# Yantrika marketing site The public site: a landing page whose primary call to action starts a live fleet with no signup, a deep page on the free VDA 5050 conformance tester, plus About, Features, Security, Docs (link-out), Get Involved, Changelog and a site search. Plain hand-written HTML/CSS/JS — no build step, no bundler, no framework, no `package.json` — consistent with the rest of this repo (`console/`, `academy/`, `docs/`). ``` marketing/ index.html landing page + the "Try it" button vda-5050-conformance-test.html the conformance tester (SEO landing page) about.html features.html security.html summary + links to docs/SECURITY.md, docs/COMPLIANCE.md docs.html link-out page to the real docs/ get-involved.html GitHub Issues + CONTRIBUTING.md (no fake contact form) changelog.html excerpt of the real CHANGELOG.md search.html client-side search over the pages above assets/styles.css shared stylesheet for all pages assets/main.js one tiny script: mobile nav toggle assets/demo.js progressive enhancement for the "Try it" button assets/og-yantrafleet.png 1200x630 OpenGraph/Twitter card (site-wide) assets/og-vda-5050-conformance.png 1200x630 card for the conformance page sitemap.xml robots.txt ``` Unlike `console/index.html` and `academy/index.html` (deliberately single-file apps), this is a multi-page site, so the CSS lives once in `assets/styles.css` rather than being duplicated into every page — still plain, hand-written, unbundled CSS, just shared. ## Two filenames that must not change * **`vda-5050-conformance-test.html`** — every HTML report `yantra-conform` generates hard-codes a footer link to `/vda-5050-conformance-test.html` (`tools/conformance/yantraconform/report_html.py`, `TOOL_URL`). Renaming this file breaks that link in every report already sitting in somebody's inbox. Both `deploy/*/validate.sh` pin the name. * **`assets/styles.css`** — every page links it; there is no fallback. ## The "Try it with a live fleet" button `index.html`'s primary call to action is a plain `
`. That is the whole feature, and it works with JavaScript off: the door (`ops/yantraops/sandbox_http.py`) sees `text/html` in the `Accept` header, answers `303` with the new sandbox's console URL in `Location:`, and the browser follows it into a running fleet. `assets/demo.js` only *enhances* it — it asks `GET /api/demo/limits` whether the demo is available and hides the block if not, and it turns a `429`/`503` into a sentence rather than a browser error page. Blocked or broken, the form underneath is untouched. For this to work at all, the deployment must: 1. serve `/api/demo/` (the shared nginx template already has the location and its rate-limit zone — see `deploy/aws/nginx/yantrafleet.conf.template`); 2. have applied `supabase/0009_demo_sandbox.sql`, deliberately, after reading its THREAT MODEL block; 3. have `yantra-sandbox` and `yantra-sandbox-reap.timer` enabled, and a service_role key in `/etc/yantrafleet-sandbox.env` for the reaper. Steps 2 and 3 are written out in both `deploy/aws/README.md` and `deploy/azure/README.md`. Until they are done the button is present but the door answers `502`; `demo.js` hides it on a deployment where `/api/demo/limits` says the demo is unavailable. ## Content sourcing Every factual claim on these pages traces back to this repository's own `README.md`, `docs/SECURITY.md`, `docs/COMPLIANCE.md`, `TEST-REPORT.md`, `tools/conformance/README.md`, `connector/README.md`, `academy/README.md`, or `ops/yantraops/sandbox_http.py`. There is no pricing page, no customer logos or testimonials, no fabricated usage numbers, and no invented functionality — none of that exists in the project. The 52-row check table on `vda-5050-conformance-test.html` is a rendering of the tester's own catalogue, i.e. the output of `yantra-conform checks --json` (`tools/conformance/yantraconform/checks.py`). Check ids, titles, "expected" lines, severities and clause references are copied from there rather than written by hand. **If the rule set changes, regenerate that table** rather than editing rows individually — `yantra-conform checks --json` prints the whole set with no broker and no network. One known wrinkle, inherited from the repo's own docs rather than introduced here: the root `README.md`'s badge says "525 tests" while the more granular `TEST-REPORT.md` sums to 633 across the same 10 suites. The marketing copy sidesteps this with the round, defensible phrasing "10 automated test suites, 500+ tests, all passing" (true under either number) everywhere except the Changelog page, which reproduces each release's own historical wording verbatim. The conformance tester's own 188 tests are quoted separately because they are a separate suite (`tools/conformance/tests`) that `TEST-REPORT.md` does not yet count. ## Two placeholders — resolved Both used to be placeholders (`https://yantrafleet.example.com` and `https://github.com/YOUR_GITHUB_USERNAME/yantrafleet`), consistently across every file listed below. Both have since been replaced with the real values below; this section is kept as a record of where each one lives and as the regression check in "Check your work". ### 1. The domain — `https://yantrika.ai` | File | Tags / lines carrying it | |---|---| | `index.html` | ``, `og:url`, `og:image`, `twitter:image`, and the JSON-LD `@graph`: `WebSite` `@id`+`url`, its `SearchAction` `urlTemplate`, `SoftwareApplication` `@id`+`url`, `FAQPage` `@id` | | `vda-5050-conformance-test.html` | ``, `og:url`, `og:image`, `twitter:image`, and the JSON-LD `@graph`: `SoftwareApplication` `@id`+`url`+`isPartOf`, `FAQPage` `@id`, both `BreadcrumbList` `item` URLs | | `about.html`, `features.html`, `security.html`, `docs.html`, `get-involved.html`, `changelog.html` | ``, `og:url`, `og:image`, `twitter:image`, and both `BreadcrumbList` `item` URLs | | `search.html` | `` | | `sitemap.xml` | every `` (8 of them) | | `robots.txt` | the `Sitemap:` line (and the comment header) | As of this writing that is 11 occurrences each in `index.html` and `vda-5050-conformance-test.html`, 6–7 in each other page, 1 in `search.html`, 9 in `sitemap.xml` and 2 in `robots.txt`. Do not count by hand — use the grep under "Check your work". **One more place, outside this directory.** The conformance tester bakes a site URL into every report it generates: `tools/conformance/yantraconform/report_html.py` sets `HOME_URL` and `TOOL_URL` to `https://yantrika.ai/...`. That is a real domain, not the `example.com` placeholder, so the grep below will not catch it. If `yantrika.ai` is your domain there is nothing to do; if it is not, change it there as well or every report you hand a vendor links to somebody else's site. ### 2. The GitHub repository URL — `https://github.com/yantrika-ai/yantrika` Same placeholder convention also used in `deploy/aws/user-data.sh` and `deploy/azure/custom-data.sh` (their `REPO_URL=` default), both of which have been updated too. It appears in: | File | Where | |---|---| | every `*.html` | the nav "View on GitHub" button, the footer "GitHub repository" and "License (MIT)" links | | `index.html` | additionally: the `git clone` line in the quickstart block, the JSON-LD `codeRepository`, the two `deploy/*/README.md` links | | `vda-5050-conformance-test.html` | additionally: the `git clone` line, the JSON-LD `codeRepository`, and three `/tree/main/tools/conformance` links | | `docs.html`, `get-involved.html`, `about.html`, `changelog.html` | additionally: the in-page links to `CONTRIBUTING.md`, `LICENSE`, `CHANGELOG.md`, `docker/`, `deploy/*/README.md` and issue templates, which point at GitHub precisely because nginx does not serve the repo root | | `assets/demo.js` | the fallback link shown when the live demo is unavailable | If either value ever needs to change again, from `marketing/`: ```bash grep -rl 'https://yantrika.ai' . | xargs sed -i 's#https://yantrika\.ai#https://YOUR-NEW-DOMAIN#g' grep -rl 'github.com/yantrika-ai/yantrika' . | xargs sed -i 's#github\.com/yantrika-ai/yantrika#github.com/YOUR-ORG/YOUR-REPO#g' # and, if the domain changes, the links inside every generated conformance report: sed -i 's#https://yantrika\.ai#https://YOUR-NEW-DOMAIN#g' \ ../tools/conformance/yantraconform/report_html.py ``` (`sed -i ''` on macOS.) ### Check your work ```bash # from the repo root — both must print nothing before you publish grep -rn 'yantrafleet.example.com' marketing/ || echo "domain: clean" grep -rn 'YOUR_GITHUB_USERNAME' marketing/ || echo "repo: clean" ``` A site published with the placeholder domain still in its canonical tags is *worse* than one with no tags at all: every page tells the crawler its real address is somewhere else. ## Preview locally These are plain files. Serve from the **repo root**, not from `marketing/`, so the links into `/docs/`, `/console/` and `/academy/` resolve the way they do in production: ```bash # from the repo root python3 -m http.server 8095 # then open http://localhost:8095/marketing/ ``` Note the one difference from production: under nginx the marketing site *is* the host root, so `/docs/SECURITY.md` resolves directly; under the command above the site lives at `/marketing/` while `/docs/` is still at the root, so the root-relative links work but the page-to-page links need the `/marketing/` prefix the browser already has. Opening `index.html` straight off disk works too, for everything except the root-relative links and the demo button. ## Deploy The marketing site owns the bare host root. That is already wired up: the shared nginx template (`deploy/aws/nginx/yantrafleet.conf.template`, used by **both** the AWS and Azure kits) does ```nginx root /opt/yantrafleet/marketing; index index.html; location / { try_files $uri $uri/ =404; } ``` with `/console/`, `/academy/`, `/docs/`, `/ask`, `/health` and `/api/demo/` as sibling locations. Nothing in this directory needs its own deploy step — the boot scripts clone the repo to `/opt/yantrafleet` and nginx serves it. **Link rule, learned the hard way:** nginx serves *only* the locations in that template. It does **not** serve the repo root or `deploy/`, so a link to `../LICENSE`, `../CONTRIBUTING.md`, `../CHANGELOG.md` or `../deploy/aws/README.md` is dead in production even though it works when you preview from the repo root. Those four are linked to GitHub instead, deliberately. Links into `/docs/` are fine (that location exists) and are written root-relative — `/docs/SECURITY.md`, not `../docs/SECURITY.md` — so they cannot be misread. Before adding any link out of this directory, check it against that template. ### Standalone static host Because it's just static files, this directory can also be deployed on its own — GitHub Pages, Netlify, Cloudflare Pages, S3 + CloudFront, a plain vhost. Two things stop working, and you should decide what to do about each: - **`/docs/`, `/console/`, `/academy/`** — repoint those links at wherever those actually live, or drop them. - **the "Try it with a live fleet" button** — there is no `/api/demo/` endpoint on a static host, so the form would post into a 404. Either point the form's `action` at the full URL of a host that *is* running `yantra-sandbox`, or remove the `
` block from `index.html` (and the `demo.js` `