worker (src/).

Deploying

Cloudflare Workers Builds is connected to this repository. A push to main deploys the site; there is nothing to run by hand.

• Worker name: odd-fog-f7c1 (see wrangler.jsonc)
• Custom domains: devshipit.com and www.devshipit.com (www redirects to the apex)

What the Worker does

Static files come from the assets binding. The Worker runs first on every request and adds:

• Markdown for agents. A page requested with Accept: text/markdown gets its Markdown twin (index.md next to each index.html) with Content-Type: text/markdown and Vary: Accept. HTML responses also carry Vary: Accept. q-values are honoured; a client that accepts neither format gets 406.
• Agent-friendly errors. Unknown paths return 404: the HTML 404 page for browsers, a short Markdown explanation for agents, and RFC 9457 problem details (application/problem+json, with code and hint) for clients that prefer JSON. A page asked for only as JSON gets a 406 problem pointing at the API.
• JSON API. GET /api/v1/* is a public, read-only API (src/api.js): profile, projects, services, FAQ, contact and every page as Markdown. It is described by OpenAPI 3.1 at /openapi.json (src/openapi.js), listed in /.well-known/api-catalog (RFC 9727) and advertised with Link: rel="service-desc" (RFC 8631) on every page.
• MCP server. POST /mcp is a public, read-only Model Context Protocol server (Streamable HTTP, JSON responses, no sign-in) with the tools get_profile, list_projects, list_services, get_faq, list_pages, get_page and get_contact. Its data comes from public/agent/site.json and the Markdown pages. Discovery: /.well-known/mcp/server-card.json (also /.well-known/mcp, /.well-known/mcp.json, /.well-known/mcp/server-cards.json).
• Headers. Security headers (CSP and friends) and cache rules, because _headers does not apply once a Worker runs first.

Machine-readable files: /llms.txt (with when-to-use guidance), /llms-full.txt, /sitemap.xml, /robots.txt, /site.webmanifest, and JSON-LD on every page (Person, ProfessionalService, WebSite, FAQPage, BreadcrumbList, the product behind each case study).

CLI

cli/ is the devshipit command-line client for the API, with no dependencies: npx github:devag7/devshipit projects. See cli/README.md. To publish it to npm: cd cli && npm publish (needs an npm account).

Search engines

.github/workflows/indexnow.yml waits for each deploy that touches public/, then submits the sitemap URLs to IndexNow (Bing and others) with the key file in public/. Run it by hand with npm run indexnow.

Editing

The pages are generated. site/build.py writes every HTML page, its Markdown twin, llms.txt, llms-full.txt, agent/site.json, the sitemap, robots.txt, the web manifest and the icon sprite into public/. Legal and case-study copy comes verbatim from site/source/. Styles and scripts in public/assets/ (site.css, home.css, app.js, site.js, home.js) are edited by hand.

npm --prefix site ci             # icon sources
pip install markdownify==1.2.3   # Markdown twins
python3 site/build.py            # regenerate public/

Change copy in site/build.py (or site/source/), rebuild, and commit both. CI rebuilds and fails if public/ differs from what the generator produces.

Tests

npm test                       # negotiation, Worker, API, OpenAPI, MCP, CLI, SEO and generated-file checks
npx wrangler dev               # run the Worker locally on http://localhost:8787
npm run verify                 # check the live site (or: node scripts/verify-live.mjs http://localhost:8787)