feat(export): Volltranskript als Anhang in DOCX/PDF (meeting/phone-call/interview)

LLM komprimiert das Transkript bisher auf wenige Sätze + 20 Excerpt-Einträge
— der Rest fiel aus dem Dokument. Pipeline reicht Whisper-Segmente jetzt an
build_docx/build_pdf durch; Profile mit `docx.include_full_transcript: true`
hängen das komplette Transkript (9 pt) am Ende an.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

mac-worker/app/core/docx_export.py

@@ -19,7 +19,11 @@ from docx.shared import Pt
 from . import profiles
 
 
-def build_docx(data: dict[str, Any], profile_name: str | None = None) -> bytes:
+def build_docx(
+    data: dict[str, Any],
+    profile_name: str | None = None,
+    full_transcript: list[dict] | None = None,
+) -> bytes:
     profile = profiles.get(profile_name)
     cfg = profile.docx or {}
 
@@ -37,6 +41,9 @@ def build_docx(data: dict[str, Any], profile_name: str | None = None) -> bytes:
     for section in cfg.get("sections", []) or []:
         _render_section(doc, data, section)
 
+    if cfg.get("include_full_transcript") and full_transcript:
+        _render_full_transcript(doc, full_transcript)
+
     if data.get("_parse_error") and data.get("_raw"):
         doc.add_heading("Roh-Antwort des Modells", level=1)
         doc.add_paragraph(data["_raw"])
@@ -46,6 +53,23 @@ def build_docx(data: dict[str, Any], profile_name: str | None = None) -> bytes:
     return buf.getvalue()
 
 
+def _render_full_transcript(doc, segments: list[dict]) -> None:
+    """Hängt den Volltranskript-Anhang an: eine Überschrift, dann ein Absatz
+    pro Whisper-Segment in 9 pt (damit das Dokument nicht explodiert).
+    Leere Segmente werden übersprungen.
+    """
+    doc.add_page_break()
+    doc.add_heading("Volltranskript", level=1)
+    for seg in segments:
+        text = (seg.get("text") if isinstance(seg, dict) else str(seg)) or ""
+        text = text.strip()
+        if not text:
+            continue
+        p = doc.add_paragraph()
+        run = p.add_run(text)
+        run.font.size = Pt(9)
+
+
 def _render_section(doc, data: dict, section: dict) -> None:
     val = data.get(section["key"])
     if val in (None, "", [], {}):

mac-worker/app/core/pdf_export.py

@@ -69,7 +69,11 @@ def _para(text: str, style_name: str = "Body2") -> Paragraph:
     return Paragraph(_esc(text).replace("\n", "<br/>"), _STYLES[style_name])
 
 
-def build_pdf(data: dict[str, Any], profile_name: str | None = None) -> bytes:
+def build_pdf(
+    data: dict[str, Any],
+    profile_name: str | None = None,
+    full_transcript: list[dict] | None = None,
+) -> bytes:
     profile = profiles.get(profile_name)
     cfg = profile.docx or {}
 
@@ -108,6 +112,9 @@ def build_pdf(data: dict[str, Any], profile_name: str | None = None) -> bytes:
     for section in cfg.get("sections", []) or []:
         _render_section(story, data, section)
 
+    if cfg.get("include_full_transcript") and full_transcript:
+        _render_full_transcript(story, full_transcript)
+
     # Parse-Error-Fallback
     if data.get("_parse_error") and data.get("_raw"):
         story.append(Paragraph("Roh-Antwort des Modells", _STYLES["SectionHeading"]))
@@ -117,6 +124,27 @@ def build_pdf(data: dict[str, Any], profile_name: str | None = None) -> bytes:
     return buf.getvalue()
 
 
+def _render_full_transcript(story: list, segments: list[dict]) -> None:
+    """Anhang nach allen Sektionen: ein Absatz pro Segment in 9 pt.
+    Vor dem Block ein PageBreak, damit das Volltranskript auf einer
+    eigenen Seite beginnt und das Protokoll nicht zerreißt.
+    """
+    from reportlab.platypus import PageBreak  # local import — selten genutzt
+
+    story.append(PageBreak())
+    story.append(Paragraph("Volltranskript", _STYLES["SectionHeading"]))
+    transcript_style = ParagraphStyle(
+        name="TranscriptLine", parent=_STYLES["Body2"],
+        fontSize=9, leading=12, spaceAfter=3,
+    )
+    for seg in segments:
+        text = (seg.get("text") if isinstance(seg, dict) else str(seg)) or ""
+        text = text.strip()
+        if not text:
+            continue
+        story.append(Paragraph(_esc(text).replace("\n", "<br/>"), transcript_style))
+
+
 def _render_section(story: list, data: dict, section: dict) -> None:
     val = data.get(section["key"])
     if val in (None, "", [], {}):