Truplexy — marketing site

Every channel. One platform. The homepage positions Truplexy as an AI assistant on every channel a business uses, plus one platform where the team manages every conversation, as in the launch film.

Astro 7 + React islands + Tailwind v4, with GSAP/Lenis for motion. Structure mirrors zenova-v2.

Commands

Command	What it does
npm run dev	Dev server on http://localhost:4321
npm run build	Static build to dist/
npm run check	Type-check (astro check)
npm run verify:seo	Sitemap, canonical, noindex and robots.txt audit of dist/ (run after build)
npm run verify	Honesty audit of the built home, /channels/* and /docs/* pages (run after build)
npm run brand	Regenerate trimmed logos and favicons from /logo
npm run og	Render public/og-card.png from scripts/og-card.html

Deploying

vercel.json deploys this site and the dashboard (dashboard/) as one Vercel project with two services: dashboard on app. hosts (app.truplexy.com), site on every other host. vercel dev runs both locally (the dashboard at http://app.localhost:3000).

Search (SEO)

• /sitemap.xml is built by @astrojs/sitemap from every page in the build except the 404, so a new or deleted page appears or disappears with the next deploy. A small integration in astro.config.mjs (sitemapAtRoot) publishes the result at /sitemap.xml: the single <urlset> while the site fits in one file, or the sitemap index once it passes 45,000 URLs (each sitemap-N.xml chunk stays under the protocol's 50,000 URL / 50 MB limits). /sitemap-index.xml, the old location, 308-redirects to it.
• Each lastmod is the date of the last commit touching that page's sources (PAGE_SOURCES / PREFIX_SOURCES in astro.config.mjs). It is left out when the build runs from a shallow clone, because every file would then show the same date. Add new pages to PAGE_SOURCES if they read from more than their own file. changefreq and priority are omitted on purpose: Google ignores them.
• /robots.txt is generated by src/pages/robots.txt.ts from the two lists in src/data/crawlers.ts: SEARCH_CRAWLERS (Googlebot, Bingbot, OAI-SearchBot, ChatGPT-User, PerplexityBot, Perplexity-User, Claude-SearchBot, Claude-User, Applebot, DuckDuckBot) may fetch everything except /404, and TRAINING_CRAWLERS (GPTBot, ClaudeBot, Google-Extended, CCBot …) are blocked. Search access and model training are separate decisions: to allow a training crawler, move its name between the lists. It names https://truplexy.com/sitemap.xml. To keep a page out of search, pass noindex to Base.astro (that also drops its canonical tag; add its route to the sitemap filter) rather than blocking it in robots.txt, which cannot stop a URL from being indexed.
• Identity: the home page opens with a visible "What is Truplexy?" section (src/components/home/WhatIs.astro) that answers what it is, who it is for, which channels it supports and how it differs from a chatbot, and says Truplexy is an independent product of Zenova Agency, not Perplexity AI. Channel names come from the registry. The JSON-LD in Base.astro (Organization, WebSite, SoftwareApplication, WebPage and BreadcrumbList, linked by @id) says the same. It deliberately has no ratings, reviews, offers or sameAs: add sameAs to the Organization once official profiles exist.
• npm run verify:seo (after npm run build) checks dist/: sitemap well-formedness and limits, unique absolute https URLs, every URL built, indexable and self-canonical, no indexable page missing from the sitemap, no broken internal links, that every search crawler may fetch every URL while training crawlers stay blocked, one per page, valid linked JSON-LD with no ratings/reviews/offers, and the home page's "What is Truplexy?" and independence statements.

Where things live

• src/data/channels.ts — channel registry. Every badge, diagram node and footer list reads status from here. Flip a channel to 'live' and the whole page follows.
• src/data/demo.ts — every sample conversation, from one fictional cast. The build fails if a demo uses a channel that isn't live.
• src/data/site.ts — name, domain, email and CTA links.
• src/components/home/* — one file per homepage section, in page order in src/pages/index.astro.
• src/islands/* — the interactive parts: the hero card stack (sample conversation playback) and the unified inbox.

Website SDK

sdk/ holds the website chat SDK: the <truplexy-chat> widget (@truplexy/web), React and Vue wrappers, the server route (@truplexy/server) and a WordPress plugin. It is its own npm workspace. Run cd sdk && npm install && npm test, and see sdk/README.md. The typed server client and contact-based identity flow are documented in sdk/CHAT_API.md; API_REFERENCE.md is the API contract.

Channel pages and docs

Both use the docs layout (src/layouts/Docs.astro + src/styles/docs.css: hero, sticky "On this page" list, .docs article), and each has one job:

• /channels/<id> (src/pages/channels/[id].astro) is the overview: a sample conversation drawn in that platform's own look (src/components/channels/ChannelDemo.astro picks a skin from src/components/channels/demo/: the website widget, a phone app, a team workspace, an inbox or a help desk; the widget skin copies sdk/packages/web/src/styles.ts), what Truplexy does there, how a conversation runs, and key facts. No code or setup detail. Content: src/data/channel-pages.ts. Developer tools (REST API, webhooks, custom integrations) are not channels customers use, so they have no overview page; channelHref() sends their links to the docs.
• Platforms with an app marketplace (WhatsApp, Messenger, Instagram, Discord, Slack, Teams, Shopify, HubSpot, Zendesk: connect: true in src/data/channels.ts) have no docs page and no server code. Truplexy's own app connects them from the dashboard in one click (Integrations → Add a channel → Connect, which calls POST /channels/connect and comes back from the platform's sign-in), and their channel page points there instead of a guide (setupHref()). npm run verify fails if one of them gets a /docs/<id> page or a link into the docs.
• /docs is the setup home: quickstart, core concepts, and every guide grouped (DOC_GROUPS). /docs/<id> (src/pages/docs/[id].astro) is a setup guide: before you start, steps, server code per language, testing and troubleshooting. Content: src/data/setup-guides.ts. Custom integrations are a section of /docs/api, WooCommerce a section of /docs/wordpress (same plugin), and Website chat has its own, longer guide at /docs/web-sdk.

Adding a channel: add it to the registry, then its overview to PAGES in channel-pages.ts. If the platform has an app marketplace, mark it connect: true here and in the dashboard catalog (connect in catalog.tsx, with its channel type) and stop there; otherwise, if it needs its own guide, add it to GUIDES in setup-guides.ts (npm run check fails on a missing overview, and on a missing demo skin in ChannelDemo.astro). Plan availability is read from COMPARISON in src/data/pricing.ts by planLabel. Sample conversations reuse the fictional store and cast of src/data/demo.ts. The server code is not written here: src/data/snippets.ts arranges the dashboard's snippet packs (dashboard/src/features/integrations/snippets/) into tabs, so the dashboard guides and these pages share one source. The provider-side steps in setup-guides.ts extend dashboard/src/features/integrations/catalog.tsx; update both when a platform's setup changes. Those snippet files must stay free of dashboard-only imports (@/…), since the site imports them.

Its install guide is src/pages/docs/web-sdk.astro (/docs/web-sdk). That page reads the framework recipes from dashboard/src/features/integrations/snippets/web-sdk.ts, the same file the dashboard's Website chat guide uses. Its live demo widget is built from sdk/packages/web/src.

Before launch

Search for PLACEHOLDER:

• src/data/site.ts — contact email, sign-up URL and demo-booking URL (currently mailto: links). The domain (url) set there feeds astro.config.mjs (site), canonical URLs, robots.txt and the sitemap.
• src/data/legal.ts — the operator, address, governing law and minimum age behind /terms, /privacy and /refunds. Have a lawyer review all three pages. Zenova Agency is unregistered for now: once it is, update entity and address there and re-check the name on the Paddle account. The contact email comes from SITE.email (see above).