Corrige busca de feeds e atualização de conteúdo

messages/en.json

@@ -773,6 +773,7 @@
     "privateNetwork": "The destination points to an unauthorized private network.",
     "responseTooLarge": "The response exceeds the maximum allowed size.",
     "tooManyRedirects": "Too many redirects.",
+    "insecureRedirect": "Refused to follow a redirect from HTTPS to HTTP.",
     "httpStatus": "HTTP {status}",
     "connectFailed": "Connection failed: {reason}",
     "feedRefreshing": "This feed is already being refreshed.",

messages/es.json

@@ -773,6 +773,7 @@
     "privateNetwork": "El destino apunta a una red privada no autorizada.",
     "responseTooLarge": "La respuesta supera el tamaño máximo permitido.",
     "tooManyRedirects": "Demasiadas redirecciones.",
+    "insecureRedirect": "Redirección de HTTPS a HTTP rechazada.",
     "httpStatus": "HTTP {status}",
     "connectFailed": "Fallo de conexión: {reason}",
     "feedRefreshing": "Este feed ya se está actualizando.",

messages/pt-BR.json

@@ -773,6 +773,7 @@
     "privateNetwork": "O destino aponta para uma rede privada não autorizada.",
     "responseTooLarge": "A resposta excede o tamanho máximo permitido.",
     "tooManyRedirects": "Redirecionamentos demais.",
+    "insecureRedirect": "Redirecionamento de HTTPS para HTTP recusado.",
     "httpStatus": "HTTP {status}",
     "connectFailed": "Falha na conexão: {reason}",
     "feedRefreshing": "Este feed já está sendo atualizado.",

prisma/postgres/migrations/20261112180000_article_content_hash/migration.sql

@@ -0,0 +1,2 @@
+-- AlterTable
+ALTER TABLE "article" ADD COLUMN "contentHash" TEXT;

prisma/postgres/schema.prisma

@@ -155,6 +155,8 @@ model Article {
   contentHtml     String?
   snippet         String?
   imageUrl        String?
+  /// Impressão do título, link e HTML crus. Muda quando o feed republica o item.
+  contentHash     String?
   publishedAt     DateTime  @default(now())
   fullContentHtml String?
   storyId         String? // artigos de feeds diferentes sobre o mesmo fato compartilham este id

prisma/schema.base.prisma

@@ -149,6 +149,8 @@ model Article {
   contentHtml     String?
   snippet         String?
   imageUrl        String?
+  /// Impressão do título, link e HTML crus. Muda quando o feed republica o item.
+  contentHash     String?
   publishedAt     DateTime  @default(now())
   fullContentHtml String?
   storyId         String? // artigos de feeds diferentes sobre o mesmo fato compartilham este id

prisma/sqlite/migrations/20261112180000_article_content_hash/migration.sql

@@ -0,0 +1,2 @@
+-- AlterTable
+ALTER TABLE "article" ADD COLUMN "contentHash" TEXT;

prisma/sqlite/schema.prisma

@@ -155,6 +155,8 @@ model Article {
   contentHtml     String?
   snippet         String?
   imageUrl        String?
+  /// Impressão do título, link e HTML crus. Muda quando o feed republica o item.
+  contentHash     String?
   publishedAt     DateTime  @default(now())
   fullContentHtml String?
   storyId         String? // artigos de feeds diferentes sobre o mesmo fato compartilham este id

src/lib/feeds/discover.ts

@@ -2,7 +2,7 @@ import "server-only";
 import { staticTranslator } from "@/i18n/static";
 import { LocalizedError } from "@/lib/localized-error";
 import { commonPathCandidates, extractFeedCandidates, platformFeedCandidates, youtubeFeedFromHtml } from "./discover-candidates";
-import { FetchError, safeFetch, type SafeFetchResult } from "./net";
+import { FetchError, privateFeedsAllowed, safeFetch, type SafeFetchResult } from "./net";
 import { looksLikeFeed, parseFeed, type ParsedFeed } from "./parse";
 
 export type DiscoveredFeed = { url: string; title: string; siteUrl: string | null; itemCount: number; feed: ParsedFeed };
@@ -108,7 +108,7 @@ class Discovery {
 
   private async probe(url: string): Promise<Probe> {
     try {
-      const res = await safeFetch(url, { timeoutMs: PROBE_TIMEOUT_MS });
+      const res = await safeFetch(url, { timeoutMs: PROBE_TIMEOUT_MS, allowPrivateNetwork: privateFeedsAllowed() });
       this.tried.add(res.url);
       if (looksLikeFeed(res.body)) return { feed: await toDiscovered(res) };
       return { html: res.body, url: res.url };
@@ -166,7 +166,7 @@ export async function discoverFeeds(input: string): Promise<DiscoveredFeed[]> {
   let pageError: unknown = null;
   discovery.markTried(url);
   try {
-    page = await safeFetch(url);
+    page = await safeFetch(url, { allowPrivateNetwork: privateFeedsAllowed() });
   } catch (error) {
     if (isFatal(error)) throw error;
     pageError = error;

src/lib/feeds/extract.ts

@@ -18,8 +18,8 @@ export type ExtractedPage = {
 };
 
 /** Baixa a página e extrai o conteúdo principal (modo leitura) e os metadados. */
-export async function extractPage(url: string): Promise<ExtractedPage> {
-  const res = await safeFetch(url, { accept: "text/html,application/xhtml+xml" });
+export async function extractPage(url: string, opts: { allowPrivateNetwork?: boolean } = {}): Promise<ExtractedPage> {
+  const res = await safeFetch(url, { accept: "text/html,application/xhtml+xml", allowPrivateNetwork: opts.allowPrivateNetwork ?? false });
   const { document } = parseHTML(res.body);
   const pageTitle = document.querySelector("title")?.textContent?.trim() || null;
   const base = document.createElement("base");

src/lib/feeds/net.ts

@@ -1,4 +1,5 @@
 import "server-only";
+import { isIP } from "node:net";
 import { pinnedFetch } from "@/lib/network";
 import { LocalizedError, type ServerErrorKey } from "@/lib/localized-error";
 
@@ -34,10 +35,38 @@ export function parseRetryAfter(value: string | null, now = Date.now()): number
   return Number.isNaN(date) ? undefined : Math.max(0, Math.round((date - now) / 1000));
 }
 
+/** Feeds na rede local (RSS-Bridge). Não vale para capa nem extração de artigo. */
+export const privateFeedsAllowed = () => process.env.ALLOW_PRIVATE_FEEDS === "true";
+
+const MULTI_PART_SUFFIX = new Set([
+  "com.br", "net.br", "org.br", "gov.br",
+  "co.uk", "org.uk", "com.au", "co.jp", "co.nz", "com.mx", "com.ar", "com.tr", "co.za",
+  "com.cn", "com.tw", "co.kr", "com.sg", "com.pt", "com.co",
+]);
+
+function registrableDomain(hostname: string): string {
+  const host = hostname.replace(/\.$/, "").toLowerCase();
+  if (isIP(host)) return host;
+  const parts = host.split(".").filter(Boolean);
+  if (parts.length < 2) return host;
+  const tail = parts.slice(-2).join(".");
+  if (MULTI_PART_SUFFIX.has(tail) && parts.length >= 3) return parts.slice(-3).join(".");
+  return tail;
+}
+
+/** Mesmo site (domínio registrável), para aceitar uma URL nova de feed depois de um 301/308. */
+export function sameSite(a: string, b: string): boolean {
+  try {
+    return registrableDomain(new URL(a).hostname) === registrableDomain(new URL(b).hostname);
+  } catch {
+    return false;
+  }
+}
+
 /** fetch com timeout, limite de tamanho, redirects checados e proteção SSRF. */
 export async function safeFetch(
   input: string,
-  opts: { headers?: Record<string, string>; timeoutMs?: number; maxBytes?: number; accept?: string } = {},
+  opts: { headers?: Record<string, string>; timeoutMs?: number; maxBytes?: number; accept?: string; allowPrivateNetwork?: boolean } = {},
 ): Promise<SafeFetchResult> {
   const { timeoutMs = 15000, maxBytes = 5 * 1024 * 1024 } = opts;
   let url = new URL(input);
@@ -55,25 +84,29 @@ export async function safeFetch(
           headers: {
             "user-agent": USER_AGENT,
             accept: opts.accept ?? "application/rss+xml, application/atom+xml, application/xml, text/xml, text/html;q=0.8, */*;q=0.5",
-            ...opts.headers,
+            // Cache condicional só no primeiro hop: um 304 do destino do redirect não é "o feed não mudou".
+            ...(hop === 0 ? opts.headers : {}),
           },
         },
-        process.env.ALLOW_PRIVATE_FEEDS === "true",
+        opts.allowPrivateNetwork ?? false,
       );
     } catch (error) {
       if (error instanceof LocalizedError) throw error;
       throw new FetchError("connectFailed", { reason: (error instanceof Error ? error.message : String(error)).slice(0, 300) });
     }
     try {
       const res = connection.response;
+      const movedPermanently = redirects > 0 && allPermanent;
+      // 304 não é redirect, mesmo com Location.
+      if (res.status === 304) return { status: 304, url: url.toString(), headers: res.headers, body: "", movedPermanently };
       if (res.status >= 300 && res.status < 400 && res.headers.get("location")) {
-        url = new URL(res.headers.get("location")!, url);
+        const next = new URL(res.headers.get("location")!, url);
+        if (url.protocol === "https:" && next.protocol !== "https:") throw new FetchError("insecureRedirect");
+        url = next;
         redirects++;
         allPermanent &&= res.status === 301 || res.status === 308;
         continue;
       }
-      const movedPermanently = redirects > 0 && allPermanent;
-      if (res.status === 304) return { status: 304, url: url.toString(), headers: res.headers, body: "", movedPermanently };
       if (!res.ok) throw new FetchError("httpStatus", { status: res.status }, res.status, parseRetryAfter(res.headers.get("retry-after")));
 
       const reader = res.body?.getReader();
@@ -99,14 +132,25 @@ export async function safeFetch(
   throw new FetchError("tooManyRedirects");
 }
 
-function decode(buffer: Buffer, headers: { get(name: string): string | null }): string {
-  const fromHeader = /charset=([^;]+)/i.exec(headers.get("content-type") ?? "")?.[1];
+function unquoteCharset(value: string | undefined): string | undefined {
+  if (!value) return undefined;
+  const cleaned = value.trim().replace(/^["']+|["']+$/g, "").trim().toLowerCase();
+  return cleaned || undefined;
+}
+
+/** Decodifica o corpo usando o charset do header (sem aspas) ou da declaração do XML. */
+export function decodeFeedBody(buffer: Buffer, contentType: string | null): string {
+  const fromHeader = unquoteCharset(/charset=\s*("[^"]*"|'[^']*'|[^;\s]+)/i.exec(contentType ?? "")?.[1]);
   const head = buffer.subarray(0, 1024).toString("latin1");
-  const fromBody = /encoding=["']([^"']+)["']/i.exec(head)?.[1] ?? /<meta[^>]+charset=["']?([\w-]+)/i.exec(head)?.[1];
-  const charset = (fromHeader ?? fromBody ?? "utf-8").trim().toLowerCase();
+  const fromBody = unquoteCharset(/encoding=["']([^"']+)["']/i.exec(head)?.[1] ?? /<meta[^>]+charset=["']?([\w-]+)/i.exec(head)?.[1]);
+  const charset = fromHeader ?? fromBody ?? "utf-8";
   try {
     return new TextDecoder(charset).decode(buffer);
   } catch {
     return new TextDecoder("utf-8").decode(buffer);
   }
 }
+
+function decode(buffer: Buffer, headers: { get(name: string): string | null }): string {
+  return decodeFeedBody(buffer, headers.get("content-type"));
+}