diff --git a/backend/Dockerfile b/backend/Dockerfile
index a9b991e..7a8c461 100644
--- a/backend/Dockerfile
+++ b/backend/Dockerfile
@@ -18,11 +18,13 @@ RUN uv sync --frozen --no-dev --no-install-project
 FROM base AS runtime
 
 COPY . .
-RUN uv sync --frozen --no-dev \
+ENV PLAYWRIGHT_BROWSERS_PATH=/ms-playwright
+RUN uv sync --frozen --extra live --no-dev \
+    && playwright install --with-deps chromium \
     && sed -i 's/\r$//' /app/scripts/startup.sh \
     && chmod +x /app/scripts/startup.sh \
     && useradd --create-home --uid 1000 app \
-    && chown -R app:app /app
+    && chown -R app:app /app /ms-playwright
 
 USER app
 EXPOSE 8000
diff --git a/backend/app/discovery/urlgen.py b/backend/app/discovery/urlgen.py
index e7fa2d3..5b73888 100644
--- a/backend/app/discovery/urlgen.py
+++ b/backend/app/discovery/urlgen.py
@@ -27,11 +27,17 @@ def generate_search_url(
     base = urlsplit(category_url)
     query = dict(parse_qsl(base.query, keep_blank_values=True))
     extra: list[tuple[str, str]] = []
+    xf_values: list[str] = []
     for name, value in fragments:
-        if name in query and name != "xf":
+        if name == "xf":
+            xf_values.append(value)
+            continue
+        if name in query:
             query[name] = value
         else:
             extra.append((name, value))
+    if xf_values:
+        extra.append(("xf", "~".join(xf_values)))
     encoded = urlencode(list(query.items()) + extra, doseq=True, safe="+", quote_via=quote)
     url = urlunsplit((base.scheme, base.netloc, base.path, encoded, base.fragment))
     return url, internal_only
@@ -40,11 +46,15 @@ def generate_search_url(
 def _encode_mapping(mapping: dict[str, Any], criterion: dict[str, Any]) -> list[tuple[str, str]]:
     encoding = mapping.get("encoding") or {}
     param = str(encoding.get("url_param") or "xf")
+    if encoding.get("encoding_type") == "RANGE_PARAMETERS" or param in {"bpmin", "bpmax"}:
+        return _encode_range(encoding, criterion)
+    if encoding.get("encoding_type") == "RANGE_TOKEN":
+        return _encode_range_token(mapping, criterion)
     if param == "xf":
         filter_key = str(mapping["external_filter_key"])
-        source_value = str(mapping.get("source_value") or _threshold_value(mapping, criterion))
+        source_value = _xf_value(mapping, criterion)
         return [("xf", f"{filter_key}_{_plus(source_value)}")]
-    if param in {"minp", "maxp", "prmin", "prmax"}:
+    if param in {"minp", "maxp", "prmin", "prmax", "asuch", "fs"}:
         return [(param, str(criterion.get("value")))]
     template = encoding.get("value_template")
     if template:
@@ -52,6 +62,63 @@ def _encode_mapping(mapping: dict[str, Any], criterion: dict[str, Any]) -> list[
     raise ValueError(f"unsupported encoding for mapping {mapping.get('criterion_key')}")
 
 
+def _encode_range(encoding: dict[str, Any], criterion: dict[str, Any]) -> list[tuple[str, str]]:
+    operator = str(criterion.get("operator") or "<=")
+    value = criterion.get("value")
+    min_param = str(encoding.get("min_parameter") or encoding.get("url_param") or "bpmin")
+    max_param = str(encoding.get("max_parameter") or "bpmax")
+    if operator == "between":
+        if isinstance(value, dict):
+            items: list[tuple[str, str]] = []
+            if value.get("min") is not None:
+                items.append((min_param, str(value["min"])))
+            if value.get("max") is not None:
+                items.append((max_param, str(value["max"])))
+            return items
+        if isinstance(value, list) and len(value) == 2:
+            return [(min_param, str(value[0])), (max_param, str(value[1]))]
+    if operator in {">=", ">"}:
+        return [(min_param, str(value))]
+    return [(max_param, str(value))]
+
+
+def _encode_range_token(mapping: dict[str, Any], criterion: dict[str, Any]) -> list[tuple[str, str]]:
+    filter_key = str(mapping["external_filter_key"])
+    looked_up = _lookup_option(mapping, criterion.get("value"))
+    if looked_up is not None:
+        return [("xf", f"{filter_key}_{_plus(looked_up)}")]
+    operator = str(criterion.get("operator") or "<=")
+    value = criterion.get("value")
+    if operator == "between":
+        if isinstance(value, dict):
+            raw = f"{value.get('min') or ''}*{value.get('max') or ''}"
+        elif isinstance(value, list) and len(value) == 2:
+            raw = f"{value[0]}*{value[1]}"
+        else:
+            raw = f"*{value}"
+    elif operator in {">=", ">"}:
+        raw = f"{value}*"
+    else:
+        raw = f"*{value}"
+    return [("xf", f"{filter_key}_{raw}")]
+
+
+def _lookup_option(mapping: dict[str, Any], value: Any) -> str | None:
+    options = (mapping.get("encoding") or {}).get("options") or {}
+    if value in options:
+        return str(options[value])
+    if str(value) in options:
+        return str(options[str(value)])
+    return None
+
+
+def _xf_value(mapping: dict[str, Any], criterion: dict[str, Any]) -> str:
+    looked_up = _lookup_option(mapping, criterion.get("value"))
+    if looked_up is not None:
+        return looked_up
+    return str(mapping.get("source_value") or _threshold_value(mapping, criterion))
+
+
 def _threshold_value(mapping: dict[str, Any], criterion: dict[str, Any]) -> str:
     option = mapping.get("source_value")
     if option:
@@ -72,8 +139,12 @@ def decode_xf_roundtrip(url: str) -> list[tuple[str, str]]:
     parts = urlsplit(url)
     decoded: list[tuple[str, str]] = []
     for name, value in parse_qsl(parts.query, keep_blank_values=True):
-        if name == "xf" and "_" in value:
-            key, raw = value.split("_", 1)
+        if name != "xf":
+            continue
+        for part in value.replace("|", "~").split("~"):
+            if "_" not in part:
+                continue
+            key, raw = part.split("_", 1)
             decoded.append((key, raw.replace("+", " ")))
     return decoded
 
diff --git a/backend/app/services/catalog_service.py b/backend/app/services/catalog_service.py
index 3029d60..27d519c 100644
--- a/backend/app/services/catalog_service.py
+++ b/backend/app/services/catalog_service.py
@@ -100,7 +100,7 @@ class CatalogService:
                 SourceCategoryRecord.product_category_id == product_category_id,
             )
         ).first()
-        if source_category is not None and source_category.taxonomy_status == TaxonomyStatus.NOT_LEARNED.value:
+        if source_category is not None and self._should_learn(source_category, source_id):
             from app.services.discovery_service import TaxonomyService
 
             TaxonomyService(self.session).discover(source_id, source_category.external_category_key)
@@ -172,7 +172,33 @@ class CatalogService:
             if definition.id in seen:
                 continue
             criteria.append(self._criterion_out(definition, None, None, [], None))
-        return jsonable({"product_category_id": product_category_id, "source": source_id, "criteria": criteria})
+        return jsonable(
+            {
+                "product_category_id": product_category_id,
+                "source": source_id,
+                "taxonomy_status": None if source_category is None else source_category.taxonomy_status,
+                "criteria": criteria,
+            }
+        )
+
+    def _should_learn(self, source_category: SourceCategoryRecord, source_id: str) -> bool:
+        if source_category.taxonomy_status == TaxonomyStatus.NOT_LEARNED.value:
+            return True
+        if source_category.taxonomy_status != TaxonomyStatus.READY.value:
+            return False
+        from app.settings import get_settings
+
+        if get_settings().geizhals_use_fixtures or source_id != "geizhals":
+            return False
+        keys = {
+            row.external_filter_key
+            for row in self.session.scalars(
+                select(SourceFilterDefinitionRecord).where(
+                    SourceFilterDefinitionRecord.source_category_id == source_category.id
+                )
+            )
+        }
+        return keys <= {"17023", "price"}
 
     def list_templates(self) -> list[dict[str, Any]]:
         rows, _total = self.catalog.list_profiles(200, 0)
diff --git a/backend/app/services/discovery_service.py b/backend/app/services/discovery_service.py
index 2bc0137..e81dea3 100644
--- a/backend/app/services/discovery_service.py
+++ b/backend/app/services/discovery_service.py
@@ -14,6 +14,7 @@ from app.discovery.diff import diff_run, next_lifecycle
 from app.discovery.enums import (
     CandidateStatus,
     DiffStatus,
+    FilterType,
     LifecycleStatus,
     MappingStatus,
     MatchStatus,
@@ -99,7 +100,20 @@ class TaxonomyService:
         if catalog.get_source(source_id) is None:
             raise AppError(404, "NOT_FOUND", "source not found", {"source_id": source_id})
         adapter = self.adapter or discovery_adapter(source_id)
-        payload = adapter.discover_taxonomy(category_key)
+        try:
+            payload = adapter.discover_taxonomy(category_key)
+        except AppError as exc:
+            category = self.session.scalars(
+                select(SourceCategoryRecord).where(
+                    SourceCategoryRecord.source_id == source_id,
+                    SourceCategoryRecord.external_category_key == category_key,
+                )
+            ).first()
+            if category is not None and exc.code == "SOURCE_BLOCKED":
+                category.taxonomy_status = TaxonomyStatus.BLOCKED.value
+                category.last_taxonomy_scan_at = now_utc()
+                self.session.flush()
+            raise
         now = now_utc()
         category = self.session.scalars(
             select(SourceCategoryRecord).where(
@@ -148,7 +162,7 @@ class TaxonomyService:
                     label=str(raw.get("label") or raw["external_filter_key"]),
                     filter_type=filter_type,
                     canonical_unit=raw.get("canonical_unit"),
-                    status=MappingStatus.VERIFIED.value,
+                    status=MappingStatus.LEARNED_UNVERIFIED.value,
                 )
                 self.session.add(filt)
                 self.session.flush()
@@ -161,9 +175,14 @@ class TaxonomyService:
             self._sync_options(filt, raw.get("options") or [])
             if raw.get("range"):
                 self._sync_range(filt, raw["range"])
-        category.taxonomy_status = (
-            TaxonomyStatus.READY.value if payload.get("filters") else TaxonomyStatus.BROKEN.value
-        )
+            self._sync_learned_mapping(category, filt, raw)
+        requested = str(payload.get("taxonomy_status") or "")
+        if requested in {item.value for item in TaxonomyStatus}:
+            category.taxonomy_status = requested
+        else:
+            category.taxonomy_status = (
+                TaxonomyStatus.READY.value if payload.get("filters") else TaxonomyStatus.BROKEN.value
+            )
         self.session.flush()
         return self.filters(source_id, category.external_category_key)
 
@@ -275,6 +294,89 @@ class TaxonomyService:
         if removed:
             self._break_mappings(filt.source_category_id, filt.external_filter_key)
 
+    def _sync_learned_mapping(
+        self,
+        category: SourceCategoryRecord,
+        filt: SourceFilterDefinitionRecord,
+        raw: dict[str, Any],
+    ) -> None:
+        encoding = raw.get("encoding") or {"url_param": "xf"}
+        definition = upsert_definition(
+            self.session,
+            {
+                "key": f"{category.source_id}:{category.external_category_key}:{filt.external_filter_key}",
+                "origin": "SOURCE",
+                "source": category.source_id,
+                "product_category_id": category.product_category_id,
+                "label": filt.label,
+                "value_type": "number"
+                if filt.filter_type in {FilterType.NUMERIC_RANGE.value, FilterType.NUMERIC_THRESHOLD.value}
+                else "string",
+                "unit": filt.canonical_unit,
+                "category": category.product_category_id,
+            },
+        )
+        mapping = self.session.scalars(
+            select(CriterionMappingRecord).where(
+                CriterionMappingRecord.source_category_id == category.id,
+                CriterionMappingRecord.external_filter_key == filt.external_filter_key,
+            )
+        ).first()
+        options = raw.get("options") or []
+        source_value = None
+        if options:
+            source_value = str(options[0].get("query_value") or options[0].get("source_value"))
+        now = now_utc()
+        if mapping is None:
+            mapping = CriterionMappingRecord(
+                id=uuid4(),
+                criterion_definition_id=definition.id,
+                criterion_key=definition.key,
+                source_id=category.source_id,
+                source_category_id=category.id,
+                external_filter_key=filt.external_filter_key,
+                source_value=source_value,
+                encoding=encoding,
+                status=MappingStatus.LEARNED_UNVERIFIED.value,
+                definition_hash=sha256_text(canonical_json({"encoding": encoding, "source_value": source_value})),
+                created_at=now,
+                updated_at=now,
+            )
+            self.session.add(mapping)
+            self.session.flush()
+        else:
+            mapping.criterion_definition_id = definition.id
+            mapping.criterion_key = definition.key
+            mapping.source_value = source_value
+            mapping.encoding = encoding
+            mapping.definition_hash = sha256_text(canonical_json({"encoding": encoding, "source_value": source_value}))
+            mapping.updated_at = now
+        probe_value = source_value or 0
+        if options and options[0].get("normalized_value") is not None:
+            probe_value = options[0]["normalized_value"]
+        operator = (options[0].get("operator") if options else None) or "<="
+        if filt.filter_type == FilterType.NUMERIC_RANGE.value:
+            operator = ">="
+            probe_value = 1000
+        verified = verify_roundtrip(
+            category_url=category.source_url or f"https://geizhals.de/?cat={category.external_category_key}",
+            criterion={
+                "criterion": definition.key,
+                "operator": operator,
+                "value": probe_value,
+            },
+            mapping={
+                "criterion_key": definition.key,
+                "external_filter_key": filt.external_filter_key,
+                "source_value": source_value,
+                "encoding": encoding,
+            },
+        )
+        if mapping.status != MappingStatus.DISABLED.value and verified:
+            mapping.status = MappingStatus.VERIFIED.value
+        elif mapping.status == MappingStatus.BROKEN.value:
+            mapping.status = MappingStatus.LEARNED_UNVERIFIED.value
+
     def _sync_range(self, filt: SourceFilterDefinitionRecord, raw: dict[str, Any]) -> None:
         row = self.session.get(SourceFilterRangeRecord, filt.id)
         if row is None:
diff --git a/backend/app/settings.py b/backend/app/settings.py
index bd04c9b..d0947fa 100644
--- a/backend/app/settings.py
+++ b/backend/app/settings.py
@@ -16,6 +16,8 @@ class Settings(BaseSettings):
     engine_build: str = "dev"
     clock_now: datetime | None = None
     cors_allowed_origins: str = "http://localhost:5173,http://127.0.0.1:5173"
+    geizhals_use_fixtures: bool = False
+    geizhals_playwright: bool = True
 
     @property
     def cors_origins(self) -> list[str]:
diff --git a/backend/app/sources/geizhals/adapter.py b/backend/app/sources/geizhals/adapter.py
index dcbc3d0..b140afd 100644
--- a/backend/app/sources/geizhals/adapter.py
+++ b/backend/app/sources/geizhals/adapter.py
@@ -7,6 +7,10 @@ from urllib.parse import parse_qs, urlparse
 
 from app.discovery.enums import FilterType
 from app.errors import AppError
+from app.settings import get_settings
+from app.sources.geizhals.live_page import load_category_page, load_result_page
+from app.sources.geizhals.result_parser import parse_result_html, with_page
+from app.sources.geizhals.taxonomy_parser import parse_taxonomy_html
 
 FIXTURE_DIR = Path(__file__).resolve().parents[3] / "fixtures" / "geizhals"
 
@@ -14,12 +18,31 @@ FIXTURE_DIR = Path(__file__).resolve().parents[3] / "fixtures" / "geizhals"
 class GeizhalsAdapter:
     source_key = "geizhals"
 
-    def __init__(self, fixture_dir: Path | None = None, live: bool = False) -> None:
+    def __init__(self, fixture_dir: Path | None = None, live: bool | None = None) -> None:
         self.fixture_dir = fixture_dir or FIXTURE_DIR
-        self.live = live
+        settings = get_settings()
+        self.live = (not settings.geizhals_use_fixtures) if live is None else live
 
     def discover_taxonomy(self, category_ref: str) -> dict[str, Any]:
-        payload = self._load(f"{category_ref}-taxonomy.json")
+        if not self.live:
+            return self._load(f"{category_ref}-taxonomy.json")
+        url = f"https://geizhals.de/?cat={category_ref}&promode=true"
+        html, mode, status = load_category_page(
+            url,
+            category_ref=category_ref,
+            playwright=get_settings().geizhals_playwright,
+        )
+        payload = parse_taxonomy_html(html, category_ref=category_ref, source_url=url)
+        if payload.get("blocked") or not payload.get("filters"):
+            raise AppError(
+                422,
+                "TAXONOMY_EMPTY",
+                "Geizhals-Taxonomie konnte nicht aus der Kategorieseite gelesen werden",
+                {"url": url, "http_status": status, "mode": mode},
+            )
+        payload["discovery_mode"] = mode
+        payload["http_status"] = status
+        payload["taxonomy_status"] = "DEGRADED" if mode == "recorded_html" else "READY"
         return payload
 
     def build_query(self, criteria: list[dict[str, Any]], mappings: list[dict[str, Any]]) -> dict[str, Any]:
@@ -34,13 +57,32 @@ class GeizhalsAdapter:
         return {"url": url, "internal_only": internal_only}
 
     def fetch_results(self, query: dict[str, Any], pagination: dict[str, Any]) -> dict[str, Any]:
-        if self.live:
-            raise AppError(422, "LIVE_FETCH_DISABLED", "live Geizhals fetch is not enabled in this process")
         page = int(pagination.get("page") or 1)
-        category = parse_qs(urlparse(query["url"]).query).get("cat", ["nb"])[0]
-        return self._load(f"{category}-search-page{page}.json")
+        if not self.live:
+            category = parse_qs(urlparse(query["url"]).query).get("cat", ["nb"])[0]
+            return self._load(f"{category}-search-page{page}.json")
+        url = with_page(str(query["url"]), page)
+        html, mode, status, blocked = load_result_page(url, playwright=get_settings().geizhals_playwright)
+        if blocked:
+            return {
+                "page": page,
+                "page_count": page,
+                "result_count": 0,
+                "http_status": status or 403,
+                "blocked": True,
+                "challenge": "captcha" if status in {401, 403} else None,
+                "items": [],
+                "discovery_mode": mode,
+            }
+        payload = parse_result_html(html, page_url=url)
+        payload["page"] = page
+        payload["http_status"] = status
+        payload["discovery_mode"] = mode
+        return payload
 
     def parse_results(self, response: dict[str, Any]) -> list[dict[str, Any]]:
+        if response.get("html") and not response.get("items"):
+            return list(parse_result_html(str(response["html"]), page_url=str(response.get("url") or "")).get("items") or [])
         return list(response.get("items") or [])
 
     def normalize_result(self, result: dict[str, Any]) -> dict[str, Any]:
diff --git a/backend/app/sources/geizhals/live_page.py b/backend/app/sources/geizhals/live_page.py
new file mode 100644
index 0000000..043ea50
--- /dev/null
+++ b/backend/app/sources/geizhals/live_page.py
@@ -0,0 +1,104 @@
+from __future__ import annotations
+
+from pathlib import Path
+
+from app.errors import AppError
+from app.sources.geizhals.taxonomy_parser import is_challenge_html
+
+FIXTURE_DIR = Path(__file__).resolve().parents[3] / "fixtures" / "geizhals"
+BROWSER_HEADERS = {
+    "User-Agent": (
+        "Mozilla/5.0 (Windows NT 10.0; Win64; x64) "
+        "AppleWebKit/537.36 (KHTML, like Gecko) Chrome/128.0.0.0 Safari/537.36"
+    ),
+    "Accept": "text/html,application/xhtml+xml,application/xml;q=0.9,*/*;q=0.8",
+    "Accept-Language": "de-DE,de;q=0.9,en;q=0.8",
+}
+EXPAND_LABELS = (
+    "Immer alle Filter anzeigen",
+    "weitere Filter anzeigen",
+)
+
+
+def load_result_page(url: str, *, playwright: bool = True) -> tuple[str, str, int, bool]:
+    status, html = 0, ""
+    try:
+        status, html = _fetch_http(url)
+        if not _blocked(status, html):
+            return html, "live_http", status, False
+    except Exception:
+        status, html = 0, ""
+    if playwright:
+        try:
+            status, html = _fetch_playwright(url)
+            if not _blocked(status, html):
+                return html, "live_playwright", status, False
+        except Exception:
+            pass
+    if html and not is_challenge_html(html) and status and status < 400:
+        return html, "live_http", status, False
+    return html, "blocked", status or 403, True
+
+
+def load_category_page(url: str, *, category_ref: str, playwright: bool = True) -> tuple[str, str, int]:
+    status, html = 0, ""
+    try:
+        status, html = _fetch_http(url)
+        if not _blocked(status, html):
+            return html, "live_http", status
+    except Exception:
+        status, html = 0, ""
+    if playwright:
+        try:
+            status, html = _fetch_playwright(url)
+            if not _blocked(status, html):
+                return html, "live_playwright", status
+        except Exception:
+            pass
+    recorded = FIXTURE_DIR / f"{category_ref}-category.html"
+    if recorded.exists():
+        return recorded.read_text(encoding="utf-8"), "recorded_html", 0
+    raise AppError(
+        423,
+        "SOURCE_BLOCKED",
+        "Geizhals blockiert den Taxonomie-Abruf (403/Challenge)",
+        {"url": url, "http_status": status},
+    )
+
+
+def _fetch_http(url: str) -> tuple[int, str]:
+    import httpx
+
+    response = httpx.get(url, headers=BROWSER_HEADERS, follow_redirects=True, timeout=25.0)
+    return response.status_code, response.text
+
+
+def _fetch_playwright(url: str) -> tuple[int, str]:
+    try:
+        from playwright.sync_api import sync_playwright
+    except ImportError as exc:
+        raise AppError(
+            423,
+            "SOURCE_BLOCKED",
+            "Geizhals verlangt einen Browser-Abruf, Playwright ist nicht installiert",
+        ) from exc
+    with sync_playwright() as playwright:
+        browser = playwright.chromium.launch(headless=True)
+        page = browser.new_page(locale="de-DE", extra_http_headers=BROWSER_HEADERS)
+        response = page.goto(url, wait_until="domcontentloaded", timeout=45000)
+        page.wait_for_timeout(4000)
+        for label in EXPAND_LABELS:
+            button = page.get_by_text(label, exact=False)
+            if button.count():
+                button.first.click(timeout=5000)
+                page.wait_for_timeout(1500)
+        html = page.content()
+        status = response.status if response is not None else 200
+        browser.close()
+        return status, html
+
+
+def _blocked(status: int, html: str) -> bool:
+    if status in {401, 403, 429}:
+        return True
+    return is_challenge_html(html)
diff --git a/backend/app/sources/geizhals/result_parser.py b/backend/app/sources/geizhals/result_parser.py
new file mode 100644
index 0000000..4caefd5
--- /dev/null
+++ b/backend/app/sources/geizhals/result_parser.py
@@ -0,0 +1,285 @@
+from __future__ import annotations
+
+import re
+from typing import Any
+from urllib.parse import parse_qsl, urlencode, urljoin, urlsplit
+
+from app.sources.geizhals.taxonomy_parser import is_challenge_html, node_text, parse_html_tree
+
+PRODUCT_ID_RE = re.compile(r"-a(\d+)\.html", re.IGNORECASE)
+PRICE_RE = re.compile(r"(\d{1,3}(?:\.\d{3})+|\d+),(\d{2})")
+OFFER_RE = re.compile(r"(\d+)\s*Angebote?", re.IGNORECASE)
+PRODUCT_COUNT_RE = re.compile(r"(\d[\d.]*)\s*Produkte", re.IGNORECASE)
+CAPACITY_RE = re.compile(r"(\d+(?:[.,]\d+)?)\s*(TB|GB)", re.IGNORECASE)
+WEIGHT_RE = re.compile(r"(\d+(?:[.,]\d+)?)\s*kg", re.IGNORECASE)
+AVAILABILITY_BY_CODE = {"0": "out_of_stock", "1": "limited", "2": "in_stock"}
+AVAILABILITY_BY_LABEL = {
+    "lagernd": "in_stock",
+    "kurzfristig lieferbar": "limited",
+    "kurzfristig": "limited",
+    "nicht verfügbar": "out_of_stock",
+}
+
+
+def parse_result_html(html: str, *, page_url: str) -> dict[str, Any]:
+    if is_challenge_html(html):
+        return {"items": [], "page_count": 1, "result_count": 0, "blocked": True}
+    root = parse_html_tree(html)
+    cards = _find_cards(root)
+    items: list[dict[str, Any]] = []
+    seen: set[str] = set()
+    for card in cards:
+        item = _parse_card(card, page_url)
+        if item is None or item["source_external_id"] in seen:
+            continue
+        seen.add(item["source_external_id"])
+        items.append(item)
+    page_count = _page_count(root)
+    result_count = _result_count(root) or len(items)
+    return {
+        "items": items,
+        "page_count": page_count,
+        "result_count": result_count,
+        "blocked": False,
+    }
+
+
+def with_page(url: str, page: int) -> str:
+    parts = urlsplit(url)
+    query = [(name, value) for name, value in parse_qsl(parts.query, keep_blank_values=True) if name != "pg"]
+    if page > 1:
+        query.append(("pg", str(page)))
+    return parts._replace(query=urlencode(query, doseq=True)).geturl()
+
+
+def _find_cards(root: Any) -> list[Any]:
+    cards: list[Any] = []
+
+    def walk(node: Any) -> None:
+        classes = node.attrs.get("class", "")
+        if "galleryview__item" in classes or "listview__item" in classes:
+            cards.append(node)
+            return
+        for child in node.children:
+            if not isinstance(child, str):
+                walk(child)
+
+    walk(root)
+    return cards
+
+
+def _parse_card(card: Any, page_url: str) -> dict[str, Any] | None:
+    link = _first_link(card, "name-link") or _first_product_link(card)
+    if link is None:
+        return None
+    href = link.attrs.get("href") or ""
+    match = PRODUCT_ID_RE.search(href)
+    if match is None:
+        return None
+    title = (link.attrs.get("title") or node_text(link)).strip()
+    if not title:
+        return None
+    source_url = urljoin(page_url, href.split("#")[0])
+    raw_attributes = _definition_list(card)
+    sku = _first_text_by_class(card, "galleryview__mpn") or _first_text_by_class(card, "listview__mpn")
+    price_text = _first_text_by_class(card, "price")
+    offers_text = _first_text_by_class(card, "offercount") or ""
+    availability_node = _first_by_attr(card, "data-availability")
+    manufacturer, model = _split_title(title)
+    return {
+        "source_external_id": match.group(1),
+        "source_url": source_url,
+        "source_title": re.sub(r"\s+", " ", title),
+        "manufacturer": manufacturer,
+        "model": model,
+        "manufacturer_sku": sku,
+        "gtin_ean": None,
+        "best_price": _parse_price(price_text),
+        "currency": "EUR" if price_text and "€" in price_text else None,
+        "availability": _availability(availability_node),
+        "seller_count": _seller_count(offers_text),
+        "raw_attributes": raw_attributes,
+        "normalized_attributes": _normalize_attributes(raw_attributes),
+    }
+
+
+def _first_link(node: Any, fragment: str) -> Any | None:
+    classes = node.attrs.get("class", "")
+    if node.tag == "a" and fragment in classes:
+        return node
+    for child in node.children:
+        if not isinstance(child, str):
+            found = _first_link(child, fragment)
+            if found is not None:
+                return found
+    return None
+
+
+def _first_product_link(node: Any) -> Any | None:
+    if node.tag == "a" and PRODUCT_ID_RE.search(node.attrs.get("href") or ""):
+        href = node.attrs.get("href") or ""
+        if "#ratings" not in href and "#offerlist" not in href:
+            return node
+    for child in node.children:
+        if not isinstance(child, str):
+            found = _first_product_link(child)
+            if found is not None:
+                return found
+    return None
+
+
+def _first_text_by_class(node: Any, fragment: str) -> str | None:
+    found = _first_by_class(node, fragment)
+    if found is None:
+        return None
+    text = node_text(found).strip()
+    return re.sub(r"\s+", " ", text) or None
+
+
+def _first_by_class(node: Any, fragment: str) -> Any | None:
+    if fragment in node.attrs.get("class", ""):
+        return node
+    for child in node.children:
+        if not isinstance(child, str):
+            found = _first_by_class(child, fragment)
+            if found is not None:
+                return found
+    return None
+
+
+def _first_by_attr(node: Any, name: str) -> Any | None:
+    if name in node.attrs:
+        return node
+    for child in node.children:
+        if not isinstance(child, str):
+            found = _first_by_attr(child, name)
+            if found is not None:
+                return found
+    return None
+
+
+def _definition_list(node: Any) -> dict[str, str]:
+    pairs: dict[str, str] = {}
+
+    def walk(current: Any) -> None:
+        if current.tag == "dl":
+            pending: str | None = None
+            for child in current.children:
+                if isinstance(child, str):
+                    continue
+                if child.tag == "dt":
+                    pending = re.sub(r"\s+", " ", node_text(child)).strip()
+                elif child.tag == "dd" and pending:
+                    pairs[pending] = re.sub(r"\s+", " ", node_text(child)).strip()
+                    pending = None
+            return
+        for child in current.children:
+            if not isinstance(child, str):
+                walk(child)
+
+    walk(node)
+    return pairs
+
+
+def _split_title(title: str) -> tuple[str | None, str | None]:
+    parts = title.split(None, 1)
+    if not parts:
+        return None, None
+    if len(parts) == 1:
+        return parts[0], None
+    return parts[0], parts[1]
+
+
+def _parse_price(text: str | None) -> str | None:
+    if not text:
+        return None
+    match = PRICE_RE.search(text.replace("\xa0", " "))
+    if match is None:
+        return None
+    whole = match.group(1).replace(".", "")
+    return f"{whole}.{match.group(2)}"
+
+
+def _seller_count(text: str) -> int | None:
+    match = OFFER_RE.search(text)
+    if match is None:
+        return None
+    return int(match.group(1))
+
+
+def _availability(node: Any | None) -> str:
+    if node is None:
+        return "unknown"
+    code = AVAILABILITY_BY_CODE.get(node.attrs.get("data-availability", ""))
+    if code:
+        return code
+    label = (node.attrs.get("aria-label") or node_text(node)).strip().casefold()
+    return AVAILABILITY_BY_LABEL.get(label, "unknown")
+
+
+def _normalize_attributes(raw: dict[str, str]) -> dict[str, Any]:
+    normalized: dict[str, Any] = {}
+    for label, value in raw.items():
+        key = label.casefold()
+        capacity = _capacity_gb(value)
+        if capacity is not None and key in {"ram", "arbeitsspeicher"}:
+            normalized["ram_gb"] = capacity
+        elif capacity is not None and key in {"speicher", "ssd", "festplatte"}:
+            normalized["storage_gb"] = capacity
+        weight = _weight_kg(value)
+        if weight is not None and "gewicht" in key:
+            normalized["weight_kg"] = weight
+    return normalized
+
+
+def _capacity_gb(text: str) -> float | int | None:
+    match = CAPACITY_RE.search(text)
+    if match is None:
+        return None
+    value = float(match.group(1).replace(",", "."))
+    if match.group(2).upper() == "TB":
+        value *= 1024
+    return int(value) if value.is_integer() else value
+
+
+def _weight_kg(text: str) -> float | None:
+    match = WEIGHT_RE.search(text)
+    if match is None:
+        return None
+    return float(match.group(1).replace(",", "."))
+
+
+def _page_count(root: Any) -> int:
+    pages = {1}
+
+    def walk(node: Any) -> None:
+        href = node.attrs.get("href") or ""
+        if "pg=" in href and ("pagination" in node.attrs.get("class", "") or _in_pagination(node)):
+            for name, value in parse_qsl(urlsplit(href).query):
+                if name == "pg" and value.isdigit():
+                    pages.add(int(value))
+        for child in node.children:
+            if not isinstance(child, str):
+                walk(child)
+
+    walk(root)
+    return max(pages)
+
+
+def _in_pagination(node: Any) -> bool:
+    current = node
+    while current is not None:
+        classes = current.attrs.get("class", "")
+        ident = current.attrs.get("id", "")
+        if "pagination" in classes or ident == "pagination":
+            return True
+        current = current.parent
+    return False
+
+
+def _result_count(root: Any) -> int | None:
+    counts = [
+        int(match.replace(".", ""))
+        for match in PRODUCT_COUNT_RE.findall(node_text(root))
+    ]
+    return max(counts) if counts else None
diff --git a/backend/app/sources/geizhals/taxonomy_parser.py b/backend/app/sources/geizhals/taxonomy_parser.py
new file mode 100644
index 0000000..6c91659
--- /dev/null
+++ b/backend/app/sources/geizhals/taxonomy_parser.py
@@ -0,0 +1,474 @@
+from __future__ import annotations
+
+import re
+from html.parser import HTMLParser
+from typing import Any
+from urllib.parse import parse_qsl, unquote, urlparse
+
+from app.discovery.enums import FilterType
+
+SKIP_OPTION_LABELS = {
+    "alle",
+    "entfernen",
+    "überspringen",
+    "aktualisieren",
+    "abonnieren",
+    "filter zurücksetzen",
+    "nur beliebteste filter anzeigen",
+    "immer alle filter anzeigen",
+    "filter überspringen",
+    "einfache filter",
+    "erweiterte filter",
+}
+RANGE_MIN_NAMES = {"bpmin", "minp", "prmin"}
+RANGE_MAX_NAMES = {"bpmax", "maxp", "prmax"}
+TEXT_NAMES = {"asuch", "fs"}
+HEADING_TAGS = {"th", "h2", "h3", "h4", "h5", "legend", "dt", "caption", "label", "strong"}
+NUMBER_RE = re.compile(r"(-?\d+(?:[.,]\d+)?)")
+THRESHOLD_RE = re.compile(r"^(ab|bis)\s+", re.IGNORECASE)
+UNIT_RE = re.compile(r"(GB|TB|GHz|MHz|Hz|cd/m²|ppi|kg|\"|″|mm|%|€|EUR)$", re.IGNORECASE)
+TYPE_MAP = {
+    "text": FilterType.ENUM.value,
+    "color": FilterType.ENUM.value,
+    "text_range": FilterType.NUMERIC_THRESHOLD.value,
+    "numeric_range": FilterType.NUMERIC_RANGE.value,
+}
+
+
+class _Node:
+    def __init__(self, tag: str, attrs: dict[str, str]) -> None:
+        self.tag = tag
+        self.attrs = attrs
+        self.children: list[_Node | str] = []
+        self.parent: _Node | None = None
+
+
+class _TreeBuilder(HTMLParser):
+    VOID = {
+        "area",
+        "base",
+        "br",
+        "col",
+        "embed",
+        "hr",
+        "img",
+        "input",
+        "link",
+        "meta",
+        "param",
+        "source",
+        "track",
+        "wbr",
+    }
+
+    def __init__(self) -> None:
+        super().__init__(convert_charrefs=True)
+        self.root = _Node("document", {})
+        self._stack = [self.root]
+
+    def handle_starttag(self, tag: str, attrs: list[tuple[str, str | None]]) -> None:
+        node = _Node(tag, {key: value or "" for key, value in attrs})
+        parent = self._stack[-1]
+        node.parent = parent
+        parent.children.append(node)
+        if tag not in self.VOID:
+            self._stack.append(node)
+
+    def handle_endtag(self, tag: str) -> None:
+        for index in range(len(self._stack) - 1, 0, -1):
+            if self._stack[index].tag == tag:
+                del self._stack[index:]
+                break
+
+    def handle_data(self, data: str) -> None:
+        text = data.strip()
+        if text:
+            self._stack[-1].children.append(text)
+
+
+def parse_html_tree(html: str) -> _Node:
+    builder = _TreeBuilder()
+    builder.feed(html)
+    return builder.root
+
+
+def node_text(node: _Node) -> str:
+    return _node_text(node)
+
+
+def parse_taxonomy_html(html: str, *, category_ref: str, source_url: str) -> dict[str, Any]:
+    if is_challenge_html(html):
+        return {
+            "source": "geizhals",
+            "external_category_key": category_ref,
+            "name": category_ref,
+            "source_url": source_url,
+            "filters": [],
+            "blocked": True,
+        }
+    builder = _TreeBuilder()
+    builder.feed(html)
+    groups: dict[str, dict[str, Any]] = {}
+    _collect_declared_filters(builder.root, groups)
+    _collect_xf_links(builder.root, groups)
+    _collect_named_inputs(builder.root, groups)
+    filters = [
+        group
+        for group in groups.values()
+        if _usable_group(group)
+    ]
+    filters.sort(key=lambda item: (item["label"].lower(), item["external_filter_key"]))
+    return {
+        "source": "geizhals",
+        "external_category_key": category_ref,
+        "name": _category_name(builder.root, category_ref),
+        "canonical_category": "notebook" if category_ref == "nb" else None,
+        "source_url": source_url,
+        "filters": filters,
+        "blocked": False,
+    }
+
+
+def is_challenge_html(html: str) -> bool:
+    sample = html.lower()
+    markers = (
+        "cf-mitigated",
+        "cf-challenge",
+        "just a moment",
+        "enable javascript and cookies",
+        "checking your browser",
+        "verification successful. waiting for geizhals",
+    )
+    return any(marker in sample for marker in markers)
+
+
+def _collect_declared_filters(node: _Node, groups: dict[str, dict[str, Any]]) -> None:
+    filter_id = node.attrs.get("data-filter-id")
+    if filter_id:
+        source_type = node.attrs.get("data-filter-type") or ""
+        label = _declared_label(node) or f"Filter {filter_id}"
+        group = groups.setdefault(filter_id, _empty_group(filter_id, label, source_type))
+        if label and group["label"].startswith("Filter "):
+            group["label"] = label
+        if source_type:
+            group["filter_type"] = TYPE_MAP.get(source_type, group["filter_type"])
+            if source_type == "numeric_range":
+                group["encoding"]["encoding_type"] = "RANGE_TOKEN"
+                group["encoding"]["value_template"] = "{min}*{max}"
+                group.setdefault("range", {"minimum": None, "maximum": None, "step": None, "unit": group.get("canonical_unit")})
+        _collect_xf_links(node, groups, required_key=filter_id, allow_histogram=source_type == "numeric_range")
+        return
+    for child in node.children:
+        if isinstance(child, _Node):
+            _collect_declared_filters(child, groups)
+
+
+def _collect_xf_links(
+    node: _Node,
+    groups: dict[str, dict[str, Any]],
+    required_key: str | None = None,
+    allow_histogram: bool = False,
+) -> None:
+    if node.tag == "a":
+        href = node.attrs.get("href") or ""
+        label = node.attrs.get("aria-label") or _node_text(node)
+        if _usable_option(node, label, allow_histogram=allow_histogram):
+            for key, query_value in _xf_tokens(href):
+                if required_key is not None and key != required_key:
+                    continue
+                if required_key is None and not _has_structural_label(node):
+                    continue
+                group_label = _declared_label(_filter_root(node)) or _nearest_group_label(node)
+                group = groups.setdefault(key, _empty_group(key, group_label, None))
+                if group_label and (group["label"].startswith("Filter ") or group["label"] == key):
+                    group["label"] = group_label
+                option = _option_from_link(label, query_value)
+                existing = {item["source_value"] for item in group["options"]}
+                if option["source_value"] not in existing:
+                    group["options"].append(option)
+                    group["encoding"].setdefault("options", {})
+                    group["encoding"]["options"][option["source_value"]] = query_value
+                    if option.get("normalized_value") is not None:
+                        group["encoding"]["options"][str(option["normalized_value"])] = query_value
+                if group["filter_type"] not in {FilterType.NUMERIC_RANGE.value, FilterType.NUMERIC_THRESHOLD.value}:
+                    group["filter_type"] = _classify_options(group["options"])
+                if group["canonical_unit"] is None:
+                    group["canonical_unit"] = option.get("unit")
+    for child in node.children:
+        if isinstance(child, _Node):
+            _collect_xf_links(child, groups, required_key=required_key, allow_histogram=allow_histogram)
+
+
+def _collect_named_inputs(node: _Node, groups: dict[str, dict[str, Any]]) -> None:
+    inputs: list[_Node] = []
+    _find_inputs(node, inputs)
+    names = {item.attrs.get("name", ""): item for item in inputs if item.attrs.get("name")}
+    if names.keys() & RANGE_MIN_NAMES or names.keys() & RANGE_MAX_NAMES:
+        min_name = next((name for name in RANGE_MIN_NAMES if name in names), "bpmin")
+        max_name = next((name for name in RANGE_MAX_NAMES if name in names), "bpmax")
+        label = _nearest_group_label(names.get(min_name) or names.get(max_name) or node) or "Preis"
+        groups["price"] = {
+            "external_filter_key": "price",
+            "label": "Preis" if "preis" not in label.lower() else label,
+            "filter_type": FilterType.NUMERIC_RANGE.value,
+            "canonical_unit": "EUR",
+            "options": [],
+            "range": {"minimum": 0, "maximum": None, "step": 1, "unit": "EUR"},
+            "encoding": {
+                "url_param": min_name,
+                "min_parameter": min_name,
+                "max_parameter": max_name,
+                "encoding_type": "RANGE_PARAMETERS",
+            },
+        }
+    for name in TEXT_NAMES:
+        if name not in names:
+            continue
+        label = _nearest_group_label(names[name]) or "Suchbegriffe"
+        groups[name] = {
+            "external_filter_key": name,
+            "label": label,
+            "filter_type": FilterType.TEXT.value,
+            "canonical_unit": None,
+            "options": [],
+            "encoding": {"url_param": name, "encoding_type": "TEXT_PARAMETER", "value_template": "{value}"},
+        }
+
+
+def _find_inputs(node: _Node, found: list[_Node]) -> None:
+    if node.tag == "input":
+        found.append(node)
+    for child in node.children:
+        if isinstance(child, _Node):
+            _find_inputs(child, found)
+
+
+def _xf_tokens(href: str) -> list[tuple[str, str]]:
+    if not href or href.startswith("#") or href.startswith("javascript:"):
+        return []
+    query = urlparse(href).query
+    tokens: list[tuple[str, str]] = []
+    for name, value in parse_qsl(query, keep_blank_values=True):
+        if name != "xf" or not value:
+            continue
+        for part in unquote(value).replace("|", "~").split("~"):
+            if "_" not in part:
+                continue
+            key, raw = part.split("_", 1)
+            if key.isdigit() and raw:
+                tokens.append((key, raw.replace("+", " ")))
+    return tokens
+
+
+def _option_from_link(label: str, query_value: str) -> dict[str, Any]:
+    cleaned = re.sub(r"\(\d[\d.]*\s*(produkte)?\)\s*$", "", label, flags=re.IGNORECASE).strip()
+    cleaned = re.sub(r"\s+", " ", cleaned)
+    operator = None
+    if cleaned.lower().startswith("ab "):
+        operator = ">="
+    elif cleaned.lower().startswith("bis "):
+        operator = "<="
+    number = _first_number(cleaned)
+    unit = _unit(cleaned)
+    return {
+        "source_value": cleaned or query_value,
+        "query_value": query_value,
+        "normalized_value": number,
+        "operator": operator,
+        "unit": unit,
+    }
+
+
+def _classify_options(options: list[dict[str, Any]]) -> str:
+    if not options:
+        return FilterType.UNKNOWN.value
+    labels = [str(item["source_value"]).lower() for item in options]
+    if all(item in {"ja", "nein", "egal", "true", "false"} for item in labels):
+        return FilterType.BOOLEAN.value
+    if all(THRESHOLD_RE.match(str(item["source_value"])) for item in options):
+        return FilterType.NUMERIC_THRESHOLD.value
+    return FilterType.ENUM.value
+
+
+def _empty_group(key: str, label: str | None, source_type: str | None) -> dict[str, Any]:
+    encoding_type = "OPTION_TOKEN"
+    if source_type == "numeric_range":
+        encoding_type = "RANGE_TOKEN"
+    return {
+        "external_filter_key": key,
+        "label": label or f"Filter {key}",
+        "filter_type": TYPE_MAP.get(source_type or "", FilterType.ENUM.value),
+        "canonical_unit": None,
+        "options": [],
+        "encoding": {
+            "url_param": "xf",
+            "encoding_type": encoding_type,
+            "options": {},
+            **({"value_template": "{min}*{max}"} if encoding_type == "RANGE_TOKEN" else {}),
+        },
+    }
+
+
+def _declared_label(node: _Node | None) -> str | None:
+    if node is None:
+        return None
+    attr_id = node.attrs.get("id", "")
+    if attr_id.startswith("filter-element__title-label"):
+        return _clean_label(_node_text(node)) or None
+    for child in node.children:
+        if isinstance(child, _Node):
+            found = _declared_label(child)
+            if found:
+                return found
+    return None
+
+
+def _filter_root(node: _Node) -> _Node | None:
+    current: _Node | None = node
+    while current is not None:
+        if current.attrs.get("data-filter-id"):
+            return current
+        current = current.parent
+    return None
+
+
+def _nearest_group_label(node: _Node | None) -> str | None:
+    current = node
+    while current is not None:
+        declared = _declared_label(current)
+        if declared:
+            return declared
+        if current.tag in HEADING_TAGS:
+            text = _clean_label(_node_text(current))
+            if text:
+                return text
+        previous = _previous_element(current)
+        while previous is not None:
+            if previous.tag in HEADING_TAGS:
+                text = _clean_label(_node_text(previous))
+                if text:
+                    return text
+            previous = _previous_element(previous)
+        labeled = current.attrs.get("aria-label") or current.attrs.get("data-filter-name") or current.attrs.get("title")
+        if labeled:
+            text = _clean_label(labeled)
+            if text and len(text.split()) <= 6:
+                return text
+        current = current.parent
+    return None
+
+
+def _has_structural_label(node: _Node) -> bool:
+    if _filter_root(node) is not None:
+        return True
+    current: _Node | None = node
+    while current is not None:
+        if current.tag in {"th", "td", "tr", "table", "dt", "dd"}:
+            return _nearest_group_label(node) is not None
+        current = current.parent
+    return False
+
+
+def _usable_option(node: _Node, label: str, *, allow_histogram: bool) -> bool:
+    classes = f"{node.attrs.get('class', '')} {node.attrs.get('id', '')}".lower()
+    if "skip-link" in classes or "add-to-columns" in classes:
+        return False
+    if "bucket" in classes and not allow_histogram:
+        return False
+    cleaned = re.sub(r"\s+", " ", label).strip()
+    if not cleaned or _skip_option(cleaned):
+        return False
+    if "überspringen" in cleaned.lower():
+        return False
+    if not allow_histogram and "produkte" in cleaned.lower():
+        return False
+    if len(cleaned.split()) > 8:
+        return False
+    return True
+
+
+def _usable_group(group: dict[str, Any]) -> bool:
+    if group["filter_type"] == FilterType.TEXT.value:
+        return True
+    if group.get("range") and group["filter_type"] == FilterType.NUMERIC_RANGE.value:
+        return True
+    if not group.get("options") and not group.get("range"):
+        return False
+    label = str(group.get("label") or "")
+    if label.startswith("Filter ") and len(group.get("options") or []) < 2:
+        return False
+    if len(label.split()) > 8:
+        return False
+    return True
+
+
+def _previous_element(node: _Node) -> _Node | None:
+    if node.parent is None:
+        return None
+    siblings = [child for child in node.parent.children if isinstance(child, _Node)]
+    try:
+        index = siblings.index(node)
+    except ValueError:
+        return None
+    return siblings[index - 1] if index > 0 else None
+
+
+def _node_text(node: _Node) -> str:
+    parts: list[str] = []
+    for child in node.children:
+        if isinstance(child, str):
+            parts.append(child)
+        else:
+            parts.append(_node_text(child))
+    return " ".join(part for part in parts if part).strip()
+
+
+def _clean_label(text: str) -> str:
+    cleaned = re.sub(r"\s+", " ", text).strip().rstrip(":")
+    cleaned = re.sub(r"\s*(überspringen|entfernen)$", "", cleaned, flags=re.IGNORECASE).strip()
+    if not cleaned or _skip_option(cleaned):
+        return ""
+    return cleaned
+
+
+def _skip_option(label: str) -> bool:
+    return label.strip().lower() in SKIP_OPTION_LABELS
+
+
+def _first_number(text: str) -> float | int | None:
+    match = NUMBER_RE.search(text.replace(",", "."))
+    if match is None:
+        return None
+    raw = match.group(1)
+    value = float(raw)
+    return int(value) if value.is_integer() else value
+
+
+def _unit(text: str) -> str | None:
+    match = UNIT_RE.search(text.replace(" ", ""))
+    if match is None:
+        if "kg" in text.lower():
+            return "kg"
+        return None
+    unit = match.group(1)
+    if unit in {'"', "″"}:
+        return "inch"
+    return unit
+
+
+def _category_name(root: _Node, fallback: str) -> str:
+    titles: list[str] = []
+
+    def walk(node: _Node) -> None:
+        if node.tag in {"h1", "title"}:
+            text = _node_text(node)
+            if text:
+                titles.append(text)
+        for child in node.children:
+            if isinstance(child, _Node):
+                walk(child)
+
+    walk(root)
+    if titles:
+        return titles[0].split("»")[0].strip() or fallback
+    return "Notebooks" if fallback == "nb" else fallback
diff --git a/backend/fixtures/geizhals/nb-category.html b/backend/fixtures/geizhals/nb-category.html
new file mode 100644
index 0000000..c7cc3bc
--- /dev/null
+++ b/backend/fixtures/geizhals/nb-category.html
@@ -0,0 +1,132 @@
+<!doctype html>
+<html lang="de">
+<head>
+  <title>Notebooks Preisvergleich Geizhals Deutschland</title>
+</head>
+<body>
+  <h1>Notebooks</h1>
+  <div data-filter-type="text_range" data-filter-id="17023">
+    <span id="filter-element__title-label--17023">RAM</span>
+    <a href="?cat=nb&xf=17023_ab+16GB&promode=true">ab 16GB</a>
+    <a href="?cat=nb&xf=17023_ab+32GB&promode=true">ab 32GB</a>
+  </div>
+  <form>
+    <label>Suchbegriffe</label>
+    <input name="asuch" type="text">
+    <label>Preisbereich (€)</label>
+    <input name="bpmin" type="number" placeholder="Preis ab">
+    <input name="bpmax" type="number" placeholder="Preis bis">
+  </form>
+  <table>
+    <tr>
+      <th>Hersteller:</th>
+      <td>
+        <a href="?cat=nb&xf=1_Apple">Apple</a>
+        <a href="?cat=nb&xf=1_ASUS">ASUS</a>
+        <a href="?cat=nb&xf=1_Lenovo">Lenovo</a>
+      </td>
+    </tr>
+    <tr>
+      <th>Display-Größe:</th>
+      <td>
+        <a href="?cat=nb&xf=12128_13">13.x"</a>
+        <a href="?cat=nb&xf=12128_14">14.x"</a>
+        <a href="?cat=nb&xf=12128_15">15.x"</a>
+        <a href="?cat=nb&xf=12128_16">16.x"</a>
+      </td>
+    </tr>
+    <tr>
+      <th>Display-Auflösung:</th>
+      <td>
+        <a href="?cat=nb&xf=9_1920x1080">1920x1080 (Full HD)</a>
+        <a href="?cat=nb&xf=9_2560x1600">2560x1600 (WQXGA)</a>
+      </td>
+    </tr>
+    <tr>
+      <th>CPU-Kerne:</th>
+      <td>
+        <a href="?cat=nb&xf=12822_8">8 (Octa-Core)</a>
+        <a href="?cat=nb&xf=12822_10">10</a>
+        <a href="?cat=nb&xf=12822_14">14-Core</a>
+      </td>
+    </tr>
+    <tr>
+      <th>CPU-Kerne ab:</th>
+      <td>
+        <a href="?cat=nb&xf=12822_8">ab 8</a>
+        <a href="?cat=nb&xf=12822_10">ab 10</a>
+      </td>
+    </tr>
+    <tr>
+      <th>CPU-Threads ab:</th>
+      <td>
+        <a href="?cat=nb&xf=12823_12">ab 12</a>
+        <a href="?cat=nb&xf=12823_16">ab 16</a>
+      </td>
+    </tr>
+    <tr>
+      <th>CPU-Turbotakt:</th>
+      <td>
+        <a href="?cat=nb&xf=9682_4500">ab 4.5GHz</a>
+        <a href="?cat=nb&xf=9682_5000">ab 5GHz</a>
+      </td>
+    </tr>
+    <tr>
+      <th>RAM ab:</th>
+      <td>
+        <a href="?cat=nb&xf=12_16384">ab 16GB</a>
+        <a href="?cat=nb&xf=12_32768">ab 32GB</a>
+        <a href="?cat=nb&xf=12_65536">ab 64GB</a>
+      </td>
+    </tr>
+    <tr>
+      <th>RAM-Typ:</th>
+      <td>
+        <a href="?cat=nb&xf=7520_DDR5">DDR5</a>
+        <a href="?cat=nb&xf=7520_LPDDR5">LPDDR5</a>
+      </td>
+    </tr>
+    <tr>
+      <th>Speichertyp:</th>
+      <td>
+        <a href="?cat=nb&xf=82_SSD">SSD</a>
+        <a href="?cat=nb&xf=82_M.2">M.2</a>
+      </td>
+    </tr>
+    <tr>
+      <th>SSD-Kapazität:</th>
+      <td>
+        <a href="?cat=nb&xf=2991_1000">ab 1TB</a>
+        <a href="?cat=nb&xf=2991_2000">ab 2TB</a>
+      </td>
+    </tr>
+    <tr>
+      <th>Grafik-Typ:</th>
+      <td>
+        <a href="?cat=nb&xf=8149_zzx">dedizierte Grafik (dGPU)</a>
+      </td>
+    </tr>
+    <tr>
+      <th>GPU - Speichergröße ab:</th>
+      <td>
+        <a href="?cat=nb&xf=7436_2048">ab 2GB</a>
+        <a href="?cat=nb&xf=7436_4096">ab 4GB</a>
+        <a href="?cat=nb&xf=7436_8192">ab 8GB</a>
+      </td>
+    </tr>
+    <tr>
+      <th>GPU - Speichergröße:</th>
+      <td>
+        <a href="?cat=nb&xf=16692_8192">8GB</a>
+        <a href="?cat=nb&xf=16692_12288">12GB</a>
+      </td>
+    </tr>
+    <tr>
+      <th>USB-C mit DisplayPort gesamt (inkl. Thunderbolt):</th>
+      <td>
+        <a href="?cat=nb&xf=13731_1">ab 1x</a>
+      </td>
+    </tr>
+  </table>
+</body>
+</html>
diff --git a/backend/fixtures/geizhals/nb-search-page.html b/backend/fixtures/geizhals/nb-search-page.html
new file mode 100644
index 0000000..4642e74
--- /dev/null
+++ b/backend/fixtures/geizhals/nb-search-page.html
@@ -0,0 +1,52 @@
+<!doctype html>
+<html lang="de">
+<body>
+  <a href="#productlist">6431 Produkte</a>
+  <nav class="pagination">
+    <a class="pagination__page" href="?cat=nb&amp;pg=2">Seite 2</a>
+    <a class="pagination__page last-page" href="?cat=nb&amp;pg=4">Seite 4</a>
+  </nav>
+  <div class="galleryview__item card">
+    <h3 class="galleryview__name">
+      <a class="galleryview__name-link" href="acme-notebook-14-a1000001.html" title="Acme Notebook 14, 32GB RAM, 1TB SSD">
+        Acme Notebook 14, 32GB RAM, 1TB SSD
+      </a>
+    </h3>
+    <span class="galleryview__mpn">ACM-14-32</span>
+    <dl>
+      <dt>Betriebssystem</dt><dd>Linux</dd>
+      <dt>Display</dt><dd>14"</dd>
+      <dt>CPU</dt><dd>Acme 8-Core</dd>
+      <dt>RAM</dt><dd>32GB</dd>
+      <dt>Speicher</dt><dd>1TB SSD</dd>
+      <dt>Grafik</dt><dd>Acme GPU</dd>
+    </dl>
+    <div class="galleryview__price">
+      <a class="galleryview__price-link" href="acme-notebook-14-a1000001.html#offerlist">
+        <span class="price">€ 1298,00</span>
+      </a>
+      <span class="availability" data-availability="2" aria-label="lagernd"></span>
+    </div>
+    <a class="galleryview__offercount-link">8 Angebote</a>
+  </div>
+  <div class="galleryview__item card">
+    <h3 class="galleryview__name">
+      <a class="galleryview__name-link" href="acme-notebook-13-a1000002.html" title="Acme Notebook 13, 8GB RAM">
+        Acme Notebook 13, 8GB RAM
+      </a>
+    </h3>
+    <span class="galleryview__mpn">ACM-13-8</span>
+    <dl>
+      <dt>RAM</dt><dd>8GB</dd>
+      <dt>Speicher</dt><dd>512GB SSD</dd>
+    </dl>
+    <div class="galleryview__price">
+      <a class="galleryview__price-link" href="acme-notebook-13-a1000002.html#offerlist">
+        <span class="price">€ 899,00</span>
+      </a>
+      <span class="availability" data-availability="1" aria-label="kurzfristig lieferbar"></span>
+    </div>
+    <a class="galleryview__offercount-link">3 Angebote</a>
+  </div>
+</body>
+</html>
diff --git a/backend/pyproject.toml b/backend/pyproject.toml
index 8e9046c..e346213 100644
--- a/backend/pyproject.toml
+++ b/backend/pyproject.toml
@@ -6,6 +6,7 @@ requires-python = ">=3.13"
 dependencies = [
     "alembic>=1.16",
     "fastapi>=0.118",
+    "httpx>=0.28",
     "psycopg[binary]>=3.2",
     "pydantic>=2.11",
     "pydantic-settings>=2.11",
@@ -18,6 +19,9 @@ test = [
     "httpx>=0.28",
     "pytest>=8.4",
 ]
+live = [
+    "playwright>=1.48",
+]
 
 [build-system]
 requires = ["setuptools>=75"]
diff --git a/backend/tests/conftest.py b/backend/tests/conftest.py
index 80f8fa7..714493e 100644
--- a/backend/tests/conftest.py
+++ b/backend/tests/conftest.py
@@ -1,4 +1,5 @@
 import json
+import os
 from datetime import datetime, timezone
 from decimal import Decimal
 from pathlib import Path
@@ -6,6 +7,8 @@ from uuid import UUID, uuid4
 
 import pytest
 
+os.environ.setdefault("GEIZHALS_USE_FIXTURES", "1")
+
 from app.domain.enums import Availability, NormalizationStatus, OfferCondition, ValueType
 from app.domain.models import (
     AttributeDefinition,
@@ -17,7 +20,17 @@ from app.domain.models import (
 )
 from app.domain.validation import parse_profile, parse_scoring_model
 
+from app.settings import get_settings
+
 ROOT = Path(__file__).resolve().parents[1]
+
+
+@pytest.fixture(autouse=True)
+def _geizhals_fixtures(monkeypatch):
+    monkeypatch.setenv("GEIZHALS_USE_FIXTURES", "1")
+    get_settings.cache_clear()
+    yield
+    get_settings.cache_clear()
 AS_OF = datetime(2026, 10, 3, 12, 0, tzinfo=timezone.utc)
 CUTOFF = datetime(2026, 10, 3, 12, 0, tzinfo=timezone.utc)
 
diff --git a/backend/tests/unit/test_discovery_urlgen.py b/backend/tests/unit/test_discovery_urlgen.py
index af70bee..9483fb8 100644
--- a/backend/tests/unit/test_discovery_urlgen.py
+++ b/backend/tests/unit/test_discovery_urlgen.py
@@ -68,6 +68,33 @@ def test_broken_mapping_is_not_used():
     assert "ram_gb" in internal
 
 
+def test_multiple_xf_filters_join_with_tilde():
+    url, _ = generate_search_url(
+        category_url="https://geizhals.de/?cat=nb",
+        criteria_items=[
+            {"criterion": "ram_gb", "operator": ">=", "value": 16},
+            {"criterion": "gpu", "operator": ">=", "value": 8},
+        ],
+        mappings=[
+            {
+                "criterion_key": "ram_gb",
+                "external_filter_key": "12",
+                "source_value": "16384",
+                "encoding": {"url_param": "xf"},
+                "status": "VERIFIED",
+            },
+            {
+                "criterion_key": "gpu",
+                "external_filter_key": "7436",
+                "source_value": "8192",
+                "encoding": {"url_param": "xf"},
+                "status": "VERIFIED",
+            },
+        ],
+    )
+    assert "xf=12_16384~7436_8192" in url or "xf=7436_8192~12_16384" in url
+
+
 def test_roundtrip_verifies_xf_and_maxp():
     assert verify_roundtrip(
         category_url="https://geizhals.de/?cat=nb",
diff --git a/backend/tests/unit/test_geizhals_result_parser.py b/backend/tests/unit/test_geizhals_result_parser.py
new file mode 100644
index 0000000..f9ebde1
--- /dev/null
+++ b/backend/tests/unit/test_geizhals_result_parser.py
@@ -0,0 +1,42 @@
+from pathlib import Path
+
+from app.sources.geizhals.result_parser import parse_result_html, with_page
+
+HTML = (Path(__file__).resolve().parents[2] / "fixtures" / "geizhals" / "nb-search-page.html").read_text(
+    encoding="utf-8"
+)
+
+
+def test_parser_reads_observed_gallery_cards() -> None:
+    payload = parse_result_html(HTML, page_url="https://geizhals.de/?cat=nb")
+    assert payload["blocked"] is False
+    assert payload["page_count"] == 4
+    assert payload["result_count"] == 6431
+    assert [item["source_external_id"] for item in payload["items"]] == ["1000001", "1000002"]
+    first = payload["items"][0]
+    assert first["source_url"] == "https://geizhals.de/acme-notebook-14-a1000001.html"
+    assert first["source_title"].startswith("Acme Notebook 14")
+    assert first["manufacturer"] == "Acme"
+    assert first["manufacturer_sku"] == "ACM-14-32"
+    assert first["best_price"] == "1298.00"
+    assert first["currency"] == "EUR"
+    assert first["availability"] == "in_stock"
+    assert first["seller_count"] == 8
+    assert first["raw_attributes"]["RAM"] == "32GB"
+    assert first["raw_attributes"]["Betriebssystem"] == "Linux"
+    assert first["normalized_attributes"]["ram_gb"] == 32
+    assert first["normalized_attributes"]["storage_gb"] == 1024
+
+
+def test_challenge_html_is_not_turned_into_candidates() -> None:
+    payload = parse_result_html(
+        "<html><body>Enable JavaScript and cookies to continue</body></html>",
+        page_url="https://geizhals.de/?cat=nb",
+    )
+    assert payload["blocked"] is True
+    assert payload["items"] == []
+
+
+def test_with_page_uses_geizhals_pg_parameter() -> None:
+    assert with_page("https://geizhals.de/?cat=nb&xf=17023_ab+16GB", 1) == "https://geizhals.de/?cat=nb&xf=17023_ab+16GB"
+    assert "pg=3" in with_page("https://geizhals.de/?cat=nb&xf=17023_ab+16GB", 3)
diff --git a/backend/tests/unit/test_geizhals_taxonomy_parser.py b/backend/tests/unit/test_geizhals_taxonomy_parser.py
new file mode 100644
index 0000000..1561736
--- /dev/null
+++ b/backend/tests/unit/test_geizhals_taxonomy_parser.py
@@ -0,0 +1,76 @@
+from pathlib import Path
+
+from app.discovery.urlgen import generate_search_url
+from app.sources.geizhals.taxonomy_parser import parse_taxonomy_html
+
+HTML = (Path(__file__).resolve().parents[2] / "fixtures" / "geizhals" / "nb-category.html").read_text(
+    encoding="utf-8"
+)
+
+
+def test_parser_reads_href_filters_and_price_inputs() -> None:
+    payload = parse_taxonomy_html(
+        HTML,
+        category_ref="nb",
+        source_url="https://geizhals.de/?cat=nb",
+    )
+    keys = {item["external_filter_key"] for item in payload["filters"]}
+    assert {"12", "17023", "price", "12128", "7436", "82", "asuch"} <= keys
+    assert len(payload["filters"]) >= 12
+    live_ram = next(item for item in payload["filters"] if item["external_filter_key"] == "17023")
+    assert live_ram["label"] == "RAM"
+    ram = next(item for item in payload["filters"] if item["external_filter_key"] == "12")
+    assert ram["label"] == "RAM ab"
+    assert ram["filter_type"] == "NUMERIC_THRESHOLD"
+    assert any(option["source_value"] == "ab 16GB" for option in ram["options"])
+    assert ram["encoding"]["options"]["ab 16GB"] == "16384"
+    price = next(item for item in payload["filters"] if item["external_filter_key"] == "price")
+    assert price["filter_type"] == "NUMERIC_RANGE"
+    assert price["encoding"]["min_parameter"] == "bpmin"
+    assert price["encoding"]["max_parameter"] == "bpmax"
+
+
+def test_learned_ram_and_price_encode_observed_geizhals_query() -> None:
+    payload = parse_taxonomy_html(
+        HTML,
+        category_ref="nb",
+        source_url="https://geizhals.de/?cat=nb",
+    )
+    ram = next(item for item in payload["filters"] if item["external_filter_key"] == "12")
+    price = next(item for item in payload["filters"] if item["external_filter_key"] == "price")
+    url, internal = generate_search_url(
+        category_url="https://geizhals.de/?cat=nb",
+        criteria_items=[
+            {"criterion": "geizhals:nb:12", "operator": ">=", "value": 16},
+            {"criterion": "geizhals:nb:price", "operator": "<=", "value": 1700},
+        ],
+        mappings=[
+            {
+                "criterion_key": "geizhals:nb:12",
+                "external_filter_key": "12",
+                "source_value": "16384",
+                "encoding": ram["encoding"],
+                "status": "VERIFIED",
+            },
+            {
+                "criterion_key": "geizhals:nb:price",
+                "external_filter_key": "price",
+                "source_value": None,
+                "encoding": price["encoding"],
+                "status": "VERIFIED",
+            },
+        ],
+    )
+    assert internal == []
+    assert "xf=12_16384" in url
+    assert "bpmax=1700" in url
+
+
+def test_challenge_html_is_not_treated_as_taxonomy() -> None:
+    payload = parse_taxonomy_html(
+        "<html><body>Enable JavaScript and cookies to continue</body></html>",
+        category_ref="nb",
+        source_url="https://geizhals.de/?cat=nb",
+    )
+    assert payload["blocked"] is True
+    assert payload["filters"] == []
diff --git a/backend/tests/unit/test_settings.py b/backend/tests/unit/test_settings.py
index a64eb3c..0adc6ab 100644
--- a/backend/tests/unit/test_settings.py
+++ b/backend/tests/unit/test_settings.py
@@ -12,3 +12,9 @@ def test_database_url_uses_psycopg_and_explicit_fields() -> None:
     assert settings.database_url == (
         "postgresql+psycopg://buyingagent:secret@postgres:5432/buyingagent_test"
     )
+
+
+def test_geizhals_live_defaults_are_off_for_fixtures_flag() -> None:
+    settings = Settings(geizhals_use_fixtures=True, geizhals_playwright=False)
+    assert settings.geizhals_use_fixtures is True
+    assert settings.geizhals_playwright is False
diff --git a/backend/uv.lock b/backend/uv.lock
index c9fb7d9..ebffc96 100644
--- a/backend/uv.lock
+++ b/backend/uv.lock
@@ -54,6 +54,7 @@ source = { editable = "." }
 dependencies = [
     { name = "alembic" },
     { name = "fastapi" },
+    { name = "httpx" },
     { name = "psycopg", extra = ["binary"] },
     { name = "pydantic" },
     { name = "pydantic-settings" },
@@ -62,6 +63,9 @@ dependencies = [
 ]
 
 [package.optional-dependencies]
+live = [
+    { name = "playwright" },
+]
 test = [
     { name = "httpx" },
     { name = "pytest" },
@@ -71,7 +75,9 @@ test = [
 requires-dist = [
     { name = "alembic", specifier = ">=1.16" },
     { name = "fastapi", specifier = ">=0.118" },
+    { name = "httpx", specifier = ">=0.28" },
     { name = "httpx", marker = "extra == 'test'", specifier = ">=0.28" },
+    { name = "playwright", marker = "extra == 'live'", specifier = ">=1.48" },
     { name = "psycopg", extras = ["binary"], specifier = ">=3.2" },
     { name = "pydantic", specifier = ">=2.11" },
     { name = "pydantic-settings", specifier = ">=2.11" },
@@ -79,7 +85,7 @@ requires-dist = [
     { name = "sqlalchemy", specifier = ">=2.0" },
     { name = "uvicorn", extras = ["standard"], specifier = ">=0.37" },
 ]
-provides-extras = ["test"]
+provides-extras = ["test", "live"]
 
 [[package]]
 name = "certifi"
@@ -125,6 +131,63 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/a0/b6/78aaf9141fb46742928c113f3cf6ef2259d538cb02b604b7656c1dc9883c/fastapi-0.142.2-py3-none-any.whl", hash = "sha256:bd5f4d81f1e93a88bcd77caf4dfe3c2dbffc3805407a0007e9a114c18b3a670b", size = 144409, upload-time = "2026-09-30T09:51:09.8Z" },
 ]
 
+[[package]]
+name = "greenlet"
+version = "3.5.6"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/3e/6e/0091f175ccd02b02bc8811bbcbcc6ac2e980be116e3b2f7a736ca322bf84/greenlet-3.5.6.tar.gz", hash = "sha256:8e67c43bdfc88d5fee6db0d3e40175b362fc95fb85f0412d233b9b203c53a575", size = 207653, upload-time = "2026-09-14T15:42:51.806Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/f1/a1/e720a38852366c589e1a46cf570b886507ad2cf591050c203365638baab0/greenlet-3.5.6-cp313-cp313-macosx_11_0_universal2.whl", hash = "sha256:f96f0e30b5a95c7631b12bfe214cbc90ec8fe8cfa36920596c10514a65743519", size = 294627, upload-time = "2026-09-14T14:24:40.102Z" },
+    { url = "https://files.pythonhosted.org/packages/eb/c3/58187858df41354a11e6a55b421e7af9059798abdab3a384cc51b8567c38/greenlet-3.5.6-cp313-cp313-manylinux_2_24_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:c75116c9de79949de23006e2d9b35ee82874c594fcf5c0311b439acaa14b8441", size = 614356, upload-time = "2026-09-14T15:12:03.399Z" },
+    { url = "https://files.pythonhosted.org/packages/ce/b9/3a7e67d5f05c9760b1ad411fa52264bd69cc08e22a2ebfb4018b90628ced/greenlet-3.5.6-cp313-cp313-manylinux_2_24_ppc64le.manylinux_2_28_ppc64le.whl", hash = "sha256:cad5782f93f7f738b62c6527b6f32a60694d924029f299a8b524758cfa53d815", size = 626756, upload-time = "2026-09-14T15:20:44.269Z" },
+    { url = "https://files.pythonhosted.org/packages/c6/7c/40400455f5b5a65bb83e94fde66d1be9e5ec518638113f8083ace746c309/greenlet-3.5.6-cp313-cp313-manylinux_2_24_s390x.manylinux_2_28_s390x.whl", hash = "sha256:a93ee7c6e8fd0f8a83525a51bd777be57ee17787e91d805bd8d6faf9dcada18e", size = 632632, upload-time = "2026-09-14T15:25:07.813Z" },
+    { url = "https://files.pythonhosted.org/packages/85/cb/ab0c123c514ed4e94c0dc9ee2e86362633e6b998cfc05de7fc9ac2eb9690/greenlet-3.5.6-cp313-cp313-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:f98e8215e172f567ce80eeaed9107fb4d32b6c44f26983d9b8334658136a205a", size = 623779, upload-time = "2026-09-14T14:36:01.104Z" },
+    { url = "https://files.pythonhosted.org/packages/f9/67/1f35cff30a6c51c3f23b63d4afcc7313ab4f97490ba3676fa78178984b27/greenlet-3.5.6-cp313-cp313-manylinux_2_39_riscv64.whl", hash = "sha256:7f731ebac68ea06d628658295cb2d217b10186329fcf9a3b6a149045059bf92e", size = 434933, upload-time = "2026-09-14T15:28:38.858Z" },
+    { url = "https://files.pythonhosted.org/packages/a5/26/fda8a5a06e7073333ccb038133c5893b9e0c4fe29d5992a17e83c241bc6e/greenlet-3.5.6-cp313-cp313-musllinux_1_2_aarch64.whl", hash = "sha256:df19e2d0b1620039af5102563fbd96e8938c7f5c3f5828528d641d9fc585525e", size = 1584930, upload-time = "2026-09-14T15:10:08.234Z" },
+    { url = "https://files.pythonhosted.org/packages/2f/37/50f8813163148d6234e08b23dcad6a9e37f01d148c8ec976e4c44ea2d918/greenlet-3.5.6-cp313-cp313-musllinux_1_2_x86_64.whl", hash = "sha256:06c0e933290fba8ffe53ead4ae1b8044b0e9754b75cebf381aa2bc3e50d82fac", size = 1647590, upload-time = "2026-09-14T14:35:51.173Z" },
+    { url = "https://files.pythonhosted.org/packages/86/da/b7669b09586365654083a62bd0724cf06cb74bd5085a15cdd161271f992f/greenlet-3.5.6-cp313-cp313-win_amd64.whl", hash = "sha256:5b602b4201b965a8354d74e232364a66ff243dd142e350d035f46169bb36e13d", size = 324086, upload-time = "2026-09-14T14:23:48.428Z" },
+    { url = "https://files.pythonhosted.org/packages/e5/5d/c9663cfe84a2a9e0aa96f066f5b0594c227ea4c647511e087e2e11d4ac0a/greenlet-3.5.6-cp313-cp313-win_arm64.whl", hash = "sha256:876077e7ebb8c84ed068e2b23d4c62ebb010d60df84b9591af1be2f39010ffb2", size = 308211, upload-time = "2026-09-14T14:28:01.634Z" },
+    { url = "https://files.pythonhosted.org/packages/66/c0/d254544ae2b8bdd311aef000fafc02828c2771b17d994b3075620ea7cc6e/greenlet-3.5.6-cp314-cp314-macosx_11_0_universal2.whl", hash = "sha256:8cddea1b8339451c2fb3388e138347b6126744f33b611bdb55b7357361cfef46", size = 295221, upload-time = "2026-09-14T14:25:11.583Z" },
+    { url = "https://files.pythonhosted.org/packages/18/18/eb54be16b9cc3971e09ca5b73334e1b8c804a4630d9addaaf218a4fe300f/greenlet-3.5.6-cp314-cp314-manylinux_2_24_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:c59acfa8eb73a1e0d484392dc002bdf001fd4ce73394e0132df3d1ab6093d7cb", size = 660992, upload-time = "2026-09-14T15:12:04.876Z" },
+    { url = "https://files.pythonhosted.org/packages/8f/b4/e193efe65671dcf294bc51fcc59efb52d154adf8612c4ea016da0d2c486c/greenlet-3.5.6-cp314-cp314-manylinux_2_24_ppc64le.manylinux_2_28_ppc64le.whl", hash = "sha256:a3b4a01c6da07ef9f80d4fe8933b994bc99747bcea3eab0330a9c34d3c12655b", size = 673428, upload-time = "2026-09-14T15:20:45.756Z" },
+    { url = "https://files.pythonhosted.org/packages/fd/21/631bb45fafde1dca782152377c0676d182ec924820064047f533a3627b28/greenlet-3.5.6-cp314-cp314-manylinux_2_24_s390x.manylinux_2_28_s390x.whl", hash = "sha256:dd0b83bed3405b586a3133629f1d1a5bc7bfd64822a3b7ab342bdc68e6dbc61b", size = 677688, upload-time = "2026-09-14T15:25:09.279Z" },
+    { url = "https://files.pythonhosted.org/packages/45/ac/28fa7a9e50f2859466214c4ac584d776db52c1604ad4dd158960a5af2a1f/greenlet-3.5.6-cp314-cp314-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:9a09d59bef1db94f384b5bcc2d523694d338f3df6b757aeeaf7baca5d0c0be88", size = 670773, upload-time = "2026-09-14T14:36:02.577Z" },
+    { url = "https://files.pythonhosted.org/packages/40/30/2b0a73e68e1e18e30b601d0d183cfdfc2beca4de5a6843c630f0fc9fb90c/greenlet-3.5.6-cp314-cp314-manylinux_2_39_riscv64.whl", hash = "sha256:fdacf26402389bdd89857ad3c045a26fe8f3314f9a8b28226f82f88463a65b77", size = 480475, upload-time = "2026-09-14T15:28:40.741Z" },
+    { url = "https://files.pythonhosted.org/packages/c3/cd/fb7d6cdd86ff3427c1494854f0e35437eba05142be91f530f6da75e09e19/greenlet-3.5.6-cp314-cp314-musllinux_1_2_aarch64.whl", hash = "sha256:8b7c73d1cef3d9ae963e9ff03f6222df43efbb9054ffd2f1969c935b7fc84c02", size = 1631900, upload-time = "2026-09-14T15:10:09.745Z" },
+    { url = "https://files.pythonhosted.org/packages/f6/40/143bdbb20a516628cb15074ae52ed17d850b450292609c7a6fccac6dbece/greenlet-3.5.6-cp314-cp314-musllinux_1_2_x86_64.whl", hash = "sha256:8b27df301f56e3b3d2298095c8f7d6b68f2521f6b1693e901fa039bdbae34424", size = 1693740, upload-time = "2026-09-14T14:35:52.959Z" },
+    { url = "https://files.pythonhosted.org/packages/c9/9e/019642432e6ae283301df1361227d47610709d2dc69a38f95edef266d713/greenlet-3.5.6-cp314-cp314-win_amd64.whl", hash = "sha256:f8f0bd690e1a41294ac87905e8121c81a3761ec2583c768f13467428606c8c7a", size = 327473, upload-time = "2026-09-14T14:28:12.948Z" },
+    { url = "https://files.pythonhosted.org/packages/e9/7f/8aafc7bf70c948786dba7221d0dc0838e5329bebc6d434ef2208b4f0e760/greenlet-3.5.6-cp314-cp314-win_arm64.whl", hash = "sha256:8cda13494d86a4f12429641117cb6ac4bbbc9c30a33f711f7d3a2e5fbe4b0b7e", size = 311095, upload-time = "2026-09-14T14:28:00.7Z" },
+    { url = "https://files.pythonhosted.org/packages/14/7e/7a205688a5b3074933b18a906608d46d106e9a79d776bdab5a4abf4b4feb/greenlet-3.5.6-cp314-cp314t-macosx_11_0_universal2.whl", hash = "sha256:97c5a53e8c1754df58e73f047a99e287d4da1bdfe64b0072fb25c87000897951", size = 305352, upload-time = "2026-09-14T14:21:31.962Z" },
+    { url = "https://files.pythonhosted.org/packages/78/cb/9c4a57a9d9dd0256e20b8f7f4f06554c2c92badebf0ab73ce344321b78b9/greenlet-3.5.6-cp314-cp314t-manylinux_2_24_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:fea4427d1ffdb3b523d7daa6712038428a4c16c450b9777bdd1221cfee0eab49", size = 672671, upload-time = "2026-09-14T15:12:06.347Z" },
+    { url = "https://files.pythonhosted.org/packages/97/52/c6729681ebbd298f4decd28746815acc8a0b0a0fde21d2df33776fd4d042/greenlet-3.5.6-cp314-cp314t-manylinux_2_24_ppc64le.manylinux_2_28_ppc64le.whl", hash = "sha256:73a29b5ba642e35433166a03a3e02935e7238c4b3467fbd77523b99edea23e5b", size = 679489, upload-time = "2026-09-14T15:20:47.291Z" },
+    { url = "https://files.pythonhosted.org/packages/71/76/3c11c21e0716b1f1dc7c1a4b3d690abb1d3b448c69a9d32049fecb64010a/greenlet-3.5.6-cp314-cp314t-manylinux_2_24_s390x.manylinux_2_28_s390x.whl", hash = "sha256:61a61b4a95a4f97922c3a6f5606d3e360851584bd47e500a5161373c53810e3d", size = 681303, upload-time = "2026-09-14T15:25:11.088Z" },
+    { url = "https://files.pythonhosted.org/packages/58/c5/2b6c721ba8b8963da42d5a0f57f25b8aaeb1fe9bdd156875e57f3be648a2/greenlet-3.5.6-cp314-cp314t-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:460e70b033aba8ed47e2ac9b5d0d2157b05a34fbfa30a241400aef4118902cdc", size = 676608, upload-time = "2026-09-14T14:36:03.959Z" },
+    { url = "https://files.pythonhosted.org/packages/3f/26/3ae402202452cd5941bbbd483e5a74297e2397e7aa3182c2a5e3ab7d5666/greenlet-3.5.6-cp314-cp314t-manylinux_2_39_riscv64.whl", hash = "sha256:fe3170a69fe039b18ad18171e66faa9a75f6fe9d78f968fd9b54e09fbd714d81", size = 510112, upload-time = "2026-09-14T15:28:42.112Z" },
+    { url = "https://files.pythonhosted.org/packages/b2/04/0d018e0d05bcdde19a0fcb907834155f1fc853a9bedd3f3f5e6acadcae19/greenlet-3.5.6-cp314-cp314t-musllinux_1_2_aarch64.whl", hash = "sha256:ca80a49b53ed1d22f7282da7255f7bb2fd1935fd0f623d8613fda38745f18961", size = 1641479, upload-time = "2026-09-14T15:10:11.216Z" },
+    { url = "https://files.pythonhosted.org/packages/59/bb/f02ef9073919158f6403fe3701d4ed4403d646720e7201dfc6e9d264bac3/greenlet-3.5.6-cp314-cp314t-musllinux_1_2_x86_64.whl", hash = "sha256:916f92f2a8db10508f739d0b5e00b83defe5d1115a997c54532a6d7cf8c95404", size = 1698758, upload-time = "2026-09-14T14:35:54.336Z" },
+    { url = "https://files.pythonhosted.org/packages/08/a5/1f48fe647473a2dcccfd1839b2ff2c78eb57009be776b4da071e901c9bff/greenlet-3.5.6-cp314-cp314t-win_amd64.whl", hash = "sha256:886bcf1870af74c32bc310fd00a6b803445e17e51b7d5a107c7b35c0f362cc16", size = 331574, upload-time = "2026-09-14T14:27:18.451Z" },
+    { url = "https://files.pythonhosted.org/packages/cd/72/3882855a75838faeb54a58aeef4fd77d20b2a86d4bad570c70d41b565dcf/greenlet-3.5.6-cp315-cp315-macosx_11_0_universal2.whl", hash = "sha256:3ac3494c381dab876cad7d0b22f3a722f3e0c8deb3a65b9e7f35ad7f58b8fcb3", size = 295926, upload-time = "2026-09-14T14:27:21.16Z" },
+    { url = "https://files.pythonhosted.org/packages/10/1f/be4d957d8a9b90bcbe8db206548a42134d96222d43e5ed3fc4708fb6e24b/greenlet-3.5.6-cp315-cp315-manylinux_2_24_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:602024dae6d77e161f4b89491b62ca1d4f19949d79d47b2db057e476d21179d6", size = 666910, upload-time = "2026-09-14T15:12:07.901Z" },
+    { url = "https://files.pythonhosted.org/packages/a1/af/60d62571a7d6de961e4ce7625d6c2faf359345659fc782d2cdf517c34577/greenlet-3.5.6-cp315-cp315-manylinux_2_24_ppc64le.manylinux_2_28_ppc64le.whl", hash = "sha256:f8e63209c3e1e828ee6a457529b4a6d8b05d050fe0ae03a7ae49e967c5d312e0", size = 677415, upload-time = "2026-09-14T15:20:48.817Z" },
+    { url = "https://files.pythonhosted.org/packages/f5/41/b3114c97c10e796010f00a30f51c81470072bca4b53e396ccca87484fcf7/greenlet-3.5.6-cp315-cp315-manylinux_2_24_s390x.manylinux_2_28_s390x.whl", hash = "sha256:9133d68624b1f2e89ec2f554d56aea8a5b0d7168cd9320200ba58d4d794845a4", size = 681337, upload-time = "2026-09-14T15:25:12.812Z" },
+    { url = "https://files.pythonhosted.org/packages/fb/16/ac9e547b611539aaed1870eb1d6ddc57abdd5924b3a99bb9b5f0b44176b8/greenlet-3.5.6-cp315-cp315-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:ccadce0130fd813ec86ebfe969a6c58b42acc1d0fe55a47525375b740e07b605", size = 675927, upload-time = "2026-09-14T14:36:05.34Z" },
+    { url = "https://files.pythonhosted.org/packages/48/1b/d41861c2fa00968e39e467a495ca8db9ce9b6310a5d9b57561b3d0dc48fa/greenlet-3.5.6-cp315-cp315-manylinux_2_39_riscv64.whl", hash = "sha256:5adcbbfe78bdc242c71740a02e0991cc1b2f34d33c8bb15ca45eee8fd1140942", size = 487339, upload-time = "2026-09-14T15:28:43.497Z" },
+    { url = "https://files.pythonhosted.org/packages/c4/b1/b7ba08d6431121741f1d30be0d5d292e76873325179a63586cd9217b62f6/greenlet-3.5.6-cp315-cp315-musllinux_1_2_aarch64.whl", hash = "sha256:9297fb9c39b9a2c039dbcd306c410bd6906b95244dec3bba4318d36c718c164c", size = 1637236, upload-time = "2026-09-14T15:10:12.442Z" },
+    { url = "https://files.pythonhosted.org/packages/af/c5/3b1cbc68f0c082022fc8717f7fe4b8b13b8d583c52352be37f4e9f55bcd2/greenlet-3.5.6-cp315-cp315-musllinux_1_2_x86_64.whl", hash = "sha256:b374e79ffa7511afc11773aef40a4ccea6191fba1c856ea2f9c56738dca69d7a", size = 1698526, upload-time = "2026-09-14T14:35:56.039Z" },
+    { url = "https://files.pythonhosted.org/packages/de/56/12941ed2711400451c89d544e10f831800a2770f19dd55eac8f0f7f2003b/greenlet-3.5.6-cp315-cp315-win_amd64.whl", hash = "sha256:7969bffa322c097bd46ae595ada6a931cefda613f18ba64587e9cff4cb320756", size = 327739, upload-time = "2026-09-14T14:23:55.768Z" },
+    { url = "https://files.pythonhosted.org/packages/c5/3b/576b9ed5ac929252e340cf60b4bcb6a8515350dc20797064b1922dc4ea75/greenlet-3.5.6-cp315-cp315-win_arm64.whl", hash = "sha256:8dba0129b93e7091dfefaf4cf7000172741bff7f47bf6326fcf17f32fbb54d6b", size = 311715, upload-time = "2026-09-14T14:28:25.154Z" },
+    { url = "https://files.pythonhosted.org/packages/16/c2/86cfc5555a98e12b86966ddbd24fd39af32f71f2f785c6595b7feb2db156/greenlet-3.5.6-cp315-cp315t-macosx_11_0_universal2.whl", hash = "sha256:de3de000d459402cda015068fd135aa50c0bf6f2477a80d4da1e646f123b4e78", size = 306244, upload-time = "2026-09-14T14:27:57.565Z" },
+    { url = "https://files.pythonhosted.org/packages/14/6d/83ffc9d05a75a80ab3a7595dbb1d9604e5d4fc2996d73a8ae2dbd1284900/greenlet-3.5.6-cp315-cp315t-manylinux_2_24_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:45663c01a4de48b9a64a2ee1509d92d1dfd3afb02b2ccfc9333029d11aef996a", size = 676578, upload-time = "2026-09-14T15:12:09.468Z" },
+    { url = "https://files.pythonhosted.org/packages/5d/d6/c2cf684810e5caded075970aaadea654ecb58b8382b9aecf1d231b936894/greenlet-3.5.6-cp315-cp315t-manylinux_2_24_ppc64le.manylinux_2_28_ppc64le.whl", hash = "sha256:3deccbb57a481e3a408fe61cdfd5c13e0678fc0a30fdd09597917ca87b4be877", size = 684203, upload-time = "2026-09-14T15:20:50.261Z" },
+    { url = "https://files.pythonhosted.org/packages/f2/d1/039c353d5593a97a89699e989324c9bc86af499e6c6152fe0180f5742204/greenlet-3.5.6-cp315-cp315t-manylinux_2_24_s390x.manylinux_2_28_s390x.whl", hash = "sha256:63aff70fe5aac59c72215f42ec39fcb59ff46774fa966e717f8ecb6ee2273577", size = 686023, upload-time = "2026-09-14T15:25:14.528Z" },
+    { url = "https://files.pythonhosted.org/packages/62/19/00e1bee5d2af890dc8f400b54d0b0f9b489965f92bc12b407ff72cc6f469/greenlet-3.5.6-cp315-cp315t-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:311018b46472fb26ee85870847fb89eb64cc8aaddb617400789d87076f7cfeec", size = 681253, upload-time = "2026-09-14T14:36:06.742Z" },
+    { url = "https://files.pythonhosted.org/packages/8a/62/97ceb8e0b2ea96046cdf8e95b042715020ebb12d83ea0690db80a8f03d23/greenlet-3.5.6-cp315-cp315t-manylinux_2_39_riscv64.whl", hash = "sha256:520648db8fb92eef7b3e6013f5a6f901cdf0d6685f639c2f7a245879f865bef7", size = 517058, upload-time = "2026-09-14T15:28:44.924Z" },
+    { url = "https://files.pythonhosted.org/packages/89/58/c9275fd0ca195d1d3402931bcce8cfcc74726ff76efb1883d229e6e1a3d7/greenlet-3.5.6-cp315-cp315t-musllinux_1_2_aarch64.whl", hash = "sha256:7f924a5a9d5890649566f2f6682e0d8ad8ca23028bacffbbac36dbd7fd680176", size = 1645988, upload-time = "2026-09-14T15:10:13.758Z" },
+    { url = "https://files.pythonhosted.org/packages/e0/36/b35747582fa4f1a5453f8f3002405dbac788e450cec7674dc2d204b6ccb5/greenlet-3.5.6-cp315-cp315t-musllinux_1_2_x86_64.whl", hash = "sha256:de9923832f2d8c1a5ecd8d7260465a6ca5a86888a0d129e3bd5cf0406d2fc5bf", size = 1702383, upload-time = "2026-09-14T14:35:58.143Z" },
+    { url = "https://files.pythonhosted.org/packages/ed/69/6ec22ac9351e474d2a134d0ff9400dc80362d1c20f0721088ffffdfc205b/greenlet-3.5.6-cp315-cp315t-win_amd64.whl", hash = "sha256:2ab5f42ac6c238eb71770715e6e909ad9a1a92b6c681ccb64cd5a0f07edb953f", size = 331845, upload-time = "2026-09-14T14:27:41.723Z" },
+    { url = "https://files.pythonhosted.org/packages/30/cf/697c051fd534e223461fb8b523890e21a24eeca229cd50624cff6f02fabd/greenlet-3.5.6-cp315-cp315t-win_arm64.whl", hash = "sha256:f9fe868463ec7e1363733af77e38a5fda3e9b63940337048c945d69e0c80ff24", size = 315070, upload-time = "2026-09-14T14:22:21.476Z" },
+]
+
 [[package]]
 name = "h11"
 version = "0.16.0"
@@ -340,6 +403,25 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/63/34/ba1c580383c9eada3711951fef0795c80b829a078d72188184bcab9dd527/packaging-26.3-py3-none-any.whl", hash = "sha256:d7193f7c8e4e93f444fde0262bf90af30e16fa0ad0ad44cb553c87339b23cd1c", size = 129956, upload-time = "2026-08-04T18:15:27.159Z" },
 ]
 
+[[package]]
+name = "playwright"
+version = "1.63.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "greenlet" },
+    { name = "pyee" },
+]
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/19/dd/fbb3d34228ad753bc14464d5f2252585b73367539775a88d0a3949cf5a45/playwright-1.63.0-py3-none-macosx_10_13_x86_64.whl", hash = "sha256:84c540759e8e7f01e690e197e04060f48899d37cea322299f255843273d3385d", size = 44278407, upload-time = "2026-09-15T16:49:06.045Z" },
+    { url = "https://files.pythonhosted.org/packages/f5/9a/948b930b1a8c4ee869e5a139a2b7747caa06aab56a3f09a2f0abdcbda221/playwright-1.63.0-py3-none-macosx_11_0_arm64.whl", hash = "sha256:fd1aa00631d44d55e56e0975bf3f3f285fac4a9fd2813183f0c12a488f1a1b24", size = 42944679, upload-time = "2026-09-15T16:49:10.039Z" },
+    { url = "https://files.pythonhosted.org/packages/94/11/dc5c13fa1602371603acd461be47529c1b3513815d3a0dc98f642c291a10/playwright-1.63.0-py3-none-macosx_11_0_universal2.whl", hash = "sha256:c89fc4736502a1f0fac2c8ca5d10c0cbc1c669f1f4774a2d8507a43140e4d53f", size = 44278410, upload-time = "2026-09-15T16:49:13.861Z" },
+    { url = "https://files.pythonhosted.org/packages/27/9c/103a5037789062bdab27c7dca53f3ca6b075b572ab2cd96eec825b3aec4e/playwright-1.63.0-py3-none-manylinux1_x86_64.whl", hash = "sha256:ad21bc07516b187965a7521c5cf0df0bd657b17482eaad74335272d35a2b07de", size = 48217159, upload-time = "2026-09-15T16:49:17.404Z" },
+    { url = "https://files.pythonhosted.org/packages/f3/82/3d85505284c5a210f2da6c07b8f757524e79d1fba9cfdafe1eafb766ae59/playwright-1.63.0-py3-none-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:354e15b29503565fc598b89f16fbe070459343bef9d7498a93e304864000c6a7", size = 47902708, upload-time = "2026-09-15T16:49:21.055Z" },
+    { url = "https://files.pythonhosted.org/packages/69/8d/f74ff6b52751859f4b69caf66aa7d4a3c8d6c1f0c7dc9920da103612ee39/playwright-1.63.0-py3-none-win32.whl", hash = "sha256:660c00c62639e31b16700ba5456b351ddba55bb766b7ce8261223aa34928e482", size = 38606075, upload-time = "2026-09-15T16:49:29.546Z" },
+    { url = "https://files.pythonhosted.org/packages/76/eb/d6b8d92658038e260dbc7dd69fb3fdbe245aac283953de8b50b02bfe5f61/playwright-1.63.0-py3-none-win_amd64.whl", hash = "sha256:2f9a707a6c6c91157ed77bff2b8caeb04b3c8d46e70d585fc134298cbe4b5cc6", size = 38606082, upload-time = "2026-09-15T16:49:32.838Z" },
+    { url = "https://files.pythonhosted.org/packages/be/75/2432b3e3c7c62103b72d4c4cc8b16a56383ada372bbb0d1278591e988f71/playwright-1.63.0-py3-none-win_arm64.whl", hash = "sha256:1e4a3a838ce22fb68ad17193fcd142a19610d9d70fb9966d2239f5dddc0cc05b", size = 34523554, upload-time = "2026-09-15T16:49:36.455Z" },
+]
+
 [[package]]
 name = "pluggy"
 version = "1.6.0"
@@ -491,6 +573,18 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/30/a4/2bffa9f8e804325a09867f0e9d30795c80ea9f8d62560bd1b6ad6220eb2f/pydantic_settings-2.15.0-py3-none-any.whl", hash = "sha256:0ba092c291c94baceb5eff768aa0d56400a457585bc0175925a5a5510303da42", size = 69413, upload-time = "2026-08-07T09:24:55.839Z" },
 ]
 
+[[package]]
+name = "pyee"
+version = "13.0.1"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "typing-extensions" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/8b/04/e7c1fe4dc78a6fdbfd6c337b1c3732ff543b8a397683ab38378447baa331/pyee-13.0.1.tar.gz", hash = "sha256:0b931f7c14535667ed4c7e0d531716368715e860b988770fc7eb8578d1f67fc8", size = 31655, upload-time = "2026-02-14T21:12:28.044Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/a0/c4/b4d4827c93ef43c01f599ef31453ccc1c132b353284fc6c87d535c233129/pyee-13.0.1-py3-none-any.whl", hash = "sha256:af2f8fede4171ef667dfded53f96e2ed0d6e6bd7ee3bb46437f77e3b57689228", size = 15659, upload-time = "2026-02-14T21:12:26.263Z" },
+]
+
 [[package]]
 name = "pygments"
 version = "2.21.0"
diff --git a/docker-compose.dev-env.yml b/docker-compose.dev-env.yml
index f3a6a1b..1294f2a 100644
--- a/docker-compose.dev-env.yml
+++ b/docker-compose.dev-env.yml
@@ -33,6 +33,8 @@ services:
       ENVIRONMENT: "${ENVIRONMENT:-development}"
       ENGINE_BUILD: "${ENGINE_BUILD:-dev}"
       CORS_ALLOWED_ORIGINS: "${CORS_ALLOWED_ORIGINS:-http://localhost:5173,http://127.0.0.1:5173,http://192.168.2.49:5174}"
+      GEIZHALS_USE_FIXTURES: "${GEIZHALS_USE_FIXTURES:-0}"
+      GEIZHALS_PLAYWRIGHT: "${GEIZHALS_PLAYWRIGHT:-1}"
     ports:
       - "${BUYING_AGENT_API_PORT:-8095}:8000"
     depends_on:
@@ -84,6 +86,7 @@ services:
       ENVIRONMENT: test
       ENGINE_BUILD: test
       CLOCK_NOW: "2026-10-03T12:00:00Z"
+      GEIZHALS_USE_FIXTURES: "1"
     depends_on:
       postgres:
         condition: service_healthy
diff --git a/docker-compose.test.yml b/docker-compose.test.yml
index 28b7fb0..f6a12a0 100644
--- a/docker-compose.test.yml
+++ b/docker-compose.test.yml
@@ -34,6 +34,7 @@ services:
       ENVIRONMENT: test
       ENGINE_BUILD: test
       CLOCK_NOW: "2026-10-03T12:00:00Z"
+      GEIZHALS_USE_FIXTURES: "1"
     depends_on:
       postgres:
         condition: service_healthy
diff --git a/docker-compose.yml b/docker-compose.yml
index 23c4e52..50bb12e 100644
--- a/docker-compose.yml
+++ b/docker-compose.yml
@@ -36,6 +36,8 @@ services:
       ENVIRONMENT: "${ENVIRONMENT:-production}"
       ENGINE_BUILD: "${ENGINE_BUILD:-dev}"
       CORS_ALLOWED_ORIGINS: "${CORS_ALLOWED_ORIGINS:-http://localhost:4173,http://127.0.0.1:4173}"
+      GEIZHALS_USE_FIXTURES: "${GEIZHALS_USE_FIXTURES:-0}"
+      GEIZHALS_PLAYWRIGHT: "${GEIZHALS_PLAYWRIGHT:-1}"
     ports:
       - "${BUYING_AGENT_API_BIND:-127.0.0.1}:${BUYING_AGENT_API_PORT:-8000}:8000"
     depends_on:
diff --git a/frontend/nginx.conf b/frontend/nginx.conf
index ef13eb6..5882ece 100644
--- a/frontend/nginx.conf
+++ b/frontend/nginx.conf
@@ -8,6 +8,9 @@ server {
     proxy_pass http://api:8000;
     proxy_set_header Host $host;
     proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
+    proxy_connect_timeout 10s;
+    proxy_read_timeout 120s;
+    proxy_send_timeout 120s;
   }
 
   location /health {
diff --git a/frontend/src/api.ts b/frontend/src/api.ts
index c4d6b35..8dab3a7 100644
--- a/frontend/src/api.ts
+++ b/frontend/src/api.ts
@@ -7,7 +7,7 @@ async function request<T>(path: string, init?: RequestInit): Promise<T> {
   });
   const body = await response.json().catch(() => ({}));
   if (!response.ok) {
-    throw new Error(body.message || `HTTP ${response.status}`);
+    throw new Error(body?.error?.message || body?.message || `HTTP ${response.status}`);
   }
   return body as T;
 }
@@ -101,7 +101,12 @@ export type AvailableCriterion = {
   supports_discovery: boolean;
   scoring_attribute_key?: string | null;
 };
-export type AvailableCriteria = { product_category_id: string; source: string; criteria: AvailableCriterion[] };
+export type AvailableCriteria = {
+  product_category_id: string;
+  source: string;
+  taxonomy_status?: string | null;
+  criteria: AvailableCriterion[];
+};
 export type DiscoverySearch = {
   id: string;
   status: string;
diff --git a/frontend/src/pages/AdminPage.tsx b/frontend/src/pages/AdminPage.tsx
index 9e3b63a..275a0e6 100644
--- a/frontend/src/pages/AdminPage.tsx
+++ b/frontend/src/pages/AdminPage.tsx
@@ -33,9 +33,10 @@ export function AdminPage() {
         <p>Source: {source || "–"}</p>
         <p>Kategorie: {category || "–"}</p>
         <p>Status: {selected?.taxonomy_status || "NOT_LEARNED"}</p>
+        <p>Gelernte Filter: {filters.data?.filters.length ?? 0}</p>
         {discover.isError ? <p className="danger">{(discover.error as Error).message}</p> : null}
         <button className="button" disabled={!source || !category || discover.isPending} onClick={() => discover.mutate()}>
-          Taxonomie laden
+          {discover.isPending ? "Lernt Taxonomie …" : "Taxonomie neu lernen"}
         </button>
       </div>
       {!category ? (
diff --git a/frontend/src/pages/WizardPage.tsx b/frontend/src/pages/WizardPage.tsx
index f4da244..93bf698 100644
--- a/frontend/src/pages/WizardPage.tsx
+++ b/frontend/src/pages/WizardPage.tsx
@@ -18,6 +18,7 @@ export function WizardPage() {
   const [sourceOverride, setSourceOverride] = useState<string | null>(null);
   const [selected, setSelected] = useState<string[]>([]);
   const [draft, setDraft] = useState<Draft>({});
+  const [criterionQuery, setCriterionQuery] = useState("");
 
   const sources = useQuery({
     queryKey: ["category-sources", categoryId],
@@ -63,7 +64,12 @@ export function WizardPage() {
   const sourceRow = sourceItems.find((item) => item.source === source) || sourceItems[0];
   const needsPicker = sourceItems.filter((item) => item.status === "ACTIVE").length > 1;
 
-  const visible = useMemo(() => criteria.data?.criteria || [], [criteria.data]);
+  const visible = useMemo(() => {
+    const items = criteria.data?.criteria || [];
+    const query = criterionQuery.trim().toLowerCase();
+    if (!query) return items;
+    return items.filter((item) => item.label.toLowerCase().includes(query));
+  }, [criteria.data, criterionQuery]);
 
   if (categories.isLoading) return <LoadingState />;
   if (categories.isError) return <ErrorState message={(categories.error as Error).message} />;
@@ -84,6 +90,7 @@ export function WizardPage() {
               setCategoryId(event.target.value);
               setSourceOverride(null);
               setSelected([]);
+              setCriterionQuery("");
             }}
           >
             <option value="">Bitte wählen</option>
@@ -139,10 +146,21 @@ export function WizardPage() {
         <p className="muted">Quelle: {source}</p>
       ) : null}
 
+      {(criteria.data?.taxonomy_status || sourceRow?.taxonomy_status) === "DEGRADED" ? (
+        <p className="muted">Die Taxonomie stammt aus einer aufgezeichneten Kategorieseite. Live-Lernen war blockiert.</p>
+      ) : null}
       {criteria.isError ? <ErrorState message={(criteria.error as Error).message} onRetry={() => criteria.refetch()} /> : null}
       {criteria.isLoading ? <LoadingState /> : null}
-      {visible.length > 0 ? (
+      {(criteria.data?.criteria || []).length > 0 ? (
         <div className="card grid">
+          <p>
+            {(criteria.data?.criteria || []).length} Kriterien aus der Quelle. Wähle die relevanten aus und setze
+            Grenzwerte.
+          </p>
+          <label>
+            Kriterien suchen
+            <input value={criterionQuery} onChange={(event) => setCriterionQuery(event.target.value)} />
+          </label>
           {visible.map((item) => {
             const on = selected.includes(item.criterion_definition_id);
             return (