Apunta los scripts de sync prod al Hetzner nuevo y suma trabajo pendiente

- sync_translations_to_prod.py / sync_audio_to_prod.py / sync_carta_from_prod.py:
  migran de FEA_PROD_HOST/PASS (CDMON, password) a FEA_PROD_SSH_HOST/PASS +
  FEA_PROD_DOCKER_CONTAINER (Hetzner/Coolify, auth por clave), con wrapping
  docker exec y el fix del bug de redirecciones (wc -c/cat) resuelto en el host
  en vez de dentro del contenedor.
- fea_translate_helper.php: subcomando clone_new para clonar en ID local nuevo
  cuando el ID de prod ya está ocupado localmente.
- translate_post.py: --dry-run.
- tts_produce.py: --allow-default-voice (voz Nico solo si se permite
  explícitamente para autores sin voz clonada) + fix voice_for_author.
- minimax_tts.py: parámetro speed en t2a/_synth_chunk.
- mirror-antiguo/deploy/nginx-mirror.conf: redirects de URLs históricas de
  catálogo y vista imprimible EFFA.
- Importaciones humanas de Pagola (carta 738) + scripts/manifest de la
  fase A Enrique, y release_issue181_prod.sh / cdmon-retire-fewp1.sh.
- .gitignore: excluye docs/backups/ (dumps SQL, mismo motivo que backups/*).

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

.gitignore

@@ -17,6 +17,10 @@ joomla-php83/
 backups/*
 !backups/README.md
 
+# Backups pesados (GBs) bajo docs/ — mismo motivo que backups/*, ruta distinta
+docs/backups/*
+!docs/backups/README.md
+
 # Capturas de pantalla
 capturas/
 

docs/releases/enrique-2026-07-25-prod-dry-run.md

@@ -0,0 +1,71 @@
+# Release dry-run — Enrique Martínez Lozano / carta «Hacia el corazón»
+
+**Estado:** preparado localmente; **no ejecutado en producción**.
+
+## Validación local completada
+
+- Comentario ES: `#55041` — `El tesoro está ya en nosotros`
+  - Categorías: `Comentarios al evangelio` + `Feadulta`.
+  - Audio: `tts/55041.mp3`, 2,221,101 bytes, voz `NicoFeadulta2026`.
+  - QA Haiku: aprobado tras correcciones objetivas EN/FR/IT.
+- Traducciones del comentario: EN `#55047`, FR `#55048`, IT `#55049`, PT `#55050`.
+- Carta local ES: `#55042` — `Hacia el corazón` (slug técnico de preview conservado).
+  - Traducciones: EN `#55051`, FR `#55052`, IT `#55053`, PT `#55054`.
+  - QA Haiku: contenido aprobado tras correcciones; 24 enlaces con destino traducido se repuntaron. Los 12 enlaces externos que quedan son fallback ES: no existe post destino traducido en local (verificado por `post_title` exacto y Polylang).
+- Render comprobado en Tailscale:
+  - Comentario ES: título correcto, categorías correctas y reproductor HTML5 visible.
+  - Carta FR: título correcto; Enrique aparece tras Fray Marcos y antes de Pagola.
+
+## Evidencia de producción (sólo lectura server-side)
+
+- Carta real: `#54914`, título `Hacia el corazón`, estado `publish`.
+- Hash actual de contenido prod: `5e7eb1b880bfc6311edb78ac8b6e81639f5b4502282a1b182c610b1d01308b69`.
+- Hash candidato local: `6fdb4929fcb5b49b7f2dc1deb7c6f5be70983af866d44c5d5c27d2866c09418c`.
+- La carta candidata lleva `fea_phase_a_source_prod_id=54914`; por tanto debe **actualizar #54914**, no crearse una carta nueva.
+
+## Dry-runs que pasaron
+
+```bash
+# Crea solamente el grupo nuevo del comentario (ES + EN/FR/IT/PT): 5 posts draft
+python3 scripts/sync_translations_to_prod.py --ids 55041 --dry-run
+
+# Sube y asocia solamente el MP3 de Enrique
+python3 scripts/sync_audio_to_prod.py --ids 55041 --dry-run
+```
+
+Ambos devolvieron `ok/error=1/0` (audio) y un grupo Polylang de cinco posts (comentario).
+
+## Guardarraíl crítico
+
+**NO ejecutar:**
+
+```bash
+python3 scripts/sync_translations_to_prod.py --ids 55041,55042
+```
+
+El dry-run demostró que clonaría `#55042` como una segunda carta ES. No actualiza la carta real `#54914`.
+
+## Release que debe ejecutar Rafa (producción)
+
+1. Backup de producción y snapshot server-side de `#54914`.
+2. Crear el grupo nuevo del comentario desde `#55041` (ES + EN/FR/IT/PT), inicialmente en `draft`; anotar los IDs reales de producción.
+3. Subir `tts/55041.mp3` y asociarlo al ID ES real creado en producción.
+4. Actualizar **solamente** título/contenido de la carta real `#54914` con el candidato local `#55042`; mantener título `Hacia el corazón`, autor, fecha y estado existentes.
+5. Crear EN/FR/IT/PT de la carta contra el origen real `#54914` y enlazarlas en Polylang. No clonar otra ES.
+6. Releer por servidor los IDs creados y `#54914`; comprobar:
+   - Enrique sólo está en `Evangelio y comentarios al Evangelio`, tras Fray Marcos y antes de Pagola.
+   - No aparece en `Artículos seleccionados`.
+   - Cada carta enlaza a Enrique en su mismo idioma.
+   - El reproductor del ES carga `tts/55041.mp3`.
+7. Sólo tras esas lecturas, promover los drafts de producción a `publish`.
+
+## Rollback
+
+- Restaurar el backup de producción previo.
+- Alternativamente, restaurar el contenido/hash previo de `#54914` y despublicar el nuevo grupo + audio de Enrique.
+
+## Backups locales relacionados
+
+- `docs/backups/enrique-translations-20260725T004921Z/`
+- `docs/backups/enrique-carta-fr-final-20260725T012649Z/`
+- `docs/backups/enrique-pre-tts-20260725T013021Z/`

import_avatars_62.php

@@ -0,0 +1,50 @@
+<?php
+/**
+ * Issue #62 — Reapunta foto_perfil de cada autor a su avatar nuevo.
+ * Crea un attachment por autor y guarda el foto_perfil antiguo en _foto_perfil_pre62 (revertible).
+ * Uso: wp eval-file import_avatars_62.php [--apply]
+ *   sin --apply => dry run (no escribe nada).
+ */
+require_once ABSPATH . 'wp-admin/includes/image.php';
+
+$apply = getenv('APPLY') === '1';
+$mfile = getenv('MANIFEST') ?: 'manifest_62.json';
+$updir = wp_get_upload_dir();
+$manifest = json_decode(file_get_contents($updir['basedir'] . '/avatares/' . $mfile), true);
+
+$done = $skip = $err = 0;
+foreach ($manifest as $row) {
+    list($uid, $name, $orig, $kind, $old_attach) = $row;
+    if (!in_array($kind, ['initials','trim','logo','qs'], true)) { $skip++; continue; }
+
+    $rel  = "avatares/autores/autor-{$uid}.png";
+    $abs  = $updir['basedir'] . '/' . $rel;
+    if (!file_exists($abs)) { echo "MISSING file uid=$uid\n"; $err++; continue; }
+
+    // si foto_perfil ya apunta a este fichero, solo se ha sobrescrito el PNG: no crear attachment nuevo
+    $cur = (int) get_user_meta($uid, 'foto_perfil', true);
+    if ($cur && get_post_meta($cur, '_wp_attached_file', true) === $rel) {
+        if ($apply) wp_update_attachment_metadata($cur, wp_generate_attachment_metadata($cur, $abs));
+        $skip++; continue;
+    }
+
+    if (!$apply) { $done++; continue; }
+
+    // guardar foto_perfil antiguo una sola vez (idempotente)
+    if (get_user_meta($uid, '_foto_perfil_pre62', true) === '') {
+        update_user_meta($uid, '_foto_perfil_pre62', $old_attach);
+    }
+
+    $attach = [
+        'post_mime_type' => 'image/png',
+        'post_title'     => "Avatar {$name}",
+        'post_status'    => 'inherit',
+        'guid'           => $updir['baseurl'] . '/' . $rel,
+    ];
+    $aid = wp_insert_attachment($attach, $abs, 0, true);
+    if (is_wp_error($aid)) { echo "ERR insert uid=$uid: ".$aid->get_error_message()."\n"; $err++; continue; }
+    wp_update_attachment_metadata($aid, wp_generate_attachment_metadata($aid, $abs));
+    update_user_meta($uid, 'foto_perfil', $aid);
+    $done++;
+}
+echo ($apply ? "APLICADO" : "DRY-RUN") . ": procesados=$done omitidos=$skip errores=$err\n";

mirror-antiguo/deploy/nginx-mirror.conf

@@ -13,6 +13,21 @@ server {
     # application/octet-stream y el navegador se los descarga en vez de mostrarlos.
     default_type text/html;
 
+    # URLs históricas de catálogo: el catálogo vigente vive en Ediciones Fe Adulta.
+    # 302 primero: evita cachear un destino externo de forma irreversible durante el soak.
+    location ~ ^/(?:es/)?catalogo-de-libros-feadulta(?:/|\.html)?$ {
+        return 302 https://edicionesfeadulta.com/;
+    }
+
+    # La vista imprimible Joomla de esta página EFFA no existe en el archivo estático.
+    # Conserva la página normal del archivo; sólo la variante ?print=1 va al contenido WP vigente.
+    location = /es/effa/95-secc5cat/3304-seccion5col00.html {
+        if ($arg_print = 1) {
+            return 302 https://www.feadulta.com/seccion5col00-2/;
+        }
+        try_files $uri $uri.html $uri/index.html =404;
+    }
+
     location / {
         try_files $uri $uri.html $uri/index.html =404;
     }

mirror-antiguo/stopped-containers.txt.hecho-20260730T014132Z

@@ -0,0 +1,10 @@
+portfolio-tracker
+yt-api-bridge
+wordpress-web
+joomla-web-php83
+jellyfin
+triptyk-local-wordpress-1
+triptyk-local-db-1
+yt-summaries
+joomla-web
+wordpress-mysql

scripts/cdmon-retire-fewp1.sh

@@ -0,0 +1,37 @@
+#!/usr/bin/env bash
+# Retira exclusivamente el esquema vacío fewp1 de CDMON.
+# Por defecto solo hace dry-run. La eliminación real exige --apply y el token exacto.
+set -euo pipefail
+
+mode="dry-run"
+if [[ "${1:-}" == "--apply" ]]; then
+  mode="apply"
+fi
+
+if [[ "$mode" == "apply" && "${CONFIRM_DROP_FEWP1:-}" != "DROP_FEWP1" ]]; then
+  echo "Refusing apply: export CONFIRM_DROP_FEWP1=DROP_FEWP1 first." >&2
+  exit 2
+fi
+
+set -a
+source "$HOME/.hermes/profiles/feadulta/.env"
+set +a
+
+php='global $wpdb;
+$count=(int)$wpdb->get_var("SELECT COUNT(*) FROM information_schema.tables WHERE table_schema=\"fewp1\"");
+if ($count !== 0) { fwrite(STDERR, "ABORT: fewp1 has ".$count." tables; it is not empty.\n"); exit(3); }
+echo "PASS: fewp1 exists and has 0 tables.\n";'
+
+if [[ "$mode" == "apply" ]]; then
+  php+=' $ok=$wpdb->query("DROP DATABASE `fewp1`");
+if ($ok === false) { fwrite(STDERR, "ABORT: DROP DATABASE failed.\n"); exit(4); }
+$remaining=(int)$wpdb->get_var("SELECT COUNT(*) FROM information_schema.schemata WHERE schema_name=\"fewp1\"");
+if ($remaining !== 0) { fwrite(STDERR, "ABORT: fewp1 remains after DROP.\n"); exit(5); }
+echo "APPLIED: fewp1 removed; verified absent.\n";'
+else
+  php+=' echo "DRY-RUN: would execute DROP DATABASE `fewp1`; no production change made.\n";'
+fi
+
+printf -v quoted_php '%q' "$php"
+sshpass -p "$FEA_PROD_SSH_PASS" ssh -o StrictHostKeyChecking=accept-new "$FEA_PROD_SSH_HOST" \
+  "cd /web && wp eval $quoted_php"

scripts/fea_translate_helper.php

@@ -262,6 +262,47 @@ switch ($cmd) {
         break;
     }
 
+    case 'clone_new': {
+        // Clona en un ID local NUEVO. Se usa al importar desde prod cuando el
+        // mismo ID ya está ocupado por contenido local distinto.
+        $lang   = (string) ($argv[2] ?? '');
+        $status = (string) ($argv[3] ?? '');
+        if ($lang === '') {
+            fwrite(STDERR, "uso: clone_new <lang> <status>\n"); exit(6);
+        }
+        $payload = json_decode(file_get_contents('php://stdin'), true);
+        if (!is_array($payload) || empty($payload['title'])) {
+            fwrite(STDERR, "payload inválido por stdin\n"); exit(4);
+        }
+        $postarr = [
+            'post_title'   => wp_slash($payload['title']),
+            'post_content' => wp_slash($payload['content'] ?? ''),
+            'post_excerpt' => wp_slash($payload['excerpt'] ?? ''),
+            'post_status'  => $status ?: ($payload['status'] ?? 'draft'),
+            'post_type'    => $payload['type'] ?? 'post',
+            'post_author'  => (int) ($payload['author'] ?? 1),
+            'post_date'    => $payload['date'] ?? current_time('mysql'),
+            'post_date_gmt'=> $payload['date_gmt'] ?? current_time('mysql', true),
+            'post_name'    => $payload['slug'] ?? '',
+            'to_ping'      => '',
+            'pinged'       => '',
+        ];
+        $new_id = wp_insert_post($postarr, true);
+        if (is_wp_error($new_id)) { fwrite(STDERR, $new_id->get_error_message() . "\n"); exit(5); }
+        pll_set_post_language($new_id, $lang);
+        $cats = [];
+        foreach ((array) ($payload['cat_slugs'] ?? []) as $slug) {
+            $term = get_term_by('slug', (string) $slug, 'category');
+            if ($term && !is_wp_error($term)) $cats[] = (int) $term->term_id;
+        }
+        if (!$cats) $cats = array_values(array_unique(array_map('intval', (array) ($payload['cats'] ?? []))));
+        wp_set_post_categories($new_id, $cats);
+        set_meta_payload($new_id, normalize_meta_input($payload));
+        clean_post_cache($new_id);
+        echo $new_id;
+        break;
+    }
+
     case 'carta_sections': {
         // Descubre el cluster de una carta parseando sus propios enlaces internos
         // (misma lógica que pinta la portada, fea-carta-portada.php). No depende

scripts/import_human_pagola_translations.py

@@ -0,0 +1,135 @@
+#!/usr/bin/env python3
+"""Importa las traducciones editoriales humanas adjuntas a un issue de carta.
+
+Por defecto solo valida y muestra el plan. --apply-local crea borradores en el
+WordPress Docker local; nunca publica ni toca producción.
+"""
+from __future__ import annotations
+
+import argparse
+import hashlib
+import html
+import json
+import re
+import subprocess
+import tempfile
+import zipfile
+from pathlib import Path
+from xml.etree import ElementTree as ET
+
+ROOT = Path(__file__).resolve().parent
+HELPER = ROOT / "fea_translate_helper.php"
+CONTAINER = "wordpress-web"
+SOURCE_ID = 55465  # Pagola, Carta 738
+LANG_BY_NAME = {"2_eng": "en", "3_fr": "fr", "4_it": "it", "5_pt": "pt"}
+TRANSLATOR_LABELS = ("Translator:", "Traducteur:", "Traduzzione:", "Tradutor:")
+
+
+def docx_lines(path: Path) -> list[str]:
+    ns = {"w": "http://schemas.openxmlformats.org/wordprocessingml/2006/main"}
+    with zipfile.ZipFile(path) as zf:
+        root = ET.fromstring(zf.read("word/document.xml"))
+    return [
+        "".join(t.text or "" for t in p.findall(".//w:t", ns)).strip()
+        for p in root.findall(".//w:p", ns)
+        if "".join(t.text or "" for t in p.findall(".//w:t", ns)).strip()
+    ]
+
+
+def value_after(lines: list[str], label: str) -> str:
+    try:
+        return lines[lines.index(label) + 1]
+    except (ValueError, IndexError) as exc:
+        raise ValueError(f"Falta {label!r}") from exc
+
+
+def parse_doc(path: Path) -> dict:
+    lines = docx_lines(path)
+    title = value_after(lines, "Título:")
+    excerpt = value_after(lines, "Entradilla:")
+    author = value_after(lines, "Autor:")
+    start = lines.index("Cuerpo:") + 1
+    end = next((i for i in range(start, len(lines))
+                if any(lines[i].startswith(label) for label in TRANSLATOR_LABELS)), len(lines))
+    body = lines[start:end]
+    if not body:
+        raise ValueError("Cuerpo vacío")
+    translator = ""
+    if end < len(lines):
+        label = next(label for label in TRANSLATOR_LABELS if lines[end].startswith(label))
+        translator = lines[end][len(label):].strip()
+        if not translator and end + 1 < len(lines):
+            translator = lines[end + 1]
+    # La fuente española incluye el crédito/origen como último párrafo.
+    body.append('Publicado en: <a href="https://www.gruposdejesus.com">https://www.gruposdejesus.com</a>')
+    content = "\n".join(
+        f"<p>{line if line.startswith('Publicado en: <a ') else html.escape(line)}</p>" for line in body
+    )
+    return {
+        "title": title,
+        "excerpt": excerpt,
+        "author": author,
+        "translator": translator,
+        "content": content,
+        "source_sha256": hashlib.sha256(path.read_bytes()).hexdigest(),
+    }
+
+
+def run(cmd: list[str], stdin: str | None = None) -> str:
+    result = subprocess.run(cmd, input=stdin, text=True, capture_output=True)
+    if result.returncode:
+        raise RuntimeError(f"rc={result.returncode}: {result.stderr.strip()} {result.stdout.strip()}")
+    return result.stdout.strip()
+
+
+def helper(*args: str, stdin: str | None = None) -> str:
+    return run(["docker", "exec", "-i", CONTAINER, "php", "/tmp/fea_translate_helper.php", *args], stdin)
+
+
+def lang_for(path: Path) -> str:
+    for marker, lang in LANG_BY_NAME.items():
+        if marker in path.name:
+            return lang
+    raise ValueError(f"No reconozco idioma en {path.name}")
+
+
+def apply_meta(post_id: int, doc: Path, parsed: dict) -> None:
+    payload = json.dumps({"id": post_id, "doc": doc.name, "sha": parsed["source_sha256"],
+                          "translator": parsed["translator"]}, ensure_ascii=False)
+    php = r'''$p=json_decode(base64_decode(getenv('PAYLOAD')),true); update_post_meta($p['id'],'traduccion_automatica','0'); update_post_meta($p['id'],'traduccion_modelo','editorial-humana'); update_post_meta($p['id'],'traduccion_fuente_doc',$p['doc']); update_post_meta($p['id'],'traduccion_fuente_sha256',$p['sha']); update_post_meta($p['id'],'traduccion_editor',$p['translator']); echo 'ok';'''
+    import base64
+    encoded = base64.b64encode(payload.encode()).decode()
+    run(["docker", "exec", "-i", "-e", f"PAYLOAD={encoded}", CONTAINER,
+         "wp", "--allow-root", "eval", php])
+
+
+def main() -> int:
+    ap = argparse.ArgumentParser()
+    ap.add_argument("--dir", type=Path, required=True)
+    ap.add_argument("--apply-local", action="store_true")
+    args = ap.parse_args()
+    docs = sorted(args.dir.glob("*.docx"))
+    if len(docs) != 4:
+        raise SystemExit(f"Se esperaban 4 DOCX, encontrados {len(docs)}")
+    run(["docker", "cp", str(HELPER), f"{CONTAINER}:/tmp/fea_translate_helper.php"])
+    for doc in docs:
+        lang, parsed = lang_for(doc), parse_doc(doc)
+        existing = helper("exists", str(SOURCE_ID), lang)
+        plan = {"lang": lang, "doc": doc.name, "existing": int(existing or "0"),
+                "title": parsed["title"], "excerpt_chars": len(parsed["excerpt"]),
+                "content_chars": len(parsed["content"]), "translator": parsed["translator"]}
+        print(json.dumps(plan, ensure_ascii=False))
+        if not args.apply_local:
+            continue
+        if int(existing or "0"):
+            raise RuntimeError(f"{lang} ya existe como #{existing}; no sobrescribo")
+        payload = json.dumps({"title": parsed["title"], "excerpt": parsed["excerpt"],
+                              "content": parsed["content"], "model": "editorial-humana"}, ensure_ascii=False)
+        post_id = int(helper("create", str(SOURCE_ID), lang, "draft", stdin=payload))
+        apply_meta(post_id, doc, parsed)
+        print(json.dumps({"created": post_id, "lang": lang, "status": "draft"}))
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())

scripts/minimax_tts.py

@@ -361,12 +361,12 @@ def _split_for_tts(text, limit=CHAR_LIMIT):
     return chunks
 
 
-def _synth_chunk(text, voice_id, model):
+def _synth_chunk(text, voice_id, model, speed=1.0):
     """Una petición t2a. Devuelve (audio_bytes|None, rc, usage_chars)."""
     body = {
         "model": model,
         "text": text,
-        "voice_setting": {"voice_id": voice_id, "speed": 1.0, "vol": 1.0, "pitch": 0},
+        "voice_setting": {"voice_id": voice_id, "speed": speed, "vol": 1.0, "pitch": 0},
         "audio_setting": {"sample_rate": 32000, "bitrate": 128000, "format": "mp3", "channel": 1},
         "language_boost": "Spanish",
     }
@@ -381,13 +381,15 @@ def _synth_chunk(text, voice_id, model):
     return bytes.fromhex(audio_hex), 0, usage
 
 
-def t2a(text, voice_id, model, name):
+def t2a(text, voice_id, model, name, speed=1.0):
+    if not 0.5 <= float(speed) <= 2.0:
+        raise ValueError(f"speed fuera de rango: {speed} (permitido 0.5–2.0)")
     chunks = _split_for_tts(text)
-    print(f"Sintetizando {len(text)} car con {model} / {voice_id} "
+    print(f"Sintetizando {len(text)} car con {model} / {voice_id} a {float(speed):.2f}× "
           f"({len(chunks)} petición/es)…", flush=True)
     raw = OUT / f"{name}.raw.mp3"
     if len(chunks) == 1:
-        audio, rc, _ = _synth_chunk(chunks[0], voice_id, model)
+        audio, rc, _ = _synth_chunk(chunks[0], voice_id, model, speed)
         if audio is None:
             return rc
         raw.write_bytes(audio)
@@ -398,7 +400,7 @@ def t2a(text, voice_id, model, name):
                 import os as _os, time as _t
                 _t.sleep(int(_os.environ.get("FEA_CHUNK_PAUSE", "35")))  # respetar TPM de MiniMax
             print(f"  trozo {k + 1}/{len(chunks)} ({len(ch)} car)…", flush=True)
-            audio, rc, _ = _synth_chunk(ch, voice_id, model)
+            audio, rc, _ = _synth_chunk(ch, voice_id, model, speed)
             if audio is None:
                 for p in parts:
                     p.unlink(missing_ok=True)

scripts/phase_a_enrique_preview.py

@@ -0,0 +1,151 @@
+#!/usr/bin/env python3
+"""Fase A local: borrador de Enrique + copia draft de la carta para revisión.
+
+Por defecto no escribe WordPress. --apply-local sólo modifica el WordPress Docker local.
+Nunca publica ni toca producción.
+"""
+import argparse
+import base64
+import hashlib
+import html
+import json
+import os
+import re
+import subprocess
+import sys
+import zipfile
+from pathlib import Path
+from xml.etree import ElementTree as ET
+
+ROOT = Path(__file__).resolve().parent.parent
+DEFAULT_DOC = Path("/home/rafa/.hermes/cache/documents/doc_7909db597c74_7r_Enrique_Mart_nez_Lozano_tesoro_esta_ya_en_nosotros.docx")
+CARTA_PROD_ID = 54914
+AUTHOR_ID = 384  # Enrique Martínez Lozano, verificado localmente.
+CATEGORIES = [1650, 71]  # Artículos + Feadulta
+SLUG = "el-tesoro-esta-ya-en-nosotros"
+
+
+def docx_text(path: Path) -> list[str]:
+    ns = {"w": "http://schemas.openxmlformats.org/wordprocessingml/2006/main"}
+    with zipfile.ZipFile(path) as zf:
+        root = ET.fromstring(zf.read("word/document.xml"))
+    out = []
+    for p in root.findall(".//w:p", ns):
+        text = "".join(t.text or "" for t in p.findall(".//w:t", ns)).strip()
+        if text:
+            out.append(text)
+    return out
+
+
+def parse_source(path: Path) -> dict:
+    lines = docx_text(path)
+    def after(label: str) -> str:
+        return lines[lines.index(label) + 1]
+    title = after("Título:")
+    author = after("Autor:")
+    assert title == "EL TESORO ESTÁ YA EN NOSOTROS", title
+    assert author == "ENRIQUE MARTÍNEZ LOZANO", author
+    body_start = lines.index("Cuerpo:") + 1
+    body = lines[body_start:]
+    # Última firma del boletín no es contenido del post: el autor va en WP.
+    if body[-1].startswith("ENRIQUE MARTÍNEZ LOZANO"):
+        body = body[:-1]
+    date_line, bible, *paragraphs = body
+    intro = lines[lines.index("Entradilla:") + 1]
+    content = "\n".join(
+        [f"<p><strong>{html.escape(date_line)}</strong></p>",
+         f"<p><strong>{html.escape(bible)}</strong></p>"]
+        + [f"<p>{html.escape(p)}</p>" for p in paragraphs]
+    )
+    return {"title": title.lower().capitalize(), "author": author.title(), "intro": intro,
+            "content": content, "source_sha256": hashlib.sha256(path.read_bytes()).hexdigest()}
+
+
+def prod_card() -> dict:
+    sys.path.insert(0, str(ROOT / "scripts"))
+    import sync_translations_to_prod as sync
+    return json.loads(sync.prod_helper("read_full", str(CARTA_PROD_ID)))
+
+
+def carta_candidate(card: dict, anchor: str) -> str:
+    marker = '<p>&nbsp;</p>\n<p><span style="color: #ff0000;"><strong>Para unas eucaristías más participativas y actuales</strong></span></p>'
+    assert card["content"].count(marker) == 1, "Marcador de sección no único/no encontrado"
+    return card["content"].replace(marker, anchor + "\n" + marker, 1)
+
+
+def run_wp(payload: dict, apply: bool) -> dict:
+    encoded = base64.b64encode(json.dumps(payload, ensure_ascii=False).encode()).decode()
+    php = r'''
+$p=json_decode(base64_decode(getenv('FEA_PAYLOAD')), true);
+$apply=getenv('FEA_APPLY') === '1';
+$existing=get_page_by_path($p['slug'], OBJECT, 'post');
+$author=get_user_by('id',(int)$p['author_id']);
+$cats=array_map('intval',$p['categories']);
+$cat_ok=count(array_filter($cats, fn($id)=>term_exists($id,'category')))===count($cats);
+$preview=get_posts(['post_type'=>'post','post_status'=>'any','meta_key'=>'fea_phase_a_source_hash','meta_value'=>$p['source_sha256'],'numberposts'=>1]);
+$out=['dry_run'=>!$apply,'existing_slug'=>$existing?['id'=>$existing->ID,'status'=>$existing->post_status]:null,
+      'author'=>$author?['id'=>$author->ID,'name'=>$author->display_name]:null,'categories_ok'=>$cat_ok,
+      'preview_existing'=>$preview?['id'=>$preview[0]->ID,'status'=>$preview[0]->post_status]:null,
+      'candidate_anchor'=>$p['anchor']];
+if (!$author || !$cat_ok) { $out['error']='Autor o categorías no válidos'; echo wp_json_encode($out,JSON_UNESCAPED_UNICODE); exit(3); }
+if ($existing || $preview) { $out['error']='Ya existe un borrador/slug para esta Fase A; no se duplica'; echo wp_json_encode($out,JSON_UNESCAPED_UNICODE); exit(4); }
+if (!$apply) { echo wp_json_encode($out,JSON_UNESCAPED_UNICODE); exit; }
+$article_id=wp_insert_post(['post_type'=>'post','post_status'=>'draft','post_author'=>(int)$p['author_id'],
+  'post_title'=>$p['article_title'],'post_name'=>$p['slug'],'post_excerpt'=>$p['intro'],
+  'post_content'=>$p['article_content'],'post_category'=>$cats],true);
+if (is_wp_error($article_id)) { $out['error']=$article_id->get_error_message(); echo wp_json_encode($out,JSON_UNESCAPED_UNICODE); exit(5); }
+if (function_exists('pll_set_post_language')) pll_set_post_language($article_id,'es');
+update_post_meta($article_id,'fea_phase_a_source_hash',$p['source_sha256']);
+update_post_meta($article_id,'fea_phase_a_source_doc','7r_Enrique Martínez Lozano_tesoro_esta_ya_en_nosotros.docx');
+update_post_meta($article_id,'fea_phase_a_status','preview-only');
+$preview_id=wp_insert_post(['post_type'=>'post','post_status'=>'draft','post_author'=>(int)$p['card_author_id'],
+  'post_title'=>'PREVIEW — Hacia el corazón (+ Enrique Martínez Lozano)','post_excerpt'=>'Copia local de validación; no publicar.',
+  'post_content'=>$p['candidate_card_content']],true);
+if (is_wp_error($preview_id)) { wp_delete_post($article_id,true); $out['error']=$preview_id->get_error_message(); echo wp_json_encode($out,JSON_UNESCAPED_UNICODE); exit(6); }
+if (function_exists('pll_set_post_language')) pll_set_post_language($preview_id,'es');
+update_post_meta($preview_id,'fea_phase_a_source_hash',$p['source_sha256']);
+update_post_meta($preview_id,'fea_phase_a_source_prod_id',(int)$p['carta_prod_id']);
+update_post_meta($preview_id,'fea_phase_a_preview_article_id',$article_id);
+update_post_meta($preview_id,'fea_phase_a_status','preview-only');
+$out['article']=['id'=>$article_id,'status'=>'draft','slug'=>get_post_field('post_name',$article_id),'author'=>get_the_author_meta('display_name',(int)$p['author_id'])];
+$out['preview_card']=['id'=>$preview_id,'status'=>'draft','source_prod_id'=>(int)$p['carta_prod_id']];
+echo wp_json_encode($out,JSON_UNESCAPED_UNICODE);
+'''
+    cmd = [
+        "docker", "exec", "-i", "-e", f"FEA_PAYLOAD={encoded}",
+        "-e", f"FEA_APPLY={'1' if apply else '0'}", "wordpress-web",
+        "wp", "--allow-root", "eval", php,
+    ]
+    result = subprocess.run(cmd, text=True, capture_output=True, check=False)
+    if result.returncode:
+        raise RuntimeError(f"WP local rc={result.returncode}: {result.stderr}\n{result.stdout}")
+    return json.loads(result.stdout)
+
+
+def main():
+    ap = argparse.ArgumentParser()
+    ap.add_argument("--doc", type=Path, default=DEFAULT_DOC)
+    ap.add_argument("--apply-local", action="store_true")
+    ap.add_argument("--output", type=Path, required=True)
+    args = ap.parse_args()
+    source = parse_source(args.doc)
+    card = prod_card()  # lectura server-side únicamente.
+    article_url = "http://localhost:8080/" + SLUG + "/"
+    anchor = (f'<p><strong><a href="{article_url}">Enrique Martínez Lozano: '
+              f'{html.escape(source["title"])}.</a></strong> {html.escape(source["intro"])}</p>')
+    payload = {
+        "source_sha256": source["source_sha256"], "article_title": source["title"],
+        "article_content": source["content"], "intro": source["intro"], "slug": SLUG,
+        "author_id": AUTHOR_ID, "categories": CATEGORIES, "anchor": anchor,
+        "candidate_card_content": carta_candidate(card, anchor), "card_author_id": card["author"],
+        "carta_prod_id": CARTA_PROD_ID,
+    }
+    result = run_wp(payload, args.apply_local)
+    report = {"source": source, "prod_card": {"id": card["id"], "title": card["title"], "status": card["status"]},
+              "result": result, "mode": "apply-local" if args.apply_local else "dry-run"}
+    args.output.parent.mkdir(parents=True, exist_ok=True)
+    args.output.write_text(json.dumps(report, ensure_ascii=False, indent=2) + "\n", encoding="utf-8")
+    print(json.dumps(report, ensure_ascii=False, indent=2))
+
+if __name__ == "__main__":
+    main()

scripts/release_issue181_prod.sh

@@ -0,0 +1,167 @@
+#!/usr/bin/env bash
+# Release Fase 1+2, issue #181: 68 traducciones publicadas + 16 MP3 TTS.
+# Por defecto es sólo dry-run. --apply requiere confirmación explícita de Rafa.
+set -euo pipefail
+
+ROOT="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
+cd "$ROOT"
+
+MODE="dry-run"
+if [[ "${1:-}" == "--apply" ]]; then
+  MODE="apply"
+elif [[ "${1:-}" == "--verify" ]]; then
+  MODE="verify"
+elif [[ "${1:-}" != "" && "${1:-}" != "--dry-run" ]]; then
+  echo "Uso: $0 [--dry-run|--verify|--apply]" >&2
+  exit 2
+fi
+
+# shellcheck disable=SC1091
+source ~/.hermes/profiles/feadulta/.env
+: "${FEA_PROD_WPLOAD:=/web/wp-load.php}"
+if [[ "$FEA_PROD_WPLOAD" != "/web/wp-load.php" ]]; then
+  echo "ABORT: FEA_PROD_WPLOAD debe ser /web/wp-load.php (recibido: $FEA_PROD_WPLOAD)" >&2
+  exit 2
+fi
+
+RELEASE_DIR="logs/release-issue-181"
+mkdir -p "$RELEASE_DIR/backups"
+STAMP="$(date -u +%Y%m%dT%H%M%SZ)"
+ORIGINS=(54875 54902 54903 54883 54884 54885 54886 54866 54867 54868 54869 54870 54871 54872 54873 54880 54914)
+AUDIO_IDS=(54875 54902 54903 54883 54884 54885 54886 54866 54867 54868 54869 54870 54871 54872 54873 54880)
+
+if [[ "$MODE" == "apply" && "${FEA_RELEASE_CONFIRM:-}" != "PUBLISH_ISSUE_181" ]]; then
+  echo "ABORT: para escribir en producción usa:" >&2
+  echo "  FEA_RELEASE_CONFIRM=PUBLISH_ISSUE_181 $0 --apply" >&2
+  exit 2
+fi
+
+preflight() {
+  python3 - "${ORIGINS[@]}" <<'PY'
+import json, sys
+sys.path.insert(0, 'scripts')
+import sync_translations_to_prod as sync
+ids = [int(x) for x in sys.argv[1:]]
+rows = []
+for pid in ids:
+    d = json.loads(sync.prod_helper('read_full', str(pid)))
+    rows.append({'id': pid, 'status': d['status'], 'lang': d['lang'], 'translations': d.get('translations', {})})
+assert len(rows) == 17
+assert all(r['status'] == 'publish' and r['lang'] == 'es' for r in rows), rows
+# Los 16 artículos no deben tener traducciones aún; la carta tampoco antes del primer release.
+assert all(r['translations'] == {'es': r['id']} for r in rows), rows
+print(json.dumps(rows, ensure_ascii=False, indent=2))
+print('preflight=PASS sources=17')
+PY
+}
+
+backup_prod_state() {
+  python3 - "$RELEASE_DIR/backups/prod-before-$STAMP.json" "${ORIGINS[@]}" -- "${AUDIO_IDS[@]}" <<'PY'
+import json, sys
+out = sys.argv[1]
+sep = sys.argv.index('--')
+origins = [int(x) for x in sys.argv[2:sep]]
+audio_ids = [int(x) for x in sys.argv[sep+1:]]
+sys.path.insert(0, 'scripts')
+import sync_translations_to_prod as translations
+import sync_audio_to_prod as audio
+payload = {
+  'origins': {str(pid): json.loads(translations.prod_helper('read_full', str(pid))) for pid in origins},
+  'audio_before': {
+    str(pid): {
+      'url': audio.prod_helper('getmeta', str(pid), 'fea_audio_url').strip(),
+      'voice': audio.prod_helper('getmeta', str(pid), 'fea_audio_voice').strip(),
+      'done': audio.prod_helper('getmeta', str(pid), 'fea_audio_done').strip(),
+    } for pid in audio_ids
+  },
+}
+with open(out, 'w', encoding='utf-8') as fh:
+    json.dump(payload, fh, ensure_ascii=False, indent=2)
+print(out)
+PY
+}
+
+verify_release() {
+  python3 - "${ORIGINS[@]}" -- "${AUDIO_IDS[@]}" <<'PY'
+import json, sys, time
+sep = sys.argv.index('--')
+origins = [int(x) for x in sys.argv[1:sep]]
+audio_ids = [int(x) for x in sys.argv[sep+1:]]
+sys.path.insert(0, 'scripts')
+import sync_translations_to_prod as translations
+import sync_audio_to_prod as audio
+
+def read_post(pid):
+    """Lectura server-side resiliente ante una respuesta SSH/PHP vacía transitoria."""
+    last = ''
+    for attempt in range(1, 4):
+        raw = translations.prod_helper('read_full', str(pid)).strip()
+        if raw:
+            try:
+                return json.loads(raw)
+            except json.JSONDecodeError as exc:
+                last = f'JSON inválido intento {attempt}: {exc}; prefijo={raw[:160]!r}'
+        else:
+            last = f'respuesta vacía intento {attempt}'
+        time.sleep(attempt)
+    raise AssertionError(f'No se pudo leer post #{pid}: {last}')
+
+for pid in origins:
+    d = read_post(pid)
+    group = d.get('translations', {})
+    assert set(group) == {'es', 'en', 'fr', 'it', 'pt'}, (pid, group)
+    for lang, tid in group.items():
+        td = read_post(tid)
+        assert td['lang'] == lang and td['status'] == 'publish', (pid, lang, tid, td['lang'], td['status'])
+for pid in audio_ids:
+    url = audio.prod_helper('getmeta', str(pid), 'fea_audio_url').strip()
+    done = audio.prod_helper('getmeta', str(pid), 'fea_audio_done').strip()
+    assert url.endswith(f'/wp-content/uploads/tts/{pid}.mp3') and done == '1', (pid, url, done)
+print('verification=PASS groups=17 translations=68 audio=16')
+PY
+}
+
+if [[ "$MODE" == "verify" ]]; then
+  echo "== Verificación server-side posterior (solo lectura) =="
+  verify_release | tee "$RELEASE_DIR/verification-$STAMP.txt"
+  echo "VERIFICACIÓN ISSUE #181 COMPLETADA"
+  exit 0
+fi
+
+echo "== Preflight server-side ($MODE): 17 fuentes ES =="
+preflight | tee "$RELEASE_DIR/preflight-$MODE-$STAMP.json"
+
+if [[ "$MODE" == "dry-run" ]]; then
+  : > "$RELEASE_DIR/translations-dry-run-$STAMP.log"
+  for origin in "${ORIGINS[@]}"; do
+    FEA_SYNC_STATUS=publish \
+    FEA_SYNC_LOG="$RELEASE_DIR/translations-dry-run-$STAMP.log" \
+    FEA_SYNC_STATE="$RELEASE_DIR/translations-$origin-state.json" \
+    python3 scripts/sync_translations_to_prod.py --origin "$origin" --dry-run
+  done
+  FEA_AUDIO_SYNC_LOG="$RELEASE_DIR/audio-dry-run-$STAMP.log" \
+  FEA_AUDIO_SYNC_STATE="$RELEASE_DIR/audio-state.json" \
+  python3 scripts/sync_audio_to_prod.py --ids "$(IFS=,; echo "${AUDIO_IDS[*]}")" --dry-run | tee "$RELEASE_DIR/audio-dry-run-$STAMP.log"
+  echo "DRY-RUN terminado: 68 traducciones publicables + 16 audios planificados."
+  exit 0
+fi
+
+echo "== Backup server-side previo =="
+backup_prod_state | tee "$RELEASE_DIR/backup-path-$STAMP.txt"
+
+echo "== Subiendo 68 traducciones como publish =="
+for origin in "${ORIGINS[@]}"; do
+  FEA_SYNC_STATUS=publish \
+  FEA_SYNC_LOG="$RELEASE_DIR/translations-apply-$STAMP.log" \
+  FEA_SYNC_STATE="$RELEASE_DIR/translations-$origin-state.json" \
+  python3 scripts/sync_translations_to_prod.py --origin "$origin"
+done
+
+echo "== Subiendo 16 audios =="
+FEA_AUDIO_SYNC_LOG="$RELEASE_DIR/audio-apply-$STAMP.log" \
+FEA_AUDIO_SYNC_STATE="$RELEASE_DIR/audio-state.json" \
+python3 scripts/sync_audio_to_prod.py --ids "$(IFS=,; echo "${AUDIO_IDS[*]}")"
+
+echo "== Verificación server-side posterior =="
+verify_release | tee "$RELEASE_DIR/verification-$STAMP.txt"
+echo "RELEASE ISSUE #181 COMPLETADO"

scripts/sync_audio_to_prod.py

@@ -3,9 +3,10 @@
 sync_audio_to_prod.py — Sube a PROD los mp3 de TTS ya generados/enlazados en
 local (fea_audio_done=1) y fija el meta fea_audio_url en prod.
 
-Prod (134.0.10.170) tiene glibc rota: scp/sftp NO funcionan (connection closed).
+Prod vive en Hetzner (Coolify/Docker) desde el cutover de agosto 2026. Subida
-Workaround: subir el binario por stdin de ssh ("cat > ruta"), igual que el resto
+del binario por stdin de ssh ("cat > ruta" dentro del contenedor vía
-de scripts que tocan ese servidor (ver feadulta-server-glibc-rota.md).
+`docker exec -i`), sin depender de scp/sftp — mismo patrón que
+sync_translations_to_prod.py y sync_carta_from_prod.py.
 
 Uso:
   python3 sync_audio_to_prod.py --carta 54254            # sincroniza toda la cola de la carta
@@ -19,8 +20,10 @@ Rollback (despublica en prod lo que este script publicó):
 from __future__ import annotations
 
 import argparse
+import base64
 import json
 import os
+import shlex
 import subprocess
 import time
 from pathlib import Path
@@ -31,13 +34,20 @@ DB_NAME = os.environ.get("FEA_DB_NAME", "wordpress_db")
 DB_USER = os.environ.get("FEA_DB_USER", "wordpress_user")
 DB_PASS = os.environ.get("FEA_DB_PASS", "wordpress_pass")
 
-PROD_HOST = os.environ.get("FEA_PROD_HOST", "feadulta@134.0.10.170")
+PROD_HOST = os.environ.get("FEA_PROD_SSH_HOST", "")
-PROD_PASS = os.environ.get("FEA_PROD_PASS", "C6c2A!mAl3Wj.BQF")
+PROD_PASS = os.environ.get("FEA_PROD_SSH_PASS", "")
-PROD_WPLOAD = os.environ.get("FEA_PROD_WPLOAD", "/web/wp-load.php")
+PROD_WPLOAD = os.environ.get("FEA_PROD_WPLOAD", "/var/www/html/wp-load.php")
-PROD_HELPER = "/tmp/fea_post_io.php"
+# Desde el cutover a Hetzner, WordPress vive dentro de Coolify/Docker. Si se
-PROD_UPLOADS_TTS = "/web/wp-content/uploads/tts"
+# define, el helper y la subida de mp3 se ejecutan dentro del contenedor.
+# Espejo de sync_translations_to_prod.py / sync_carta_from_prod.py.
+PROD_DOCKER_CONTAINER = os.environ.get("FEA_PROD_DOCKER_CONTAINER", "")
+PROD_UPLOADS_TTS = os.environ.get("FEA_PROD_UPLOADS_TTS", "/var/www/html/wp-content/uploads/tts")
 
 HELPER_SRC = Path(__file__).resolve().parent / "fea_post_io.php"
+# Se evalúa en memoria con `php -r`: no se copia un helper temporal a prod.
+HELPER_EVAL_CODE = "eval(base64_decode(" + repr(
+    base64.b64encode(HELPER_SRC.read_text(encoding="utf-8").removeprefix("<?php").encode("utf-8")).decode("ascii")
+) + "));"
 LOCAL_TTS_DIR = Path(__file__).resolve().parent.parent / "wordpress/wp-content/uploads/tts"
 
 LOG_FILE = Path(os.environ.get(
@@ -74,6 +84,17 @@ def local_meta(post_id: int, key: str) -> str:
     return r.stdout.strip()
 
 
+def prod_id_for_local(local_id: int) -> int:
+    """Resolve a remapped local post to its original production ID.
+
+    ``sync_carta_from_prod.py --remap-conflicts`` keeps this mapping in
+    ``fea_prod_source_id``. Audio files remain named with the local ID, but
+    remote paths and production post meta must use the source ID.
+    """
+    source_id = local_meta(local_id, "fea_prod_source_id")
+    return int(source_id) if source_id.isdigit() else local_id
+
+
 def carta_article_ids(carta_id: int) -> list[int]:
     q = ("SELECT post_id FROM wp_postmeta "
          f"WHERE meta_key='_carta_id' AND meta_value='{carta_id}' ORDER BY post_id;")
@@ -85,47 +106,80 @@ def carta_article_ids(carta_id: int) -> list[int]:
     return [int(x) for x in r.stdout.split() if x.isdigit()]
 
 
-# ── Prod (glibc rota: nada de scp/sftp, todo por ssh + cat) ────────────────────
+# ── Prod (Hetzner/Docker: todo por ssh, mp3 por stdin de "cat") ────────────────
 def _ssh_text(remote_cmd: str, *, stdin: str | None = None, timeout: int = 120) -> str:
+    # El Hetzner nuevo usa auth por clave (ed25519 claude-code@feadulta); sshpass
+    # solo se antepone si hay contraseña configurada (servidor viejo CDMON).
+    if PROD_PASS:
         cmd = ["sshpass", "-p", PROD_PASS, "ssh", "-o", "StrictHostKeyChecking=accept-new",
                "-o", "ConnectTimeout=20", PROD_HOST, remote_cmd]
+    else:
+        cmd = ["ssh", "-o", "StrictHostKeyChecking=accept-new",
+               "-o", "ConnectTimeout=20", PROD_HOST, remote_cmd]
     r = subprocess.run(cmd, input=stdin, capture_output=True, text=True, timeout=timeout)
     if r.returncode != 0:
         raise RuntimeError(f"ssh falló ({r.returncode}): {remote_cmd[:80]}…\n{r.stderr.strip()[:400]}")
     return r.stdout
 
 
+def _remote_wrap(inner_cmd: str) -> str:
+    """Envuelve un comando para que corra dentro del contenedor Docker de prod
+    si FEA_PROD_DOCKER_CONTAINER está definido; si no, corre en el host tal cual.
+
+    Importante: el wrapping (incluidas redirecciones como `< ruta`) debe quedar
+    DENTRO de la shell del contenedor (`sh -c '...'`), porque `docker exec
+    CONTENEDOR cmd < ruta` resuelve esa redirección en el filesystem del HOST,
+    no dentro del contenedor.
+    """
+    if PROD_DOCKER_CONTAINER:
+        return f"docker exec -i {shlex.quote(PROD_DOCKER_CONTAINER)} sh -c {shlex.quote(inner_cmd)}"
+    return inner_cmd
+
+
 def _ssh_upload_bytes(data: bytes, remote_path: str, *, timeout: int = 180) -> None:
+    remote_cmd = _remote_wrap(f"cat > {remote_path}")
+    if PROD_PASS:
         cmd = ["sshpass", "-p", PROD_PASS, "ssh", "-o", "StrictHostKeyChecking=accept-new",
-           "-o", "ConnectTimeout=20", PROD_HOST, f"cat > {remote_path}"]
+               "-o", "ConnectTimeout=20", PROD_HOST, remote_cmd]
+    else:
+        cmd = ["ssh", "-o", "StrictHostKeyChecking=accept-new",
+               "-o", "ConnectTimeout=20", PROD_HOST, remote_cmd]
     sh(cmd, input_bytes=data, timeout=timeout)
 
 
-_prod_helper_ready = False
-
-
 def prod_helper(subcmd: str, *args: str) -> str:
-    global _prod_helper_ready
+    """Ejecuta el helper PHP en memoria; no deja ficheros temporales en prod."""
-    if not _prod_helper_ready:
+    argv = " ".join(shlex.quote(x) for x in (subcmd, *args))
-        _ssh_upload_bytes(HELPER_SRC.read_bytes(), PROD_HELPER)
+    inner = (
-        _prod_helper_ready = True
+        f"FEA_WP_LOAD={shlex.quote(PROD_WPLOAD)} "
-    inner = f"FEA_WP_LOAD={PROD_WPLOAD} php {PROD_HELPER} {subcmd} " + " ".join(args)
+        f"php -r {shlex.quote(HELPER_EVAL_CODE)} {argv}"
-    return _ssh_text(inner, timeout=60)
+    )
+    if PROD_DOCKER_CONTAINER:
+        remote_cmd = (
+            f"docker exec -i -e FEA_WP_LOAD={shlex.quote(PROD_WPLOAD)} "
+            f"{shlex.quote(PROD_DOCKER_CONTAINER)} php -r {shlex.quote(HELPER_EVAL_CODE)} {argv}"
+        )
+    else:
+        remote_cmd = inner
+    return _ssh_text(remote_cmd, timeout=60)
 
 
-def prod_upload_mp3(post_id: int) -> None:
+def prod_upload_mp3(local_id: int, prod_id: int) -> None:
-    src = LOCAL_TTS_DIR / f"{post_id}.mp3"
+    src = LOCAL_TTS_DIR / f"{local_id}.mp3"
     data = src.read_bytes()
-    remote_path = f"{PROD_UPLOADS_TTS}/{post_id}.mp3"
+    remote_path = f"{PROD_UPLOADS_TTS}/{prod_id}.mp3"
     _ssh_upload_bytes(data, remote_path)
-    # Verificación de tamaño (glibc rota => sin fiarse ciegamente del rc=0 de ssh)
+    # Verificación de tamaño: no fiarse ciegamente del rc=0 de ssh. `wc -c` debe
-    remote_size = int(_ssh_text(f"wc -c < {remote_path}").strip())
+    # correr (y resolver la redirección) DENTRO del contenedor — ver _remote_wrap.
+    remote_size = int(_ssh_text(_remote_wrap(f"wc -c < {remote_path}")).strip())
     if remote_size != len(data):
-        raise RuntimeError(f"tamaño no coincide tras subir #{post_id}: local={len(data)} remoto={remote_size}")
+        raise RuntimeError(
+            f"tamaño no coincide tras subir prod#{prod_id}: local={len(data)} remoto={remote_size}"
+        )
 
 
 def prod_remove_mp3(post_id: int) -> None:
-    _ssh_text(f"rm -f {PROD_UPLOADS_TTS}/{post_id}.mp3")
+    _ssh_text(_remote_wrap(f"rm -f {PROD_UPLOADS_TTS}/{post_id}.mp3"))
 
 
 # ── Estado ───────────────────────────────────────────────────────────────────
@@ -149,25 +203,27 @@ def sync_one(post_id: int, state: dict, *, dry_run: bool) -> str:
         return "sin-audio-local"
     if not (LOCAL_TTS_DIR / f"{post_id}.mp3").exists():
         return "mp3-local-ausente"
+    prod_id = prod_id_for_local(post_id)
     if dry_run:
-        return "PLAN: subiría mp3 + setaudio"
+        return f"PLAN: local#{post_id}.mp3 → prod#{prod_id}.mp3 + setaudio"
 
     voice = local_meta(post_id, "fea_audio_voice") or "NicoFeadulta2026"
-    prod_upload_mp3(post_id)
+    prod_upload_mp3(post_id, prod_id)
-    prod_helper("setaudio", str(post_id), f"/wp-content/uploads/tts/{post_id}.mp3", voice)
+    prod_helper("setaudio", str(prod_id), f"/wp-content/uploads/tts/{prod_id}.mp3", voice)
-    if post_id not in state["synced"]:
+    if prod_id not in state["synced"]:
-        state["synced"].append(post_id)
+        state["synced"].append(prod_id)
         save_state(state)
-    return "ok"
+    return f"ok prod#{prod_id}"
 
 
 def rollback_one(post_id: int, state: dict) -> str:
-    prod_helper("unsetaudio", str(post_id))
+    prod_id = prod_id_for_local(post_id)
-    prod_remove_mp3(post_id)
+    prod_helper("unsetaudio", str(prod_id))
-    if post_id in state["synced"]:
+    prod_remove_mp3(prod_id)
-        state["synced"].remove(post_id)
+    if prod_id in state["synced"]:
+        state["synced"].remove(prod_id)
         save_state(state)
-    return "ok"
+    return f"ok prod#{prod_id}"
 
 
 def main() -> int:
@@ -194,7 +250,7 @@ def main() -> int:
     for pid in ids:
         try:
             if args.rollback:
-                if pid not in state["synced"] and not args.ids:
+                if prod_id_for_local(pid) not in state["synced"] and not args.ids:
                     res = "no-estaba-sincronizado"
                     skip += 1
                 else:

scripts/sync_carta_from_prod.py

@@ -23,6 +23,8 @@ el mismo fea_translate_helper.php sin tocar su lógica, solo invertido
 Uso:
     python3 sync_carta_from_prod.py --carta 54495 --dry-run
     python3 sync_carta_from_prod.py --carta 54495
+    python3 sync_carta_from_prod.py --ids 54875,54902,54903 --dry-run
+    python3 sync_carta_from_prod.py --ids 54875,54902,54903
 
 Tras esto, el resto del ciclo ya existente no cambia:
     translate_post.py --carta 54495 --langs en,fr,it,pt --status draft
@@ -34,8 +36,10 @@ Tras esto, el resto del ciclo ya existente no cambia:
 from __future__ import annotations
 
 import argparse
+import base64
 import json
 import os
+import shlex
 import subprocess
 import time
 from pathlib import Path
@@ -44,7 +48,11 @@ WP_CONTAINER = os.environ.get("FEA_WP_CONTAINER", "wordpress-web")
 
 PROD_HOST = os.environ.get("FEA_PROD_SSH_HOST", "")
 PROD_PASS = os.environ.get("FEA_PROD_SSH_PASS", "")
-PROD_WPLOAD = os.environ.get("FEA_PROD_WPLOAD", "/web/wp-load.php")
+PROD_WPLOAD = os.environ.get("FEA_PROD_WPLOAD", "/var/www/html/wp-load.php")
+# Desde el cutover a Hetzner, WordPress vive dentro de Coolify/Docker. Si se
+# define, el helper se ejecuta en memoria dentro del contenedor; no se escribe
+# ningún fichero temporal en prod.
+PROD_DOCKER_CONTAINER = os.environ.get("FEA_PROD_DOCKER_CONTAINER", "")
 PROD_HELPER = "/tmp/fea_translate_helper.php"
 
 HELPER_SRC = Path(__file__).resolve().parent / "fea_translate_helper.php"
@@ -74,22 +82,35 @@ def sh(cmd: list[str], *, stdin: str | None = None, timeout: int = 120) -> str:
 
 
 # ── Prod (origen, solo lectura) ─────────────────────────────────────────────
-_prod_ready = False
-
-
 def _ssh(remote_cmd: str, *, stdin: str | None = None, timeout: int = 120) -> str:
+    # El Hetzner nuevo usa auth por clave (ed25519 claude-code@feadulta); sshpass
+    # solo se antepone si hay contraseña configurada (servidor viejo CDMON).
+    if PROD_PASS:
         cmd = ["sshpass", "-p", PROD_PASS, "ssh", "-o", "StrictHostKeyChecking=accept-new",
                "-o", "ConnectTimeout=20", PROD_HOST, remote_cmd]
+    else:
+        cmd = ["ssh", "-o", "StrictHostKeyChecking=accept-new",
+               "-o", "ConnectTimeout=20", PROD_HOST, remote_cmd]
     return sh(cmd, stdin=stdin, timeout=timeout)
 
 
-def prod_helper(subcmd: str, *args: str, stdin: str | None = None) -> str:
+def prod_helper(subcmd: str, *args: str) -> str:
-    global _prod_ready
+    """Run the helper in prod memory; never upload a temporary file to prod."""
-    if not _prod_ready:
+    helper_php = HELPER_SRC.read_text(encoding="utf-8").replace("<?php", "", 1)
-        _ssh(f"cat > {PROD_HELPER}", stdin=HELPER_SRC.read_text(encoding="utf-8"))
+    encoded = base64.b64encode(helper_php.encode("utf-8")).decode("ascii")
-        _prod_ready = True
+    code = f"eval(base64_decode('{encoded}'));"
-    inner = f"FEA_WP_LOAD={PROD_WPLOAD} php {PROD_HELPER} {subcmd} " + " ".join(args)
+    if PROD_DOCKER_CONTAINER:
-    return _ssh(inner, stdin=stdin, timeout=180)
+        remote = (
+            f"docker exec -i -e FEA_WP_LOAD={shlex.quote(PROD_WPLOAD)} "
+            f"{shlex.quote(PROD_DOCKER_CONTAINER)} php -r {shlex.quote(code)} -- "
+            + " ".join(shlex.quote(part) for part in (subcmd, *args))
+        )
+    else:
+        remote = (
+            f"FEA_WP_LOAD={shlex.quote(PROD_WPLOAD)} php -r {shlex.quote(code)} -- "
+            + " ".join(shlex.quote(part) for part in (subcmd, *args))
+        )
+    return _ssh(remote, timeout=180)
 
 
 def prod_read_full(post_id: int) -> dict:
@@ -110,6 +131,17 @@ def seed_ids_from_carta(carta_id: int) -> list[int]:
     return sorted(ids)
 
 
+def seed_ids_from_csv(raw_ids: str) -> list[int]:
+    """Parse a deliberate prod→local subset for Mixbot --parte1 (#181)."""
+    try:
+        ids = sorted({int(part.strip()) for part in raw_ids.split(",") if part.strip()})
+    except ValueError as exc:
+        raise ValueError("--ids debe ser una lista CSV de IDs numéricos") from exc
+    if not ids or any(pid <= 0 for pid in ids):
+        raise ValueError("--ids debe contener al menos un ID positivo")
+    return ids
+
+
 def collect_related_posts(seed_ids: list[int]) -> tuple[dict[int, dict], list[dict[str, int]]]:
     posts: dict[int, dict] = {}
     groups: dict[tuple[tuple[str, int], ...], dict[str, int]] = {}
@@ -153,7 +185,11 @@ def local_read_safe(post_id: int) -> dict | None:
         return None
 
 
-def local_clone(post: dict) -> int:
+def local_clone(post: dict, *, preserve_id: bool = True) -> int:
+    # Conserva el ID remoto como trazabilidad incluso cuando haya que crear un
+    # ID local nuevo por una colisión entre entornos.
+    meta = dict(post.get("meta", {}))
+    meta["fea_prod_source_id"] = [str(post["id"])]
     payload = {
         "title": post["title"],
         "content": post.get("content", ""),
@@ -166,10 +202,14 @@ def local_clone(post: dict) -> int:
         "status": post.get("status"),
         "cats": post.get("cats", []),
         "cat_slugs": post.get("cat_slugs", []),
-        "meta": post.get("meta", {}),
+        "meta": meta,
     }
+    if preserve_id:
         out = local_helper("clone", str(post["id"]), post.get("lang") or "es", STATUS,
                            stdin=json.dumps(payload)).strip()
+    else:
+        out = local_helper("clone_new", post.get("lang") or "es", STATUS,
+                           stdin=json.dumps(payload)).strip()
     return int(out)
 
 
@@ -179,13 +219,9 @@ def local_save_group(group: dict[str, int]) -> dict[str, int]:
 
 
 # ── Main ─────────────────────────────────────────────────────────────────────
-def sync_carta(carta_id: int, *, dry_run: bool) -> int:
+def sync_posts(seed_ids: list[int], *, dry_run: bool, source_label: str,
-    seed_ids = seed_ids_from_carta(carta_id)
+               remap_conflicts: bool = False) -> int:
-    log(f"Cluster descubierto desde fea_parse_carta_sections(prod#{carta_id}): "
+    log(f"Lote {source_label}: {len(seed_ids)} post(s) -> {seed_ids}")
-        f"{len(seed_ids)} post(s) -> {seed_ids}")
-    if not seed_ids or seed_ids == [carta_id]:
-        log("  ⚠️ El parser no resolvió ningún artículo enlazado (¿carta sin publicar aún, "
-            "o secciones sin encabezados reconocibles?). Revisa antes de continuar.")
 
     posts, groups = collect_related_posts(seed_ids)
 
@@ -196,51 +232,82 @@ def sync_carta(carta_id: int, *, dry_run: bool) -> int:
             conflicts.append((pid, existing.get("title", ""), p.get("title", "")))
 
     if conflicts:
-        log(f"  ⚠️ {len(conflicts)} CONFLICTO(S): el ID ya existe en local con OTRO contenido "
+        log(f"  ⚠️ {len(conflicts)} CONFLICTO(S): el ID ya existe en local con OTRO contenido:")
-            f"y va a ser SOBRESCRITO:")
         for pid, old_title, new_title in conflicts:
-            log(f"      #{pid}: local actual «{old_title[:50]}»  ->  prod «{new_title[:50]}»")
+            log(f"      #{pid}: local actual «{old_title[:50]}»  |  prod «{new_title[:50]}»")
+        if not dry_run and not remap_conflicts:
+            raise RuntimeError(
+                "Importación cancelada: usar --remap-conflicts para crear IDs locales nuevos; "
+                "nunca se sobrescriben posts locales por un choque de IDs entre entornos."
+            )
 
     if dry_run:
         for pid, p in posts.items():
-            log(f"  PULL prod#{pid} [{p.get('lang','?')}] status={p.get('status')} "
+            action = "REMAPPING to new local ID" if pid in {c[0] for c in conflicts} else "PULL"
+            log(f"  {action} prod#{pid} [{p.get('lang','?')}] status={p.get('status')} "
                 f"slug={p.get('slug','')} «{p.get('title','')[:50]}»")
         for group in groups:
             log(f"  GROUP {group}")
         log("DRY-RUN: nada escrito en local.")
         return 0
 
+    conflict_ids = {pid for pid, _, _ in conflicts}
+    id_map: dict[int, int] = {}
     for pid, p in posts.items():
-        new_id = local_clone(p)
+        new_id = local_clone(p, preserve_id=pid not in conflict_ids)
+        id_map[pid] = new_id
         if new_id != pid:
-            log(f"  ⚠️ prod#{pid} se clonó como local#{new_id} — ID NO preservado, revisar a mano.")
+            log(f"  remap prod#{pid} -> local#{new_id} [{p.get('lang','?')}] «{p['title'][:45]}»")
         else:
             log(f"  clone prod#{pid} -> local#{new_id} [{p.get('lang','?')}] «{p['title'][:45]}»")
 
     for group in groups:
-        if len(group) < 2:
+        local_group = {lang: id_map.get(pid, pid) for lang, pid in group.items()}
+        if len(local_group) < 2:
             continue
-        saved = local_save_group(group)
+        saved = local_save_group(local_group)
         log(f"  group enlazado en local {saved}")
 
-    log(f"FIN sync prod->local. carta={carta_id} posts={len(posts)} conflictos_previos={len(conflicts)}")
+    log(f"FIN sync prod->local. fuente={source_label} posts={len(posts)} conflictos_previos={len(conflicts)}")
     return 0
 
 
+def sync_carta(carta_id: int, *, dry_run: bool, remap_conflicts: bool = False) -> int:
+    seed_ids = seed_ids_from_carta(carta_id)
+    if not seed_ids or seed_ids == [carta_id]:
+        log("  ⚠️ El parser no resolvió ningún artículo enlazado (¿carta sin publicar aún, "
+            "o secciones sin encabezados reconocibles?). Revisa antes de continuar.")
+    return sync_posts(seed_ids, dry_run=dry_run, source_label=f"carta prod#{carta_id}",
+                      remap_conflicts=remap_conflicts)
+
+
+def sync_ids(raw_ids: str, *, dry_run: bool, remap_conflicts: bool = False) -> int:
+    return sync_posts(seed_ids_from_csv(raw_ids), dry_run=dry_run,
+                      source_label="IDs explícitos de Mixbot --parte1",
+                      remap_conflicts=remap_conflicts)
+
+
 def main() -> int:
     ap = argparse.ArgumentParser(
         description="Copia una carta (y su cluster de artículos) de PROD a LOCAL preservando IDs.")
-    ap.add_argument("--carta", type=int, required=True, help="ID del post ES de la carta en PROD.")
+    group = ap.add_mutually_exclusive_group(required=True)
+    group.add_argument("--carta", type=int, help="ID del post ES de la carta en PROD.")
+    group.add_argument("--ids", help="Lista CSV explícita de posts ES para Mixbot --parte1 (#181).")
     ap.add_argument("--dry-run", action="store_true", help="Solo muestra el plan; no escribe en local.")
+    ap.add_argument("--remap-conflicts", action="store_true",
+                    help="Ante IDs ocupados localmente, crea IDs nuevos y guarda fea_prod_source_id; nunca sobrescribe.")
     args = ap.parse_args()
 
-    if not PROD_HOST or not PROD_PASS:
+    if not PROD_HOST:
         raise SystemExit(
-            "Faltan FEA_PROD_SSH_HOST / FEA_PROD_SSH_PASS en el entorno.\n"
+            "Falta FEA_PROD_SSH_HOST en el entorno (FEA_PROD_SSH_PASS es opcional: "
+            "el Hetzner nuevo usa auth por clave).\n"
             "Antes de ejecutar: source ~/.hermes/profiles/feadulta/.env"
         )
 
-    return sync_carta(args.carta, dry_run=args.dry_run)
+    if args.carta:
+        return sync_carta(args.carta, dry_run=args.dry_run, remap_conflicts=args.remap_conflicts)
+    return sync_ids(args.ids, dry_run=args.dry_run, remap_conflicts=args.remap_conflicts)
 
 
 if __name__ == "__main__":

scripts/sync_translations_to_prod.py

@@ -16,8 +16,10 @@ coincidencia local↔prod cuando prod va por detrás.
 from __future__ import annotations
 
 import argparse
+import base64
 import json
 import os
+import shlex
 import subprocess
 import time
 from pathlib import Path
@@ -29,12 +31,20 @@ DB_NAME = os.environ.get("FEA_DB_NAME", "wordpress_db")
 DB_USER = os.environ.get("FEA_DB_USER", "wordpress_user")
 DB_PASS = os.environ.get("FEA_DB_PASS", "wordpress_pass")
 
-PROD_HOST = os.environ.get("FEA_PROD_HOST", "feadulta@134.0.10.170")
+PROD_HOST = os.environ.get("FEA_PROD_SSH_HOST", "")
-PROD_PASS = os.environ.get("FEA_PROD_PASS", "C6c2A!mAl3Wj.BQF")
+PROD_PASS = os.environ.get("FEA_PROD_SSH_PASS", "")
-PROD_WPLOAD = os.environ.get("FEA_PROD_WPLOAD", "/web/wp-load.php")
+PROD_WPLOAD = os.environ.get("FEA_PROD_WPLOAD", "/var/www/html/wp-load.php")
-PROD_HELPER = "/tmp/fea_translate_helper.php"
+# Desde el cutover a Hetzner, WordPress vive dentro de Coolify/Docker. Si se
+# define, el helper se ejecuta en memoria dentro del contenedor; no se escribe
+# ningún fichero temporal en prod. Espejo de sync_carta_from_prod.py.
+PROD_DOCKER_CONTAINER = os.environ.get("FEA_PROD_DOCKER_CONTAINER", "")
 
 HELPER_SRC = Path(__file__).resolve().parent / "fea_translate_helper.php"
+# El helper se ejecuta en memoria con `php -r` en prod. Así no se copia ningún
+# fichero a /tmp remoto y stdin queda disponible para el payload JSON.
+HELPER_EVAL_CODE = "eval(base64_decode(" + repr(
+    base64.b64encode(HELPER_SRC.read_text(encoding="utf-8").removeprefix("<?php").encode("utf-8")).decode("ascii")
+) + "));"
 LOCAL_HELPER_DST = "/tmp/fea_translate_helper.php"
 STATE_FILE = Path(os.environ.get("FEA_SYNC_STATE", "/tmp/feadulta-sync-state.json"))
 LOG_FILE = Path(os.environ.get("FEA_SYNC_LOG", "/tmp/feadulta-sync.log"))
@@ -107,8 +117,20 @@ def local_read_full(post_id: int) -> dict:
 
 
 def local_translation_pairs() -> list[tuple[int, int]]:
-    q = ("SELECT post_id, meta_value FROM wp_postmeta "
+    """Devuelve (traducción_local, origen_ES_en_prod).
-         "WHERE meta_key='traduccion_origen' ORDER BY CAST(meta_value AS UNSIGNED), post_id;")
+
+    Normalmente ambos entornos compartían IDs. Para importaciones remapeadas,
+    la carta ES local guarda `fea_prod_source_id`; esa referencia remota tiene
+    prioridad y evita enlazar una traducción con un post distinto de prod.
+    """
+    q = (
+        "SELECT tr.post_id, COALESCE(NULLIF(src.meta_value,''), tr.meta_value) "
+        "FROM wp_postmeta tr "
+        "LEFT JOIN wp_postmeta src ON src.post_id=CAST(tr.meta_value AS UNSIGNED) "
+        "AND src.meta_key='fea_prod_source_id' "
+        "WHERE tr.meta_key='traduccion_origen' "
+        "ORDER BY CAST(COALESCE(NULLIF(src.meta_value,''), tr.meta_value) AS UNSIGNED), tr.post_id;"
+    )
     out = sh(["docker", "exec", DB_CONTAINER, "mysql", f"-u{DB_USER}", f"-p{DB_PASS}",
               DB_NAME, "-N", "-e", q])
     pairs = []
@@ -155,21 +177,31 @@ def collect_related_posts(seed_ids: list[int]) -> tuple[dict[int, dict], list[di
 
 
 # ── Prod ─────────────────────────────────────────────────────────────────────
-_prod_ready = False
-
-
 def _ssh(remote_cmd: str, *, stdin: str | None = None, timeout: int = 120) -> str:
+    # El Hetzner nuevo usa auth por clave (ed25519 claude-code@feadulta); sshpass
+    # solo se antepone si hay contraseña configurada (servidor viejo CDMON).
+    if PROD_PASS:
         cmd = ["sshpass", "-p", PROD_PASS, "ssh", "-o", "StrictHostKeyChecking=accept-new",
                "-o", "ConnectTimeout=20", PROD_HOST, remote_cmd]
+    else:
+        cmd = ["ssh", "-o", "StrictHostKeyChecking=accept-new",
+               "-o", "ConnectTimeout=20", PROD_HOST, remote_cmd]
     return sh(cmd, stdin=stdin, timeout=timeout)
 
 
 def prod_helper(subcmd: str, *args: str, stdin: str | None = None) -> str:
-    global _prod_ready
+    """Ejecuta el helper PHP en memoria; no crea ficheros temporales en prod."""
-    if not _prod_ready:
+    argv = " ".join(shlex.quote(x) for x in (subcmd, *args))
-        _ssh(f"cat > {PROD_HELPER}", stdin=HELPER_SRC.read_text(encoding="utf-8"))
+    if PROD_DOCKER_CONTAINER:
-        _prod_ready = True
+        inner = (
-    inner = f"FEA_WP_LOAD={PROD_WPLOAD} php {PROD_HELPER} {subcmd} " + " ".join(args)
+            f"docker exec -i -e FEA_WP_LOAD={shlex.quote(PROD_WPLOAD)} "
+            f"{shlex.quote(PROD_DOCKER_CONTAINER)} php -r {shlex.quote(HELPER_EVAL_CODE)} {argv}"
+        )
+    else:
+        inner = (
+            f"FEA_WP_LOAD={shlex.quote(PROD_WPLOAD)} "
+            f"php -r {shlex.quote(HELPER_EVAL_CODE)} {argv}"
+        )
     return _ssh(inner, stdin=stdin, timeout=180)
 
 
@@ -256,7 +288,12 @@ def deploy_fixed_ids(seed_ids: list[int], *, keep_existing: set[int], dry_run: b
 
 
 # ── Main legado ──────────────────────────────────────────────────────────────
-def legacy_sync(limit: int, origin: int) -> int:
+def legacy_sync(limit: int, origin: int, *, dry_run: bool = False) -> int:
+    """Sincroniza por origen ES+idioma, sin reutilizar IDs locales en prod.
+
+    Este es el modo seguro si producción ha avanzado y sus IDs ya pueden
+    colisionar con traducciones creadas en el entorno local.
+    """
     state = load_state()
     pairs = local_translation_pairs()
     if origin:
@@ -280,6 +317,10 @@ def legacy_sync(limit: int, origin: int) -> int:
         if key in state["done"]:
             n_skip += 1
             continue
+        if dry_run:
+            log(f"  PLAN {key}: crearía traducción [{lang}] «{t['title'][:45]}»")
+            n_ok += 1
+            continue
         try:
             new_id = prod_create(src_origin, lang, t["title"], t["content"])
             state["done"][key] = new_id
@@ -292,8 +333,10 @@ def legacy_sync(limit: int, origin: int) -> int:
             n_err += 1
             log(f"  {key} ERROR: {exc}")
 
+    if not dry_run:
         save_state(state)
-    log(f"FIN sync legado. nuevos={n_ok} saltados={n_skip} errores={n_err}. Estado: {STATE_FILE}")
+    mode = "DRY-RUN" if dry_run else "SYNC"
+    log(f"FIN sync legado {mode}. nuevos/plan={n_ok} saltados={n_skip} errores={n_err}. Estado: {STATE_FILE}")
     log("Recuerda en prod: ejecutar remap_translation_cats.php si alguna quedó sin categoría traducida.")
     return 0
 
@@ -318,7 +361,7 @@ def main() -> int:
         keep_existing = set(parse_csv_ints(args.keep_existing))
         return deploy_fixed_ids(seed_ids, keep_existing=keep_existing, dry_run=args.dry_run)
 
-    return legacy_sync(args.limit, args.origin)
+    return legacy_sync(args.limit, args.origin, dry_run=args.dry_run)
 
 
 if __name__ == "__main__":

scripts/translate_post.py

@@ -271,7 +271,8 @@ def save_state(state: dict) -> None:
 
 
 # ── Orquestación ─────────────────────────────────────────────────────────────
-def process_post(post_id: int, langs: list[str], status: str, force: bool, state: dict) -> None:
+def process_post(post_id: int, langs: list[str], status: str, force: bool, state: dict,
+                 *, dry_run: bool = False) -> None:
     src = read_post(post_id)
     if src.get("lang") and src["lang"] != "es":
         log(f"#{post_id} no es ES (lang={src['lang']}) — saltado")
@@ -286,8 +287,14 @@ def process_post(post_id: int, langs: list[str], status: str, force: bool, state
             state["done"][key] = existing
             continue
         if existing and force:
+            if dry_run:
+                log(f"  {lang}: PLAN regenerar traducción previa #{existing}")
+                continue
             php_helper("unlink", str(post_id), lang)
             log(f"  {lang}: --force, eliminada traducción previa #{existing}")
+        if dry_run:
+            log(f"  {lang}: PLAN traducir con {ENGINE} y crear como {status}")
+            continue
         try:
             t0 = time.time()
             title = translate_text(src["title"], lang, is_title=True)
@@ -312,6 +319,7 @@ def main() -> int:
     ap.add_argument("--langs", default="en,fr,it,pt", help="Idiomas destino separados por coma.")
     ap.add_argument("--status", default="draft", choices=["draft", "publish"], help="Estado de la traducción.")
     ap.add_argument("--force", action="store_true", help="Regenera aunque ya exista la traducción.")
+    ap.add_argument("--dry-run", action="store_true", help="Muestra el plan sin llamar al LLM ni escribir en WordPress/estado.")
     args = ap.parse_args()
 
     langs = [l.strip() for l in args.langs.split(",") if l.strip() in LANG_NAMES]
@@ -329,9 +337,11 @@ def main() -> int:
 
     state = load_state()
     for pid in ids:
-        process_post(pid, langs, args.status, args.force, state)
+        process_post(pid, langs, args.status, args.force, state, dry_run=args.dry_run)
+    if not args.dry_run:
         save_state(state)
-    log(f"FIN. {len(state['done'])} traducciones registradas, "
+    mode = "DRY-RUN" if args.dry_run else "FIN"
+    log(f"{mode}. {len(state['done'])} traducciones registradas, "
         f"{len(state.get('errors', {}))} errores. Estado: {STATE_FILE}")
     return 0
 

scripts/tts_produce.py

@@ -29,7 +29,7 @@ sys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))
 import minimax_tts as mm          # get_post_text, add_pauses, t2a, OUT
 import translate_post as tp       # carta_article_ids
 
-VOICE = "NicoFeadulta2026"
+VOICE = os.environ.get("FEA_TTS_VOICE", "NicoFeadulta2026")
 MODEL = "speech-2.8-hd"
 CONTAINER = "wordpress-web"
 PROD = Path(__file__).resolve().parent.parent / "wordpress/wp-content/uploads/tts"
@@ -103,6 +103,8 @@ def main():
                         help="para tras N audios OK en esta ejecución (0 = sin tope)")
     parser.add_argument("--dry-run", action="store_true",
                         help="imprime la cola y sale, sin sintetizar ni gastar cuota")
+    parser.add_argument("--allow-default-voice", action="store_true",
+                        help="permite Nico para autores sin voz digitalizada (desactivado por defecto)")
     args = parser.parse_args()
     if args.ids:
         os.environ["FEA_TTS_IDS"] = args.ids
@@ -149,7 +151,15 @@ def main():
             i += 1
             continue
 
-        voice = mm.voice_for_author(author, VOICE)
+        cloned_voice = mm.voice_for_author(author, "")
+        if cloned_voice:
+            voice = cloned_voice
+        elif args.allow_default_voice:
+            voice = VOICE
+        else:
+            log(f"#{pid}: autor sin voz digitalizada (author_id={author}); omitido")
+            i += 1
+            continue
         rc = mm.t2a(mm.add_pauses(text), voice, MODEL, f"prod-{pid}")
         if rc == 0:
             src = mm.OUT / f"prod-{pid}.mp3"