#!/usr/bin/env node
/**
 * Auditoria repetível de SEO / AEO / GEO, sem dependências (Node >= 20).
 *
 *   node scripts/seo-audit.mjs --base http://localhost:3000 --out audit-antes.json
 *   node scripts/seo-compare.mjs audit-antes.json audit-depois.json
 *
 * Lê o sitemap, baixa cada URL como um robô (HTML sem JavaScript) e confere o que buscadores e sistemas de
 * resposta/IA realmente enxergam. Não inventa nota: cada verificação é passa/falha com a evidência.
 * O que depende de conta (Search Console, Bing, CrUX, citações em plataformas de IA) está em docs/seo-geo-operations.md.
 */
import { writeFileSync } from "node:fs";

const args = Object.fromEntries(
  process.argv
    .slice(2)
    .map((a, i, all) => (a.startsWith("--") ? [a.slice(2), all[i + 1]] : null))
    .filter(Boolean),
);
const BASE = (args.base ?? "https://evandro.dev.br").replace(/\/$/, "");
const SITE = "https://evandro.dev.br"; // origem canônica declarada no código
const UA = "Mozilla/5.0 (compatible; seo-audit/1.0)";
const AI_BOTS = [
  [
    "Googlebot",
    "Mozilla/5.0 (compatible; Googlebot/2.1; +http://www.google.com/bot.html)",
  ],
  [
    "bingbot",
    "Mozilla/5.0 (compatible; bingbot/2.0; +http://www.bing.com/bingbot.htm)",
  ],
  [
    "OAI-SearchBot",
    "Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko); compatible; OAI-SearchBot/1.0; +https://openai.com/searchbot",
  ],
  [
    "GPTBot",
    "Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko; compatible; GPTBot/1.1; +https://openai.com/gptbot)",
  ],
  [
    "ChatGPT-User",
    "Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko; compatible; ChatGPT-User/1.0; +https://openai.com/bot)",
  ],
  [
    "Claude-SearchBot",
    "Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko; compatible; Claude-SearchBot/1.0; +searchbot@anthropic.com)",
  ],
  [
    "ClaudeBot",
    "Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko; compatible; ClaudeBot/1.0; +claudebot@anthropic.com)",
  ],
  [
    "PerplexityBot",
    "Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko; compatible; PerplexityBot/1.0; +https://perplexity.ai/perplexitybot)",
  ],
];

const local = (url) => url.replace(SITE, BASE); // permite auditar localhost com URLs canônicas de produção
const get = async (url, ua = UA, init = {}) => {
  const t0 = performance.now();
  const res = await fetch(url, {
    headers: { "user-agent": ua, "accept-encoding": "gzip, br" },
    redirect: "manual",
    ...init,
  });
  const body = init.method === "HEAD" ? "" : await res.text();
  return { res, body, ms: Math.round(performance.now() - t0) };
};

const attrs = (tag) => {
  const out = {};
  for (const m of tag.matchAll(
    /([a-zA-Z_:][\w:.-]*)\s*=\s*(?:"([^"]*)"|'([^']*)')/g,
  ))
    out[m[1].toLowerCase()] = m[2] ?? m[3] ?? "";
  return out;
};
const decode = (s) =>
  s
    .replace(/&/g, "&")
    .replace(/"/g, '"')
    .replace(/'|'/g, "'")
    .replace(/&lt;/g, "<")
    .replace(/&gt;/g, ">")
    .trim();
const strip = (html) =>
  decode(
    html
      .replace(/<(script|style|noscript|template|svg)[\s\S]*?<\/\1>/gi, " ")
      .replace(/<!--[\s\S]*?-->/g, " ")
      .replace(/<[^>]+>/g, " ")
      .replace(/\s+/g, " "),
  );
const words = (s) => (s.match(/[\p{L}\p{N}][\p{L}\p{N}'’-]*/gu) ?? []).length;

function parsePage(html) {
  const metas = [...html.matchAll(/<meta\b[^>]*>/gi)].map((m) => attrs(m[0]));
  const meta = (key, val = "name") =>
    metas.find((m) => m[val]?.toLowerCase() === key)?.content;
  const links = [...html.matchAll(/<link\b[^>]*>/gi)].map((m) => attrs(m[0]));
  const headings = [
    ...html.matchAll(/<h([1-6])\b[^>]*>([\s\S]*?)<\/h\1>/gi),
  ].map((m) => ({ level: +m[1], text: strip(m[2]) }));
  const imgs = [...html.matchAll(/<img\b[^>]*>/gi)].map((m) => attrs(m[0]));
  const anchors = [...html.matchAll(/<a\b[^>]*>/gi)].map((m) => attrs(m[0]));
  const ld = [
    ...html.matchAll(
      /<script[^>]*type=["']application\/ld\+json["'][^>]*>([\s\S]*?)<\/script>/gi,
    ),
  ].map((m) => m[1]);
  const main = html.match(/<main\b[\s\S]*?<\/main>/i)?.[0] ?? html;
  const firstH1 = main.search(/<h1\b/i);
  const afterH1 = firstH1 >= 0 ? main.slice(firstH1) : main;
  const firstP = [...afterH1.matchAll(/<p\b[^>]*>([\s\S]*?)<\/p>/gi)]
    .map((m) => strip(m[1]))
    .find((t) => words(t) >= 8);
  return {
    title: decode(html.match(/<title[^>]*>([\s\S]*?)<\/title>/i)?.[1] ?? ""),
    description: meta("description"),
    robots: meta("robots"),
    viewport: meta("viewport"),
    lang: html.match(/<html[^>]*\blang=["']([^"']+)/i)?.[1],
    canonical: links.find((l) => l.rel === "canonical")?.href,
    alternates: links
      .filter((l) => l.rel === "alternate" && l.hreflang)
      .map((l) => l.hreflang),
    og: Object.fromEntries(
      metas
        .filter((m) => m.property?.startsWith("og:"))
        .map((m) => [m.property, m.content]),
    ),
    tw: Object.fromEntries(
      metas
        .filter((m) => m.name?.startsWith("twitter:"))
        .map((m) => [m.name, m.content]),
    ),
    headings,
    imgs,
    anchors,
    ld,
    wordCount: words(strip(main)),
    firstAnswerWords: firstP ? words(firstP) : 0,
  };
}

function jsonLdFacts(blocks) {
  const nodes = [];
  const errors = [];
  for (const raw of blocks) {
    try {
      const data = JSON.parse(raw);
      const list = data["@graph"] ?? (Array.isArray(data) ? data : [data]);
      nodes.push(...list);
    } catch (e) {
      errors.push(String(e.message).slice(0, 80));
    }
  }
  const ids = new Set(nodes.map((n) => n["@id"]).filter(Boolean));
  const types = nodes.flatMap((n) => [].concat(n["@type"] ?? []));
  const dangling = [];
  const walk = (o) => {
    if (Array.isArray(o)) return o.forEach(walk);
    if (o && typeof o === "object") {
      const keys = Object.keys(o);
      if (
        keys.length === 1 &&
        keys[0] === "@id" &&
        !ids.has(o["@id"]) &&
        !String(o["@id"]).startsWith(SITE)
      )
        dangling.push(o["@id"]);
      Object.values(o).forEach(walk);
    }
  };
  walk(nodes);
  return { nodes, ids, types, errors, dangling };
}

const checks = []; // {scope, id, group, ok, detail}
const add = (scope, group, id, ok, detail = "") =>
  checks.push({ scope, group, id, ok: Boolean(ok), detail });

/** Chaves de Open Graph que toda página precisa ter. */
const OG_KEYS = [
  "og:title",
  "og:description",
  "og:url",
  "og:image",
  "og:type",
  "og:site_name",
  "og:locale",
];

/** Compressão do HTML e tempo até o primeiro byte. */
function checkTechnical(scope, res, ms) {
  const encoding = res.headers.get("content-encoding");
  add(
    scope,
    "técnico",
    "html-comprimido",
    /br|gzip/.test(encoding ?? "") || BASE.startsWith("http://localhost"),
    encoding ?? "sem compressão",
  );
  add(
    scope,
    "técnico",
    "ttfb-menor-800ms",
    ms < 800,
    `${ms} ms (HTML completo)`,
  );
}

/** Idioma, viewport e título (tamanho e unicidade no site). */
function checkHead(scope, p, siteTitles) {
  add(
    scope,
    "on-page",
    "lang-pt-BR",
    p.lang?.toLowerCase() === "pt-br",
    p.lang ?? "ausente",
  );
  add(scope, "on-page", "viewport", Boolean(p.viewport));
  add(
    scope,
    "on-page",
    "title-30-65",
    p.title.length >= 30 && p.title.length <= 65,
    `${p.title.length} car.: ${p.title}`,
  );
  add(scope, "on-page", "title-unico", !siteTitles.has(p.title), p.title);
  siteTitles.add(p.title);
}

/** Descrição (tamanho e unicidade), canônica própria e indexabilidade. */
function checkSnippet(scope, url, res, p, siteDescs) {
  add(
    scope,
    "on-page",
    "description-70-165",
    !!p.description &&
      p.description.length >= 70 &&
      p.description.length <= 165,
    `${p.description?.length ?? 0} car.`,
  );
  add(
    scope,
    "on-page",
    "description-unica",
    !!p.description && !siteDescs.has(p.description),
  );
  if (p.description) siteDescs.add(p.description);
  add(
    scope,
    "on-page",
    "canonical-propria",
    p.canonical === url || p.canonical === `${url}/`,
    p.canonical ?? "ausente",
  );
  const noindex = (value) => /noindex/i.test(value ?? "");
  add(
    scope,
    "on-page",
    "indexavel",
    !noindex(p.robots) && !noindex(res.headers.get("x-robots-tag")),
  );
}

/** A hierarquia de títulos pula algum nível (de h1 direto para h3, por exemplo)? */
function skipsHeadingLevel(headings) {
  let previous = 0;
  for (const h of headings) {
    if (h.level - previous > 1) return true;
    previous = h.level;
  }
  return false;
}

/** Links internos (mesmo site, sem protocolo relativo) de uma lista de âncoras. */
const isInternalLink = (a) =>
  a.href &&
  (a.href.startsWith("/") || a.href.startsWith(SITE)) &&
  !a.href.startsWith("//");

/** Links internos suficientes e links externos que abrem em outra aba com noopener. */
function checkLinks(scope, p) {
  const internal = p.anchors.filter(isInternalLink);
  const destinations = new Set(internal.map((a) => a.href.split("#")[0])).size;
  add(
    scope,
    "on-page",
    "links-internos-5+",
    destinations >= 5,
    `${destinations} destinos`,
  );
  const external = p.anchors.filter((a) => a.target === "_blank");
  // noreferrer também implica noopener
  const unsafe = external.filter(
    (a) => !/noopener|noreferrer/.test(a.rel ?? ""),
  );
  add(
    scope,
    "on-page",
    "externos-com-noopener",
    unsafe.length === 0,
    `${external.length} links _blank; sem rel: ${unsafe
      .map((a) => a.href)
      .slice(0, 2)
      .join(" ")}`,
  );
}

/** Títulos (um h1, hierarquia), imagens com alt e links. */
function checkStructure(scope, p) {
  const h1 = p.headings.filter((h) => h.level === 1);
  add(scope, "on-page", "um-h1", h1.length === 1, `${h1.length} h1`);
  add(
    scope,
    "on-page",
    "hierarquia-de-titulos",
    !skipsHeadingLevel(p.headings),
  );
  add(
    scope,
    "on-page",
    "imagens-com-alt",
    p.imgs.every((i) => i.alt !== undefined),
    `${p.imgs.filter((i) => i.alt === undefined).length}/${p.imgs.length} sem alt`,
  );
  checkLinks(scope, p);
}

/** A imagem de compartilhamento responde 200 e é uma imagem. */
async function checkOgImage(scope, src) {
  const img = await get(local(src), UA, { method: "GET" }).catch(() => null);
  add(
    scope,
    "social",
    "og-image-acessivel",
    img?.res.status === 200 &&
      /^image\//.test(img.res.headers.get("content-type") ?? ""),
    `${img?.res.status} ${img?.res.headers.get("content-type")}`,
  );
}

/** Open Graph completo, cartão do Twitter e a imagem de compartilhamento. */
async function checkSocial(scope, p) {
  const missingOg = OG_KEYS.filter((k) => !p.og[k]);
  add(
    scope,
    "social",
    "og-completo",
    missingOg.length === 0,
    missingOg.length ? `faltam ${missingOg.join(",")}` : "",
  );
  add(
    scope,
    "social",
    "twitter-card",
    p.tw["twitter:card"] === "summary_large_image" &&
      !!p.tw["twitter:title"] &&
      !!p.tw["twitter:image"],
  );
  if (p.og["og:image"]) await checkOgImage(scope, p.og["og:image"]);
}

/** JSON-LD válido, referências que resolvem, tipos proibidos, breadcrumb e Person com sameAs. Devolve os fatos. */
function checkStructuredData(scope, p) {
  const ld = jsonLdFacts(p.ld);
  add(
    scope,
    "dados-estruturados",
    "json-ld-valido",
    p.ld.length > 0 && ld.errors.length === 0,
    ld.errors.join("; ") || `${p.ld.length} blocos`,
  );
  add(
    scope,
    "dados-estruturados",
    "referencias-resolvem",
    ld.dangling.length === 0,
    ld.dangling.join(", "),
  );
  add(
    scope,
    "dados-estruturados",
    "sem-faqpage-nem-localbusiness",
    !ld.types.some((t) => /FAQPage|LocalBusiness|ProfessionalService/.test(t)),
    ld.types.join(","),
  );
  if (scope !== "/")
    add(
      scope,
      "dados-estruturados",
      "breadcrumb",
      ld.types.includes("BreadcrumbList"),
    );
  add(
    scope,
    "dados-estruturados",
    "person-com-sameas",
    !ld.types.includes("Person") ||
      ld.nodes.some(
        (n) =>
          n["@type"] === "Person" &&
          Array.isArray(n.sameAs) &&
          // ao menos um perfil profissional no ar (o GitHub do autor foi suspenso; sobrou o LinkedIn)
          n.sameAs.length >= 1,
      ),
  );
  return ld;
}

/** Conteúdo no HTML sem JavaScript, resposta direta cedo e autor consistente. */
function checkAeo(scope, p, ld) {
  add(
    scope,
    "aeo-geo",
    "conteudo-no-html-sem-js",
    p.wordCount >= 250,
    `${p.wordCount} palavras no <main> do HTML`,
  );
  add(
    scope,
    "aeo-geo",
    "resposta-direta-cedo",
    p.firstAnswerWords >= 15 && p.firstAnswerWords <= 70,
    `${p.firstAnswerWords} palavras no 1º parágrafo após o H1`,
  );
  const names = new Set(
    ld.nodes.filter((n) => n["@type"] === "Person").map((n) => n.name),
  );
  add(
    scope,
    "aeo-geo",
    "entidade-autor-consistente",
    names.size <= 1 && (scope === "/" ? names.size === 1 : true),
    [...names].join(","),
  );
}

async function auditUrl(url, siteTitles, siteDescs) {
  const { res, body, ms } = await get(local(url));
  const scope = url.replace(SITE, "") || "/";
  add(scope, "técnico", "status-200", res.status === 200, `HTTP ${res.status}`);
  if (res.status !== 200) return null;
  checkTechnical(scope, res, ms);
  const p = parsePage(body);
  checkHead(scope, p, siteTitles);
  checkSnippet(scope, url, res, p, siteDescs);
  checkStructure(scope, p);
  await checkSocial(scope, p);
  const ld = checkStructuredData(scope, p);
  checkAeo(scope, p, ld);
  return { scope, p, ld };
}

/** Sitemap e robots.txt do site; devolve as URLs do sitemap. */
async function checkSiteFiles() {
  const sm = await get(`${BASE}/sitemap.xml`);
  const urls = [...sm.body.matchAll(/<loc>([^<]+)<\/loc>/g)].map((m) => m[1]);
  add(
    "site",
    "técnico",
    "sitemap-200",
    sm.res.status === 200 && urls.length > 0,
    `${urls.length} URLs`,
  );

  const robots = await get(`${BASE}/robots.txt`);
  add(
    "site",
    "técnico",
    "robots-com-sitemap",
    /sitemap:\s*https?:\/\//i.test(robots.body),
  );
  add(
    "site",
    "técnico",
    "robots-nao-bloqueia-tudo",
    !/disallow:\s*\/\s*$/im.test(robots.body),
  );
  add("site", "técnico", "http-para-https-ou-local", true);
  return urls;
}

/** Acesso dos robôs de busca e de IA, e-mail visível para robôs e a política do llms.txt. */
async function checkBots() {
  for (const [name, ua] of AI_BOTS) {
    const r = await get(`${BASE}/`, ua);
    add(
      "site",
      "aeo-geo",
      `acesso-${name}`,
      r.res.status === 200,
      `HTTP ${r.res.status}`,
    );
  }
  // A Cloudflare (Email Address Obfuscation) troca e-mails por /cdn-cgi/l/email-protection para quem é robô: buscadores e IA não veem o contato
  const botHome = await get(`${BASE}/`, AI_BOTS[0][1]);
  add(
    "site",
    "aeo-geo",
    "email-visivel-para-robos",
    !/email-protection|\[email(&#160;|\s)protected\]/.test(botHome.body),
    /email-protection/.test(botHome.body)
      ? "HTML servido a robôs tem e-mail ofuscado"
      : "",
  );
  const llms = await get(`${BASE}/llms.txt`);
  // informativo: a política do projeto exclui llms.txt (nenhum sistema de IA o usa, segundo o Google); ausência é o esperado
  add(
    "site",
    "aeo-geo",
    "llms-txt-ausente-por-politica",
    llms.res.status === 404,
    `HTTP ${llms.res.status}`,
  );
}

/** Audita cada URL do sitemap, uma depois da outra; páginas que não respondem 200 ficam como `null`. */
async function auditPages(urls) {
  const titles = new Set();
  const descs = new Set();
  const pages = [];
  for (const url of urls) pages.push(await auditUrl(url, titles, descs));
  return pages;
}

/** Destinos internos únicos (caminhos) de todas as páginas auditadas. */
function internalTargets(pages) {
  const targets = new Set();
  for (const pg of pages.filter(Boolean))
    for (const a of pg.p.anchors) {
      const h = a.href?.split("#")[0];
      if (h?.startsWith("/") && !h.startsWith("//")) targets.add(h);
    }
  return targets;
}

/** Destinos que respondem erro (ou não respondem), com o status. */
async function findBroken(targets) {
  const broken = [];
  for (const t of targets) {
    const r = await get(`${BASE}${t}`, UA, { method: "HEAD" }).catch(
      () => null,
    );
    if (!r || r.res.status >= 400)
      broken.push(`${t} (${r?.res.status ?? "erro"})`);
  }
  return broken;
}

/** Páginas do sitemap (fora a home) que nenhuma página audita linka. */
function findOrphans(urls, targets) {
  const inSitemap = new Set(urls.map((u) => u.replace(/\/$/, "")));
  const linked = [...targets].map((t) => t.replace(/\/$/, ""));
  return [...inSitemap].filter((u) => {
    const path = u.replace(SITE, "") || "/";
    return path !== "/" && !linked.includes(path);
  });
}

/** Links internos quebrados (amostra: todos os destinos únicos das páginas auditadas) e páginas órfãs. */
async function checkInternalLinks(urls, pages) {
  const targets = internalTargets(pages);
  const brokenList = await findBroken(targets);
  add(
    "site",
    "técnico",
    "sem-links-internos-quebrados",
    brokenList.length === 0,
    `${brokenList.length}/${targets.size} quebrados: ${brokenList.join(", ")}`,
  );
  const orphanCandidates = findOrphans(urls, targets);
  add(
    "site",
    "técnico",
    "sem-paginas-orfas-no-sitemap",
    orphanCandidates.length === 0,
    orphanCandidates.join(", "),
  );
}

/** Links únicos para os repositórios do autor encontrados nas páginas auditadas. */
function githubLinks(pages) {
  const links = new Set();
  for (const pg of pages.filter(Boolean))
    for (const a of pg.p.anchors)
      if (
        /^https:\/\/github\.com\/evandrodevbr\/[\w.-]+\/?$/.test(a.href ?? "")
      )
        links.add(a.href.replace(/\/$/, ""));
  return links;
}

/** Repositórios que dão 404 (ou não respondem) para visitantes. */
async function findBrokenRepos(links) {
  const broken = [];
  for (const url of links) {
    const r = await get(url, UA, { method: "HEAD" }).catch(() => null);
    if (!r || r.res.status === 404)
      broken.push(url.replace("https://github.com/evandrodevbr/", ""));
  }
  return broken;
}

/** Links externos para os repositórios do autor: 404 = projeto com "Ver código" quebrado (repositório privado). */
async function checkGithubLinks(pages) {
  const ghLinks = githubLinks(pages);
  const ghBroken = await findBrokenRepos(ghLinks);
  add(
    "site",
    "técnico",
    "links-github-acessiveis",
    ghBroken.length === 0,
    ghBroken.length
      ? `404 para visitantes: ${ghBroken.join(", ")}`
      : `${ghLinks.size} repositórios`,
  );
}

/** Quantas verificações passaram em cada grupo. */
function summarizeGroups() {
  const groups = {};
  for (const c of checks) {
    groups[c.group] ??= { ok: 0, total: 0 };
    const g = groups[c.group];
    g.total++;
    if (c.ok) g.ok++;
  }
  return groups;
}

/** As falhas agrupadas pelo id da verificação. */
function failuresById() {
  const byId = {};
  for (const f of checks.filter((c) => !c.ok)) {
    byId[f.id] ??= [];
    byId[f.id].push(f);
  }
  return byId;
}

function printReport(urls, groups) {
  console.log(`\nAuditoria ${BASE} — ${urls.length} URLs`);
  for (const [g, v] of Object.entries(groups))
    console.log(`  ${g.padEnd(20)} ${v.ok}/${v.total}`);
  console.log(`\nFalhas (${checks.filter((c) => !c.ok).length}):`);
  for (const [id, list] of Object.entries(failuresById()))
    console.log(
      `  - ${id} (${list.length}): ${list
        .slice(0, 4)
        .map(
          (f) => `${f.scope}${f.detail ? ` [${f.detail.slice(0, 70)}]` : ""}`,
        )
        .join(" | ")}`,
    );
}

async function main() {
  const urls = await checkSiteFiles();
  await checkBots();
  const pages = await auditPages(urls);
  await checkInternalLinks(urls, pages);
  await checkGithubLinks(pages);

  const groups = summarizeGroups();
  const out = {
    base: BASE,
    at: new Date().toISOString(),
    urls: urls.length,
    groups,
    checks,
  };
  if (args.out) writeFileSync(args.out, JSON.stringify(out, null, 2));
  printReport(urls, groups);
}

main().catch((e) => {
  console.error(e);
  process.exit(1);
});