Promote develop → main (2026-05-17 00:40 UTC)

requirements.txt

@@ -16,3 +16,8 @@ Jinja2>=3.1
 weasyprint>=68.0
 python-docx>=1.2
 pikepdf>=9.0
+# PDF-Quellen (Ingestion)
+pdfplumber>=0.11
+pytesseract>=0.3
+pdf2image>=1.17
+Pillow>=10.0

scripts/migrate_pdf_source.py

@@ -0,0 +1,34 @@
+"""Idempotente Migration: Quellen-Typ pdf_document + EN-Spalten in articles.
+
+Beim Live-Promote anwenden:
+    python3 scripts/migrate_pdf_source.py /home/claude-dev/osint-data/osint.db
+"""
+import sqlite3
+import sys
+
+
+def add_col(db, table, col_def):
+    name = col_def.split()[0]
+    cols = {r[1] for r in db.execute(f"PRAGMA table_info({table})").fetchall()}
+    if name in cols:
+        return False
+    db.execute(f"ALTER TABLE {table} ADD COLUMN {col_def}")
+    return True
+
+
+def main(path):
+    with sqlite3.connect(path) as db:
+        for col in ("pdf_path TEXT", "pdf_sha256 TEXT", "processed_at TIMESTAMP"):
+            print(f"sources.{col.split()[0]}:", "added" if add_col(db, "sources", col) else "exists")
+        for col in ("headline_en TEXT", "content_en TEXT"):
+            print(f"articles.{col.split()[0]}:", "added" if add_col(db, "articles", col) else "exists")
+        db.execute("CREATE INDEX IF NOT EXISTS idx_sources_pdf_sha256 ON sources(pdf_sha256)")
+        db.commit()
+    print("DONE")
+
+
+if __name__ == "__main__":
+    if len(sys.argv) != 2:
+        print("Usage: migrate_pdf_source.py /path/to/osint.db")
+        sys.exit(1)
+    main(sys.argv[1])

src/main.py

@@ -298,6 +298,8 @@ async def lifespan(app: FastAPI):
     orchestrator.set_ws_manager(ws_manager)
     await orchestrator.start()
 
+    from services import pdf_ingest as _pdf_ingest
+    scheduler.add_job(_pdf_ingest.run_once, "interval", minutes=1, id="pdf_ingest", max_instances=1, coalesce=True)
     scheduler.add_job(check_auto_refresh, "interval", minutes=1, id="auto_refresh")
     scheduler.add_job(cleanup_expired, "interval", hours=1, id="cleanup")
     scheduler.add_job(daily_source_health_check, "cron", hour=4, minute=0, id="source_health")

src/models.py

@@ -140,7 +140,7 @@ class IncidentListItem(BaseModel):
 
 
 # Sources (Quellenverwaltung)
-SOURCE_TYPE_PATTERN = "^(rss_feed|web_source|excluded|telegram_channel|podcast_feed)$"
+SOURCE_TYPE_PATTERN = "^(rss_feed|web_source|excluded|telegram_channel|podcast_feed|pdf_document)$"
 SOURCE_CATEGORY_PATTERN = "^(nachrichtenagentur|oeffentlich-rechtlich|qualitaetszeitung|behoerde|fachmedien|think-tank|international|regional|boulevard|sonstige)$"
 SOURCE_STATUS_PATTERN = "^(active|inactive)$"
 class SourceCreate(BaseModel):

src/routers/sources.py

@@ -1,13 +1,19 @@
 """Sources-Router: Quellenverwaltung (Multi-Tenant). Klassifikation: Read-Only — Pflege in der Verwaltung."""
 import json
 import logging
+import uuid
+import re
+import os
+import hashlib
 from collections import defaultdict
-from fastapi import APIRouter, Depends, HTTPException, status
+from fastapi import APIRouter, Depends, File, Form, HTTPException, UploadFile, status
 from models import SourceCreate, SourceUpdate, SourceResponse, DiscoverRequest, DiscoverResponse, DiscoverMultiResponse, DomainActionRequest
 from auth import get_current_user
 from database import db_dependency, refresh_source_counts
 from source_rules import discover_source, discover_all_feeds, evaluate_feeds_with_claude, _extract_domain, _detect_category, domain_to_display_name, _DOMAIN_ALIASES
 import aiosqlite
+from config import DB_PATH
+from typing import Optional
 
 logger = logging.getLogger("osint.sources")
 
@@ -640,3 +646,110 @@ async def trigger_refresh_counts(
     await refresh_source_counts(db)
     return {"status": "ok"}
 
+
+# --- PDF-Upload (Kundenquelle vom Typ pdf_document) ---
+# Analog zum Verwaltungs-Upload, aber tenant-spezifisch.
+# Datei landet unter <dirname(DB_PATH)>/pdfs/{sha256}.pdf.
+# Der Worker (services.pdf_ingest) verarbeitet sie asynchron im Minutentakt.
+
+MAX_PDF_SIZE_BYTES = 50 * 1024 * 1024  # 50 MB
+PDF_DIR = os.path.join(os.path.dirname(os.path.abspath(DB_PATH)), "pdfs")
+
+
+def _pdf_dir() -> str:
+    os.makedirs(PDF_DIR, exist_ok=True)
+    return PDF_DIR
+
+
+@router.post("/upload-pdf", status_code=status.HTTP_201_CREATED)
+async def upload_pdf_source(
+    current_user: dict = Depends(get_current_user),
+    db: aiosqlite.Connection = Depends(db_dependency),
+    file: UploadFile = File(...),
+    name: Optional[str] = Form(None),
+    category: str = Form("sonstige"),
+    language: Optional[str] = Form(None),
+    notes: Optional[str] = Form(None),
+):
+    """PDF hochladen + als Kundenquelle (source_type=pdf_document) registrieren.
+
+    Idempotent ueber SHA256 innerhalb des Tenants: doppelter Upload erzeugt 409.
+    """
+    head = await file.read(8)
+    if not head.startswith(b"%PDF-"):
+        raise HTTPException(status_code=415, detail="Datei ist kein gueltiges PDF")
+
+    tenant_id = current_user.get("tenant_id")
+    sha = hashlib.sha256()
+    sha.update(head)
+    total = len(head)
+    tmp_path = os.path.join(_pdf_dir(), f".upload-{uuid.uuid4().hex}.tmp")
+    try:
+        with open(tmp_path, "wb") as out:
+            out.write(head)
+            while True:
+                chunk = await file.read(1024 * 1024)
+                if not chunk:
+                    break
+                total += len(chunk)
+                if total > MAX_PDF_SIZE_BYTES:
+                    raise HTTPException(status_code=413, detail=f"PDF ueberschreitet {MAX_PDF_SIZE_BYTES // 1024 // 1024} MB")
+                sha.update(chunk)
+                out.write(chunk)
+        sha_hex = sha.hexdigest()
+        final_path = os.path.join(_pdf_dir(), f"{sha_hex}.pdf")
+        rel_path = os.path.join("pdfs", f"{sha_hex}.pdf")
+
+        # Duplikat-Pruefung innerhalb des Tenants (oder global, falls eine
+        # gleiche PDF bereits als Grundquelle existiert -> dann sichtbar fuer alle).
+        cursor = await db.execute(
+            "SELECT id, name, tenant_id FROM sources WHERE pdf_sha256 = ? "
+            "AND (tenant_id IS NULL OR tenant_id = ?)",
+            (sha_hex, tenant_id),
+        )
+        existing = await cursor.fetchone()
+        if existing:
+            os.unlink(tmp_path)
+            scope = "global" if existing["tenant_id"] is None else "Ihrer Organisation"
+            raise HTTPException(
+                status_code=409,
+                detail=f"PDF bereits in {scope} vorhanden als Quelle '{existing['name']}' (id={existing['id']})",
+            )
+
+        if not os.path.exists(final_path):
+            os.replace(tmp_path, final_path)
+        else:
+            os.unlink(tmp_path)
+    except HTTPException:
+        if os.path.exists(tmp_path):
+            try: os.unlink(tmp_path)
+            except OSError: pass
+        raise
+    except Exception as e:
+        if os.path.exists(tmp_path):
+            try: os.unlink(tmp_path)
+            except OSError: pass
+        logger.exception("PDF-Upload (tenant) fehlgeschlagen")
+        raise HTTPException(status_code=500, detail=f"PDF-Upload fehlgeschlagen: {e}")
+
+    display_name = (name or "").strip() or re.sub(r"\.pdf$", "", file.filename or "PDF", flags=re.I)
+    display_name = display_name[:200]
+
+    cursor = await db.execute(
+        """INSERT INTO sources
+           (name, url, domain, source_type, category, status, notes, language,
+            pdf_path, pdf_sha256, added_by, tenant_id)
+           VALUES (?, NULL, NULL, 'pdf_document', ?, 'active', ?, ?, ?, ?, ?, ?)""",
+        (display_name, category, notes, language, rel_path, sha_hex,
+         current_user["username"], tenant_id),
+    )
+    src_id = cursor.lastrowid
+    await db.commit()
+
+    cursor = await db.execute("SELECT * FROM sources WHERE id = ?", (src_id,))
+    row = await cursor.fetchone()
+    result = dict(row)
+    result["is_global"] = result.get("tenant_id") is None
+    result["state_affiliated"] = bool(result.get("state_affiliated"))
+    result["alignments"] = []
+    return result

src/services/pdf_ingest.py

@@ -0,0 +1,237 @@
+"""PDF-Ingest: liest hochgeladene PDFs ein und legt sie als Pool-Artikel ab.
+
+Quellen vom Typ `pdf_document` werden in der Verwaltung angelegt
+(`processed_at IS NULL`). Dieser Service pollt sie, extrahiert den Text,
+uebersetzt nach DE+EN und schreibt EINEN Artikel (incident_id=NULL) in
+`articles`. Idempotent ueber `processed_at`.
+"""
+from __future__ import annotations
+
+import asyncio
+import json
+import logging
+import os
+import re
+from typing import Optional
+
+import aiosqlite
+
+from config import DB_PATH, CLAUDE_MODEL_FAST
+from agents.claude_client import call_claude
+
+logger = logging.getLogger("osint.pdf_ingest")
+
+MAX_CHARS_PER_PDF = 200_000  # harte Obergrenze, schuetzt vor riesigen Dumps
+TRANSLATE_INPUT_MAX = 12_000  # was wir dem LLM zum Uebersetzen geben (Cost-Control)
+
+
+def _extract_text_pdfplumber(path: str) -> str:
+    import pdfplumber
+    parts: list[str] = []
+    with pdfplumber.open(path) as pdf:
+        for page in pdf.pages:
+            t = page.extract_text() or ""
+            if t:
+                parts.append(t)
+    return "\n\n".join(parts).strip()
+
+
+def _extract_text_ocr(path: str) -> str:
+    """Tesseract-Fallback ueber pdf2image -> Pillow -> pytesseract."""
+    from pdf2image import convert_from_path
+    import pytesseract
+    images = convert_from_path(path, dpi=200)
+    parts = []
+    for img in images:
+        # deu+eng zusammen, damit mehrsprachige PDFs gehen
+        t = pytesseract.image_to_string(img, lang="deu+eng")
+        if t and t.strip():
+            parts.append(t.strip())
+    return "\n\n".join(parts).strip()
+
+
+def _extract_text(path: str) -> tuple[str, str]:
+    """Gibt (text, method) zurueck. method: 'pdfplumber' oder 'ocr'."""
+    try:
+        text = _extract_text_pdfplumber(path)
+    except Exception as e:
+        logger.warning("pdfplumber-Extraktion fehlgeschlagen fuer %s: %s", path, e)
+        text = ""
+    if len(text) >= 50:
+        return text[:MAX_CHARS_PER_PDF], "pdfplumber"
+    logger.info("PDF hat keinen Text-Layer (oder <50 Zeichen), versuche OCR: %s", path)
+    text = _extract_text_ocr(path)
+    return text[:MAX_CHARS_PER_PDF], "ocr"
+
+
+def _derive_headline(text: str, fallback: str) -> str:
+    """Erste sinnvolle Zeile als Headline; sonst Fallback (Dateiname)."""
+    for raw in text.splitlines():
+        line = raw.strip()
+        if 5 <= len(line) <= 200:
+            return line
+    return fallback.strip() or "Untitled PDF"
+
+
+async def _translate(text: str, headline: str, target_lang: str) -> tuple[str, str]:
+    """Uebersetzt Headline + Content nach target_lang ('de' oder 'en').
+
+    Eigene mini-Funktion (statt agents.translator), weil wir je PDF nur EIN
+    Item haben und Headline+Content getrennt brauchen. Returnt (headline_t, content_t).
+    Bei Fehler oder leerem Text: ('', '').
+    """
+    if not text and not headline:
+        return "", ""
+    lang_label = {"de": "Deutsch", "en": "Englisch"}.get(target_lang, target_lang)
+    content_in = (text or "")[:TRANSLATE_INPUT_MAX]
+    prompt = f"""Du bist ein praeziser Uebersetzer fuer Sachtexte.
+Uebersetze Headline und Inhalt nach {lang_label}.
+
+WICHTIG:
+- Verwende IMMER echte UTF-8-Umlaute (ae->ä, oe->ö, ue->ü, ss->ß) bei Deutsch.
+- Behalte Eigennamen im Original.
+- Wenn der Text schon auf {lang_label} ist, gib ihn (nahezu) unveraendert zurueck.
+- Behalte die wichtigsten Inhalte; kuerze stark auf MAX 3000 Zeichen Content.
+
+Antworte AUSSCHLIESSLICH mit einem JSON-Objekt im Format:
+{{"headline": "...", "content": "..."}}
+
+Keine Markdown-Codefence, keine Einleitung.
+
+HEADLINE: {headline}
+INHALT:
+{content_in}
+"""
+    try:
+        result_text, _usage = await call_claude(prompt, tools=None, model=CLAUDE_MODEL_FAST)
+    except Exception as e:
+        logger.warning("PDF-Translator (%s) Claude-Call fehlgeschlagen: %s", target_lang, e)
+        return "", ""
+
+    raw = result_text.strip()
+    if raw.startswith("```"):
+        raw = re.sub(r"^```(?:json)?\s*", "", raw)
+        raw = re.sub(r"\s*```\s*$", "", raw).strip()
+    try:
+        data = json.loads(raw)
+    except json.JSONDecodeError:
+        m = re.search(r"\{.*\}", raw, re.DOTALL)
+        if not m:
+            logger.warning("PDF-Translator (%s) JSON nicht parsbar: %r", target_lang, raw[:200])
+            return "", ""
+        try:
+            data = json.loads(m.group(0))
+        except json.JSONDecodeError:
+            return "", ""
+    if not isinstance(data, dict):
+        return "", ""
+    return (data.get("headline") or "").strip(), (data.get("content") or "").strip()
+
+
+async def _process_one(db: aiosqlite.Connection, src: dict) -> None:
+    sid = src["id"]
+    name = src["name"] or "PDF"
+    rel_path = src["pdf_path"]
+    if not rel_path:
+        logger.warning("PDF-Source #%d ohne pdf_path, ueberspringe", sid)
+        return
+
+    abs_path = rel_path if os.path.isabs(rel_path) else os.path.join(
+        os.path.dirname(DB_PATH), rel_path
+    )
+    if not os.path.exists(abs_path):
+        logger.error("PDF-Datei fehlt fuer Source #%d: %s", sid, abs_path)
+        # auf processed_at setzen aber Notiz hinterlegen, damit kein Endlos-Retry
+        await db.execute(
+            "UPDATE sources SET processed_at = CURRENT_TIMESTAMP, "
+            "notes = COALESCE(notes,'') || ' [PDF-Datei nicht gefunden]' WHERE id = ?",
+            (sid,),
+        )
+        await db.commit()
+        return
+
+    logger.info("PDF-Ingest start: source #%d (%s)", sid, abs_path)
+
+    try:
+        text, method = await asyncio.to_thread(_extract_text, abs_path)
+    except Exception as e:
+        logger.exception("PDF-Extraktion fehlgeschlagen fuer #%d: %s", sid, e)
+        await db.execute(
+            "UPDATE sources SET processed_at = CURRENT_TIMESTAMP, "
+            "notes = COALESCE(notes,'') || ' [PDF-Extraktion fehlgeschlagen]' WHERE id = ?",
+            (sid,),
+        )
+        await db.commit()
+        return
+
+    if not text:
+        logger.warning("PDF #%d ergab keinen Text (auch OCR leer)", sid)
+        await db.execute(
+            "UPDATE sources SET processed_at = CURRENT_TIMESTAMP, "
+            "notes = COALESCE(notes,'') || ' [PDF leer/nicht lesbar]' WHERE id = ?",
+            (sid,),
+        )
+        await db.commit()
+        return
+
+    fallback_name = re.sub(r"\.pdf$", "", os.path.basename(abs_path), flags=re.I)
+    headline = _derive_headline(text, fallback_name)
+    # Hochgeladene PDFs sind meist deutsch oder englisch; LLM kann das im Prompt erkennen
+    src_lang = (src.get("language") or "").lower() or "auto"
+
+    # Wir senden parallel DE + EN
+    (de_h, de_c), (en_h, en_c) = await asyncio.gather(
+        _translate(text, headline, "de"),
+        _translate(text, headline, "en"),
+    )
+
+    # Originaltext kappen, damit articles-Tabelle handhabbar bleibt
+    content_original = text[:5000]
+
+    await db.execute(
+        """INSERT INTO articles (incident_id, headline, headline_de, headline_en,
+           source, source_url, content_original, content_de, content_en, language,
+           published_at, tenant_id, verification_status)
+           VALUES (NULL, ?, ?, ?, ?, ?, ?, ?, ?, ?, NULL, ?, 'unverified')""",
+        (
+            headline,
+            de_h or None,
+            en_h or None,
+            name,
+            f"pdf://{src.get('pdf_sha256') or sid}",
+            content_original,
+            de_c or None,
+            en_c or None,
+            src_lang if src_lang != "auto" else None,
+            src.get("tenant_id"),
+        ),
+    )
+    await db.execute(
+        "UPDATE sources SET processed_at = CURRENT_TIMESTAMP, article_count = article_count + 1, "
+        "last_seen_at = CURRENT_TIMESTAMP WHERE id = ?",
+        (sid,),
+    )
+    await db.commit()
+    logger.info("PDF-Ingest fertig: source #%d (%s, %d Zeichen)", sid, method, len(text))
+
+
+async def run_once() -> int:
+    """Verarbeitet alle pdf_document-Sources ohne processed_at. Returnt Anzahl.
+
+    Wird vom APScheduler als interval-Job aufgerufen. Pro Tick max 5 PDFs,
+    damit ein hochgeladener Stapel nicht einen einzelnen Lauf monopolisiert.
+    """
+    async with aiosqlite.connect(DB_PATH) as db:
+        db.row_factory = aiosqlite.Row
+        cursor = await db.execute(
+            "SELECT id, name, pdf_path, pdf_sha256, language, tenant_id "
+            "FROM sources WHERE source_type = 'pdf_document' AND processed_at IS NULL "
+            "ORDER BY created_at ASC LIMIT 5"
+        )
+        rows = [dict(r) for r in await cursor.fetchall()]
+        for src in rows:
+            try:
+                await _process_one(db, src)
+            except Exception:
+                logger.exception("PDF-Ingest unerwarteter Fehler bei source #%d", src["id"])
+        return len(rows)

src/static/dashboard.html

@@ -555,6 +555,7 @@
                         <input type="text" id="sources-search" class="timeline-filter-input sources-search-input" placeholder="Suche..." oninput="App.filterSources()" data-i18n-attr="placeholder:sources_modal.search_placeholder">
                     </div>
                     <div class="sources-toolbar-actions">
+                        <button class="btn btn-secondary btn-small" onclick="App.openPdfUpload()" style="margin-right:8px;">+ PDF hochladen</button>
                         <button class="btn btn-primary btn-small" onclick="App.toggleSourceForm()" data-i18n="sources_modal.add_source">+ Quelle</button>
                     </div>
                 </div>
@@ -633,6 +634,57 @@
         </div>
     </div>
 
+    <!-- Modal: PDF als Quelle hochladen -->
+    <div class="modal-overlay" id="modal-pdf-upload" role="dialog" aria-modal="true" aria-labelledby="modal-pdf-upload-title">
+        <div class="modal">
+            <div class="modal-header">
+                <div class="modal-title" id="modal-pdf-upload-title">PDF als Quelle hochladen</div>
+                <button class="modal-close" onclick="closeModal(&#39;modal-pdf-upload&#39;)" aria-label="Schliessen">&times;</button>
+            </div>
+            <form id="pdf-upload-form" enctype="multipart/form-data">
+                <div class="modal-body">
+                    <p class="text-secondary" style="margin-top:0;">
+                        Die PDF wird gespeichert und im Hintergrund verarbeitet: Text wird extrahiert (OCR-Fallback fuer gescannte Dokumente) und nach Deutsch und Englisch uebersetzt. Sie erscheint danach in Ihrer Quellenliste.
+                    </p>
+                    <div class="form-group">
+                        <label for="pdf-upload-file">PDF-Datei (max. 50 MB)</label>
+                        <input type="file" id="pdf-upload-file" accept="application/pdf,.pdf" required>
+                    </div>
+                    <div class="form-group">
+                        <label for="pdf-upload-name">Anzeige-Name (optional)</label>
+                        <input type="text" id="pdf-upload-name" maxlength="200" placeholder="leer = Dateiname">
+                    </div>
+                    <div style="display:grid;grid-template-columns:1fr 1fr;gap:12px;">
+                        <div class="form-group">
+                            <label for="pdf-upload-category">Kategorie</label>
+                            <select id="pdf-upload-category">
+                                <option value="sonstige" selected>Sonstige</option>
+                                <option value="behoerde">Behoerde</option>
+                                <option value="think-tank">Think-Tank</option>
+                                <option value="fachmedien">Fachmedien</option>
+                                <option value="international">International</option>
+                            </select>
+                        </div>
+                        <div class="form-group">
+                            <label for="pdf-upload-language">Sprache (optional)</label>
+                            <input type="text" id="pdf-upload-language" placeholder="z.B. Deutsch, Englisch">
+                        </div>
+                    </div>
+                    <div class="form-group">
+                        <label for="pdf-upload-notes">Notizen</label>
+                        <input type="text" id="pdf-upload-notes" placeholder="Optional">
+                    </div>
+                    <div id="pdf-upload-error" class="error-msg" style="display:none"></div>
+                    <div id="pdf-upload-progress" class="text-secondary" style="display:none;margin-top:8px;">Laedt hoch &hellip;</div>
+                </div>
+                <div class="modal-footer">
+                    <button type="button" class="btn btn-secondary" onclick="closeModal(&#39;modal-pdf-upload&#39;)">Abbrechen</button>
+                    <button type="submit" class="btn btn-primary" id="pdf-upload-submit">Hochladen</button>
+                </div>
+            </form>
+        </div>
+    </div>
+
     <!-- Modal: Content-Viewer (wiederverwendbar für Lagebild, Faktencheck, Quellenübersicht, Timeline) -->
     <div class="modal-overlay" id="modal-content-viewer" role="dialog" aria-modal="true" aria-labelledby="content-viewer-title">
         <div class="modal modal-content-viewer">

src/static/js/api.js

@@ -22,6 +22,31 @@ const API = {
         };
     },
 
+    async upload(path, formData) {
+        const token = localStorage.getItem("osint_token");
+        const headers = {};
+        if (token) headers["Authorization"] = `Bearer ${token}`;
+        const response = await fetch(`${this.baseUrl}${path}`, {
+            method: "POST",
+            headers,
+            body: formData,
+        });
+        if (response.status === 401) {
+            localStorage.removeItem("osint_token");
+            localStorage.removeItem("osint_username");
+            window.location.href = "/";
+            return;
+        }
+        if (!response.ok) {
+            const data = await response.json().catch(() => ({}));
+            let d = data.detail;
+            if (Array.isArray(d)) d = d.map(e => e.msg || JSON.stringify(e)).join("; ");
+            else if (typeof d === "object" && d !== null) d = JSON.stringify(d);
+            throw new Error(d || `Fehler ${response.status}`);
+        }
+        return response.json();
+    },
+
     async _request(method, path, body = null, externalSignal = null) {
         const controller = new AbortController();
         const timeout = setTimeout(() => controller.abort(), 30000);

src/static/js/app.js

@@ -3106,6 +3106,70 @@ async handleRefresh() {
 
     _discoveredData: null,
 
+    openPdfUpload() {
+        const form = document.getElementById("pdf-upload-form");
+        if (form) form.reset();
+        const err = document.getElementById("pdf-upload-error");
+        if (err) { err.style.display = "none"; err.textContent = ""; }
+        const prog = document.getElementById("pdf-upload-progress");
+        if (prog) prog.style.display = "none";
+        openModal("modal-pdf-upload");
+        this._bindPdfUploadFormOnce();
+    },
+
+    _bindPdfUploadFormOnce() {
+        const form = document.getElementById("pdf-upload-form");
+        if (!form || form.dataset.bound === "1") return;
+        form.dataset.bound = "1";
+        form.addEventListener("submit", async (e) => {
+            e.preventDefault();
+            const errEl = document.getElementById("pdf-upload-error");
+            const progEl = document.getElementById("pdf-upload-progress");
+            const submitBtn = document.getElementById("pdf-upload-submit");
+            errEl.style.display = "none";
+
+            const fileInput = document.getElementById("pdf-upload-file");
+            const f = fileInput && fileInput.files && fileInput.files[0];
+            if (!f) {
+                errEl.textContent = "Bitte eine PDF-Datei auswaehlen.";
+                errEl.style.display = "block";
+                return;
+            }
+            if (f.size > 50 * 1024 * 1024) {
+                errEl.textContent = "Datei ueberschreitet 50 MB.";
+                errEl.style.display = "block";
+                return;
+            }
+
+            const fd = new FormData();
+            fd.append("file", f);
+            const nm = (document.getElementById("pdf-upload-name").value || "").trim();
+            if (nm) fd.append("name", nm);
+            fd.append("category", document.getElementById("pdf-upload-category").value || "sonstige");
+            const lng = (document.getElementById("pdf-upload-language").value || "").trim();
+            if (lng) fd.append("language", lng);
+            const nt = (document.getElementById("pdf-upload-notes").value || "").trim();
+            if (nt) fd.append("notes", nt);
+
+            submitBtn.disabled = true;
+            progEl.style.display = "block";
+            try {
+                await API.upload("/sources/upload-pdf", fd);
+                closeModal("modal-pdf-upload");
+                if (typeof UI !== "undefined" && UI.showToast) {
+                    UI.showToast("PDF hochgeladen -- Verarbeitung laeuft im Hintergrund", "success");
+                }
+                await App.loadSources();
+            } catch (err) {
+                errEl.textContent = err && err.message ? err.message : "Upload fehlgeschlagen";
+                errEl.style.display = "block";
+            } finally {
+                submitBtn.disabled = false;
+                progEl.style.display = "none";
+            }
+        });
+    },
+
     toggleSourceForm(show) {
         const form = document.getElementById('sources-add-form');
         if (!form) return;