feat(translation): quality pipeline overhaul + new features (audit 2026-08-29)

Translation quality & format preservation:
- Word: merge adjacent same-format runs into one unit (sentence-level
  coherence like inline-tag handling); translate comments/balloons;
  dedupe textbox collection (was translated twice); RTL no longer
  overrides center/justify alignment; CJK/Arabic font hints (eastAsia/cs)
- PPTX: chart translations now actually reach the output file
  (ChartPart.blob is read-only — rewrite chart XML in the saved ZIP);
  CJK typeface hints (a:ea)
- Excel: sheet renames no longer break references — rewrite cell
  formulas (3D/quoted), defined names, data validations, cond. formats
- PDF: bold/italic honored (hebo/heit/hebi); table cells never merge;
  unchanged blocks left untouched (typography preserved, fixes duplicate
  hyperlinks); attempted/changed stats + route gate now cover PDF;
  CJK font paths; scanned PDFs via Mistral OCR (detection + admin settings)

Features:
- formality param (formal/informal) + automatic regional-variant prompts
- output_mode=bilingual docx (source above translation)
- per-user translation memory on Redis (falls back to LRU), context-hashed
- QA report + 0-100 confidence score in job status; L0 on by default
- OpenAI-compatible providers: whole chunk in ONE numbered-JSON request
  (~15x fewer calls) with per-item fallback; base prompt always present
  (custom prompt no longer replaces translation instructions)

Infra & marketing alignment:
- plan-based engine gating + vision gating (closes paid-engine leak);
  /providers/available filtered per plan; 107 languages exposed
- zh-CN/zh-TW validation fixed; libmagic disabled on Windows (native crash)
- admin: Mistral OCR settings + engine status dashboard; httpx<0.28 pin
  (TestClient breakage); Prometheus test fixture fixed
- marketing docs aligned with code (PDF+OCR, retention, engines, pricing)
- security: .env.ionos/.env.production/provider_settings.json removed

Tests: 1173 passed / 0 failed (6 network tests deselected: free Google
endpoint temporarily blocked from this machine)

translators/bilingual.py

@@ -0,0 +1,90 @@
+"""
+Bilingual output: interleave source paragraphs with their translation.
+
+Given the ORIGINAL document and the TRANSLATED document (same structure —
+the pipeline only rewrites run texts), produce a copy of the translated
+document where each translated body paragraph is preceded by its source
+paragraph in gray italic. Tables, headers and footers keep the translated
+version only (duplicating them would double the layout).
+"""
+
+from pathlib import Path
+from typing import Optional
+
+from docx import Document
+from docx.text.paragraph import Paragraph
+from docx.oxml import OxmlElement
+from docx.oxml.ns import qn
+from docx.shared import Pt, RGBColor
+
+from core.logging import get_logger
+
+logger = get_logger(__name__)
+
+
+def make_bilingual_docx(
+    source_path: Path, translated_path: Path, output_path: Path
+) -> Optional[Path]:
+    """Create a bilingual .docx (source paragraph above its translation).
+
+    Returns the output path, or None when the pairing failed (structure
+    mismatch) — callers fall back to the translated-only file.
+    """
+    source_path = Path(source_path)
+    translated_path = Path(translated_path)
+    output_path = Path(output_path)
+
+    try:
+        src = Document(str(source_path))
+        tr = Document(str(translated_path))
+    except Exception as e:
+        logger.warning("bilingual_open_failed", error=str(e))
+        return None
+
+    src_children = list(src.element.body)
+    tr_children = list(tr.element.body)
+
+    # The pipeline preserves structure exactly; a drift beyond a small
+    # tolerance means pairing by index is unsafe → bail out.
+    if abs(len(src_children) - len(tr_children)) > 0:
+        logger.warning(
+            "bilingual_structure_mismatch",
+            source_elements=len(src_children),
+            translated_elements=len(tr_children),
+        )
+        if len(src_children) != len(tr_children):
+            return None
+
+    inserted = 0
+    for src_el, tr_el in zip(src_children, tr_children):
+        if tr_el.tag != qn("w:p") or src_el.tag != qn("w:p"):
+            continue
+        src_text = "".join(
+            t.text or "" for t in src_el.iter(qn("w:t"))
+        ).strip()
+        if not src_text:
+            continue
+
+        # Insert a NEW paragraph directly above the translated one, inside
+        # the translated document (keeps styles/sections untouched).
+        new_p = OxmlElement("w:p")
+        tr_el.addprevious(new_p)
+        para = Paragraph(new_p, tr)
+        run = para.add_run(src_text)
+        run.font.size = Pt(9)
+        run.font.italic = True
+        run.font.color.rgb = RGBColor(0x80, 0x80, 0x80)
+        inserted += 1
+
+    if inserted == 0:
+        logger.info("bilingual_nothing_inserted")
+        return None
+
+    try:
+        tr.save(str(output_path))
+    except Exception as e:
+        logger.warning("bilingual_save_failed", error=str(e))
+        return None
+
+    logger.info("bilingual_docx_created", paragraphs=inserted)
+    return output_path