Digital products, blueprints, and research

FrankX combines digital product blueprints (data/products.json), source-verified research hubs (app/research/), and proxied brand documentation (app/docs/[[...slug]]/route.ts) inside frankxai/frankx.ai-vercel-website.

Product catalog structure (data/products.json)Copy Anchor Link

The product catalog in frankx.ai-vercel-website/data/products.json defines 21 product entries rendered at /products/[slug] and linked across the homepage and domain hubs. Each record includes id, slug, name, category, badge, headline, promise, transformation[], socialProof, offer, modules[], faq[], analyticsId, and optional delivery metadata.

Unprovisioned commercial offers route their primary CTA to /waitlist?intent=<slug> rather than exposing an unconfigured checkout button.

Research publication gate (scripts/check-research-publication.mjs)Copy Anchor Link

Public research claims in frankx.ai-vercel-website are gated by scripts/check-research-publication.mjs, which validates both data/research/approved-claims.json (validateApprovedClaims) and the TypeScript source registries lib/research/sources.ts and lib/research/validated-claims.ts (findUnsupportedGeneratedGrades).

1. Approved claims ledger (validateApprovedClaims)Copy Anchor Link

Every claim in data/research/approved-claims.json (schemaVersion: 1) must satisfy:

• Stable slug & dossier: Unique kebab-case id, domain slug, text (at least 20 characters), and all 9 required dossier fields (question, scope, method, inclusionCriteria, exclusionCriteria, limitations, contraryEvidence, correctionCheck, rightsDecision, each at least 20 characters).
• Publication status: Must be one of supported-synthesis, vendor-claim, preprint-finding, or independently-replicated.
• Direct HTTPS sources: Every source requires title, locator, version, type (journal, preprint, official-report, benchmark, or documentation), and a direct HTTPS URL. Search result URLs (google.com, bing.com, scholar.*, /search) and self-referential frankx.ai URLs are rejected.
• Independent replication proof: Claims marked independently-replicated require at least two distinct studies from two independent organizations plus a replicationReceipt (protocolUrl, resultUrl, team, performedAt).
• Human maker-checker receipt: reviewedBy must be a named human reviewer (never an agent/bot identifier like ai, agent, claude, codex, gpt, or llm), distinct from draftedBy, paired with a reviewReceipt pinning repository: 'frankxai/frankx.ai-vercel-website', a 40-hex-character headSha, prNumber, reviewId, and reviewerLogin.

2. Generated grade guard (findUnsupportedGeneratedGrades)Copy Anchor Link

The script parses exported literals (domainSources, validatedClaims, researchBriefs) in lib/research/sources.ts and lib/research/validated-claims.ts and fails if any entry:

• Uses a search engine URL or self-referential /research/ citation.
• Marks type: 'journal' or 'conference' without a primary identifier (doi.org/, pubmed.ncbi.nlm.nih.gov/, or arxiv.org/abs|pdf/).
• Claims confidence: 'high', replicationStatus: 'replicated', evidenceQuality: 'rct', or crossRefCount > 0 without a backing primary source URL.
• Uses the banned PhD-grade prestige label anywhere in a string field.

# Run the research publication gate and its unit test suite
node scripts/check-research-publication.mjs
node --test scripts/tests/research-publication-gate.test.mjs

How frankx.ai/docs proxies to docs-platformCopy Anchor Link

The App Router handler at app/docs/[[...slug]]/route.ts (backed by lib/docs-proxy.ts) serves the FrankX documentation tenant at https://frankx.ai/docs:

• Path mapping & content negotiation: Bare /docs maps to /frankx/intro. Requests ending in .md or carrying Accept: text/markdown (weighted higher than text/html) resolve to /llms.mdx/frankx/<slug>/content.md. /docs/llms.txt maps to /llms-index/frankx and /docs/llms-full.txt maps to /llms-full.mdx/frankx.
• Forwarded headers & timeout budget: The proxy forwards accept, rsc, next-router-prefetch, next-router-state-tree, and next-url with a 20,000ms connect/header timeout (UPSTREAM_TIMEOUT_MS) and follows up to 3 manual redirects (UPSTREAM_REDIRECT_HOPS), rejecting any redirect that leaves the frankx docs prefix.
• Response rewriting (rewriteForProxyHost): For text/html and text/x-component responses, root-relative "/frankx/..." links are rewritten to "/docs/...", while "/_next/...", "/llms.mdx/frankx/...", and "/llms-index/frankx" rewrite to absolute DOCS_ORIGIN URLs.