Découvert via attache CDP au vrai Chrome (contourne le blocage automation) :
- set_selection = PUT /gw/v1/carts/{week}, body {meals:[{index,quantity}], extras:[]}
sélection par index de course, params (customer/subscription/sku/cutoff) dérivés
dynamiquement de /subscriptions + /deliveries (aucun id en dur)
- Recipe.course_index conservé depuis le menu pour le mapping id->index
- get_editable_weeks via /deliveries (modèle Delivery: cutoff, status, editable)
- Token lu depuis le cookie apiV2Auth (storage_state) -> auth sans navigateur, headless OK
- hf_confirm_selection: garde-fou coco + dry_run; tool attach_capture.py ajouté
- Dry-run validé: requête identique à l'appel réel capturé
133 lines
4.9 KiB
Python
133 lines
4.9 KiB
Python
"""Attache Playwright à TON Chrome (via CDP) au lieu de lancer un navigateur piloté.
|
|
|
|
Pourquoi : un navigateur lancé par Playwright porte des drapeaux d'automatisation que
|
|
HelloFresh peut rejeter au login. Ici, c'est TOI qui lances Chrome (session normale, le
|
|
login marche), et on s'y attache en lecture pour :
|
|
- capturer le trafic gateway (dont l'appel d'écriture de sélection de recettes) ;
|
|
- exporter ta session (cookies) -> .session/storage_state.json, réutilisable en headless.
|
|
|
|
Prérequis — lance Chrome avec le port de debug et un PROFIL DÉDIÉ (Chrome 149 interdit le
|
|
debug sur le profil par défaut). Ta fenêtre Chrome habituelle peut rester ouverte :
|
|
|
|
"/Applications/Google Chrome.app/Contents/MacOS/Google Chrome" \\
|
|
--remote-debugging-port=9222 --user-data-dir="$HOME/.hf-chrome-debug" \\
|
|
https://www.hellofresh.fr/my-account/deliveries/menu
|
|
|
|
Puis, dans cette fenêtre : connecte-toi (email + mot de passe) et change une recette.
|
|
|
|
Usage :
|
|
python tools/attach_capture.py # attend ~5 min puis sauvegarde
|
|
ANTICOCO_DISCOVER_WAIT=600 python tools/attach_capture.py
|
|
"""
|
|
|
|
from __future__ import annotations
|
|
|
|
import json
|
|
import os
|
|
import sys
|
|
import time
|
|
from pathlib import Path
|
|
|
|
ROOT = Path(__file__).resolve().parent.parent
|
|
sys.path.insert(0, str(ROOT))
|
|
|
|
from playwright.sync_api import sync_playwright # noqa: E402
|
|
from hellofresh import auth # noqa: E402
|
|
|
|
CDP_URL = os.environ.get("ANTICOCO_CDP_URL", "http://localhost:9222")
|
|
LOG_PATH = auth.SESSION_DIR / "discovery_log_cdp.json"
|
|
STATE_PATH = auth.SESSION_DIR / "storage_state.json"
|
|
DISCOVERED = ROOT / "config" / "endpoints_discovered.json"
|
|
|
|
|
|
def _connect(pw, timeout_s: int = 60):
|
|
"""Connexion CDP avec attente que Chrome (port debug) soit disponible."""
|
|
deadline = time.time() + timeout_s
|
|
last = None
|
|
while time.time() < deadline:
|
|
try:
|
|
return pw.chromium.connect_over_cdp(CDP_URL)
|
|
except Exception as e: # Chrome pas encore prêt
|
|
last = e
|
|
print(f" ...en attente de Chrome sur {CDP_URL}")
|
|
time.sleep(3)
|
|
raise RuntimeError(f"Impossible de se connecter à {CDP_URL} : {last}")
|
|
|
|
|
|
def main() -> None:
|
|
auth.SESSION_DIR.mkdir(parents=True, exist_ok=True)
|
|
requests_seen: list[dict] = []
|
|
|
|
def on_request(req):
|
|
if not auth._is_gateway_request(req.url):
|
|
return
|
|
entry = {"method": req.method, "url": req.url,
|
|
"has_auth": bool(req.headers.get("authorization"))}
|
|
if req.method in ("POST", "PUT", "PATCH"):
|
|
try:
|
|
entry["post_data"] = req.post_data
|
|
except Exception:
|
|
entry["post_data"] = None
|
|
requests_seen.append(entry)
|
|
print(f" [{req.method}] {req.url}")
|
|
|
|
with sync_playwright() as pw:
|
|
print(f"Connexion à Chrome via {CDP_URL} ...")
|
|
browser = _connect(pw)
|
|
contexts = browser.contexts
|
|
print(f"Connecté. {len(contexts)} contexte(s).")
|
|
|
|
def wire_page(page):
|
|
try:
|
|
page.on("request", on_request)
|
|
except Exception:
|
|
pass
|
|
|
|
for ctx in contexts:
|
|
for p in ctx.pages:
|
|
wire_page(p)
|
|
ctx.on("page", wire_page)
|
|
|
|
wait_s = int(os.environ.get("ANTICOCO_DISCOVER_WAIT", "300"))
|
|
print(f">>> Connecte-toi et change une recette. Capture {wait_s}s max "
|
|
"(Ctrl-C pour finir plus tôt).")
|
|
try:
|
|
waited = 0
|
|
while waited < wait_s and browser.is_connected():
|
|
time.sleep(2)
|
|
waited += 2
|
|
except KeyboardInterrupt:
|
|
pass
|
|
|
|
# Export de la session (cookies) tant que la connexion tient.
|
|
try:
|
|
if browser.contexts:
|
|
browser.contexts[0].storage_state(path=str(STATE_PATH))
|
|
print(f"Session exportée -> {STATE_PATH}")
|
|
except Exception as e:
|
|
print("Export storage_state impossible:", e)
|
|
|
|
_save_report(requests_seen)
|
|
|
|
|
|
def _save_report(requests_seen: list[dict]) -> None:
|
|
LOG_PATH.write_text(json.dumps(requests_seen, indent=2, ensure_ascii=False), encoding="utf-8")
|
|
print(f"\n{len(requests_seen)} requêtes gateway -> {LOG_PATH}")
|
|
|
|
noise = ("otlp/traces", "/login", "auth/email", "translations", "/bot/")
|
|
writes = [r for r in requests_seen
|
|
if r["method"] in ("POST", "PUT", "PATCH")
|
|
and not any(n in r["url"] for n in noise)]
|
|
skeleton = {
|
|
"_comment": "Brouillon issu de attach_capture (ne remplace pas config/endpoints.json).",
|
|
"set_selection_candidates": [f"[{r['method']}] {r['url']}" for r in writes],
|
|
}
|
|
DISCOVERED.write_text(json.dumps(skeleton, indent=2, ensure_ascii=False), encoding="utf-8")
|
|
print(f"Candidats d'écriture ({len(writes)}) -> {DISCOVERED}")
|
|
for r in writes:
|
|
print(f" [{r['method']}] {r['url']}")
|
|
|
|
|
|
if __name__ == "__main__":
|
|
main()
|