Open-source AI browser agent for Chrome and Firefox (monorepo) ðŸ§
JavaScript ai-agentai-agentsai-toolsbrowser-aibrowser-automationbrowser-extensionchrome-aichrome-devtoolschrome-extensionchrome-extensionsfirefox-addonfirefox-addonsfirefox-extensionfirefox-extensionsgemmagemma4minimaxoffline-aioffline-llmqwen
#!/usr/bin/env python3
"""Regression suite for social-media-downloader.js
Attach order (auto-selected, best first):
1. CDP on port 9222 — run start-chrome-debug.bat (.sh) to launch a
SECOND Chrome with its own profile. Main Chrome stays untouched.
2. browser_cookie3 from default Chrome profile (Chrome 127+'s
App-Bound Encryption usually blocks this; falls through).
3. Dedicated test profile at ~/smd-test-profile (--setup once).
Usage:
python test_smd.py # all sites
python test_smd.py --site reddit # one site
python test_smd.py --headless # hide bundled browser (modes 2/3)
python test_smd.py --update-baseline # save current counts as baseline
python test_smd.py --js PATH # explicit SMD script
python test_smd.py --setup # fallback profile login
"""
from __future__ import annotations
import argparse, json, sys, time
from datetime import datetime
from pathlib import Path
from playwright.sync_api import sync_playwright
from common import TestResult
from sites import TESTS as _FEED_TESTS
from sites_video import TESTS as _VIDEO_TESTS
from sites_advanced import TESTS as _ADVANCED_TESTS
TESTS = _FEED_TESTS + _VIDEO_TESTS + _ADVANCED_TESTS
SCRIPT_DIR = Path(__file__).resolve().parent
PROFILE_DIR = Path.home() / "smd-test-profile"
REPORT_DIR = Path.home() / "smd-test-reports"
BASELINE_PATH = SCRIPT_DIR / "baseline.json"
CDP_PORT = 9222
COOKIE_DOMAINS = [
"facebook.com", "instagram.com", "x.com", "twitter.com",
"linkedin.com", "reddit.com", "pinterest.com",
]
def find_smd_js() -> Path:
for c in [SCRIPT_DIR / "social-media-downloader.js",
SCRIPT_DIR.parent / "social-media-downloader.js"]:
if c.exists():
return c
return SCRIPT_DIR / "social-media-downloader.js"
DEFAULT_JS = find_smd_js()
def _extract_chrome_cookies():
"""Try to read cookies from default Chrome profile. Returns None on failure."""
try:
import browser_cookie3 # type: ignore
except ImportError:
print("[!] browser_cookie3 not installed; skipping cookie-attach path.")
return None
all_cookies, abe_blocked = [], False
for d in COOKIE_DOMAINS:
try:
cj = browser_cookie3.chrome(domain_name=d)
for c in cj:
expires = float(c.expires) if c.expires else -1.0
rest = getattr(c, "_rest", {}) or {}
http_only = any(k.lower() == "httponly" for k in rest)
same_site = "Lax"
for k, v in rest.items():
if k.lower() == "samesite" and isinstance(v, str):
v2 = v.capitalize()
if v2 in ("Strict", "Lax", "None"):
same_site = v2
all_cookies.append({
"name": c.name, "value": c.value,
"domain": c.domain, "path": c.path or "/",
"expires": expires, "httpOnly": http_only,
"secure": bool(c.secure), "sameSite": same_site,
})
except Exception as e:
msg = str(e)
if "Unable to get key" in msg or "decryption" in msg.lower():
abe_blocked = True
break
print(f"[!] cookie extract for {d}: {e}")
if abe_blocked:
print("[!] Chrome 127+ App-Bound Encryption blocks external cookie")
print("[!] reading. Use the CDP path instead - it does NOT require")
print("[!] you to close your main Chrome:")
print("[!] start-chrome-debug.bat (Windows)")
print("[!] ./start-chrome-debug.sh (macOS / Linux)")
print("[!] That opens a SECOND Chrome with its own profile. Log into")
print("[!] the test sites in it once; every later run reuses them.")
return None
return all_cookies or None
def _trigger_download(page, site_name: str):
"""After a test passes, trigger SocialMediaDownloader.run({limit:1}) and
capture the resulting file via Playwright's download event. Returns the
saved file path or an error string."""
target_dir = REPORT_DIR / "downloads"
target_dir.mkdir(parents=True, exist_ok=True)
try:
with page.expect_download(timeout=20000) as dl_info:
page.evaluate(
"(prefix) => { SocialMediaDownloader.run({ limit: 1, prefix }); }",
site_name,
)
download = dl_info.value
suggested = download.suggested_filename or f"{site_name}.bin"
target = target_dir / f"{site_name}__{suggested}"
download.save_as(str(target))
return str(target)
except Exception as e:
return f"download failed: {e}"
def run_all(js_path, headless=False, site_filter=None, download=False):
REPORT_DIR.mkdir(exist_ok=True)
sdir = REPORT_DIR / "screenshots"
results = []
p = sync_playwright().start()
browser = None; ctx = None
attached_cdp = False; used_browser_close = False
test_page = None
try:
try:
browser = p.chromium.connect_over_cdp(
f"http://localhost:{CDP_PORT}", timeout=3000)
ctx = browser.contexts[0] if browser.contexts else browser.new_context()
attached_cdp = True
print(f"[+] Attached to Chrome via CDP on port {CDP_PORT}.")
except Exception:
pass
if ctx is None:
cookies = _extract_chrome_cookies()
if cookies:
print(f"[+] Extracted {len(cookies)} cookies from Chrome profile.")
browser = p.chromium.launch(
headless=headless,
args=["--disable-blink-features=AutomationControlled"])
ctx = browser.new_context(
viewport={"width": 1400, "height": 900},
bypass_csp=True)
try: ctx.add_cookies(cookies)
except Exception as e: print(f"[!] Some cookies rejected: {e}")
used_browser_close = True
if ctx is None:
print(f"[!] Falling back to dedicated test profile at {PROFILE_DIR}.")
print(f"[!] Run start-chrome-debug.bat for CDP, or --setup for this profile.")
PROFILE_DIR.mkdir(exist_ok=True)
ctx = p.chromium.launch_persistent_context(
user_data_dir=str(PROFILE_DIR), headless=headless,
viewport={"width": 1400, "height": 900},
bypass_csp=True,
args=["--disable-blink-features=AutomationControlled"])
test_page = ctx.new_page()
for fn in TESTS:
name = fn.__name__.replace("test_", "")
if site_filter and name != site_filter:
continue
print(f"[+] testing {name} ...", flush=True)
try:
res = fn(test_page, js_path, sdir)
except Exception as e:
res = TestResult(site=name, url=test_page.url, passed=False, error=str(e))
# Optional: trigger a real download to verify end-to-end.
if download and res.passed and res.url_count > 0:
dl_result = _trigger_download(test_page, name)
# Record into the test result so the report shows it.
if dl_result.startswith("download failed"):
res.notes = (res.notes + " | " if res.notes else "") + dl_result
else:
res.notes = (res.notes + " | " if res.notes else "") + f"downloaded: {dl_result}"
print(f" {res.status():12s} url_count={res.url_count} profile={res.profile_detected}")
results.append(res)
finally:
try:
if test_page is not None: test_page.close()
except Exception: pass
if attached_cdp:
try:
if browser is not None: browser.close()
except Exception: pass
elif used_browser_close:
try:
if ctx is not None: ctx.close()
except Exception: pass
try:
if browser is not None: browser.close()
except Exception: pass
else:
try:
if ctx is not None: ctx.close()
except Exception: pass
p.stop()
return results
def load_baseline():
return json.loads(BASELINE_PATH.read_text()) if BASELINE_PATH.exists() else {}
def update_baseline(results):
base = {r.site: {"url_count": r.url_count, "profile": r.profile_detected}
for r in results if r.passed}
BASELINE_PATH.write_text(json.dumps(base, indent=2))
print(f"[+] baseline updated -> {BASELINE_PATH}")
def write_report(results):
REPORT_DIR.mkdir(exist_ok=True)
ts = datetime.now().strftime("%Y-%m-%d_%H%M%S")
path = REPORT_DIR / f"smd_report_{ts}.md"
baseline = load_baseline()
n_pass = sum(1 for r in results if r.passed)
n_auth = sum(1 for r in results if r.needs_login)
n_fail = len(results) - n_pass - n_auth
tags = {"PASS": "[OK]", "FAIL": "[FAIL]", "AUTH NEEDED": "[AUTH]", "ERROR": "[ERR]"}
out = [
f"# SocialMediaDownloader regression report\n",
f"Run: **{ts}**\n",
f"Summary: **{n_pass} pass / {n_fail} fail / {n_auth} auth-needed** "
f"(total: {len(results)})\n",
]
for r in results:
status = r.status()
out.append(f"\n## {tags[status]} {r.site} - {status}\n")
out.append(f"- URL: `{r.url}`")
out.append(f"- Profile detected: `{r.profile_detected}`")
out.append(f"- URLs found: **{r.url_count}**")
prev = baseline.get(r.site)
if prev and prev.get("url_count"):
prev_n = prev["url_count"]; delta = r.url_count - prev_n
pct = (delta / prev_n * 100) if prev_n else 0
warn = " WARNING" if abs(pct) > 50 else ""
out.append(f"- Baseline: {prev_n} -> {r.url_count} ({pct:+.0f}%){warn}")
if r.assertions:
out.append("- Assertions:")
for a in r.assertions: out.append(f" - {a}")
if r.failures:
out.append("- **Failures:**")
for f in r.failures: out.append(f" - {f}")
if r.error: out.append(f"- **Error:** `{r.error}`")
if r.sample_urls:
out.append("- Sample URLs found:")
for u in r.sample_urls: out.append(f" - `{u}`")
if r.screenshot_path: out.append(f"- Screenshot: `{r.screenshot_path}`")
if r.notes: out.append(f"- Note: _{r.notes}_")
out.append("\n---\n## When a site fails\n")
out.append("1. `python test_smd.py --site <name>` to watch the run")
out.append("2. Open DevTools to see what changed in the DOM")
out.append("3. Update the matching profile in `social-media-downloader.js`")
out.append("4. Re-run; if counts shifted permanently: `--update-baseline`")
path.write_text("\n".join(out), encoding="utf-8")
(REPORT_DIR / "latest.md").write_text("\n".join(out), encoding="utf-8")
return path
def setup_login():
PROFILE_DIR.mkdir(exist_ok=True)
print(f"Profile dir: {PROFILE_DIR}")
print("Log into the sites you want covered, then close the window.")
with sync_playwright() as p:
ctx = p.chromium.launch_persistent_context(
user_data_dir=str(PROFILE_DIR), headless=False,
viewport={"width": 1400, "height": 900})
page = ctx.pages[0] if ctx.pages else ctx.new_page()
page.goto("https://www.facebook.com")
try:
while ctx.pages: time.sleep(1)
except KeyboardInterrupt: pass
finally: ctx.close()
print("Login profile saved.")
def main():
ap = argparse.ArgumentParser(
description=__doc__,
formatter_class=argparse.RawDescriptionHelpFormatter)
ap.add_argument("--setup", action="store_true",
help="(fallback) log into sites in dedicated profile")
ap.add_argument("--headless", action="store_true",
help="(modes 2/3) hide bundled browser")
ap.add_argument("--site",
help="one site (pinterest reddit instagram twitter facebook linkedin)")
ap.add_argument("--update-baseline", action="store_true",
help="save current counts as baseline")
ap.add_argument("--download", action="store_true",
help="after each passing test, download the top media item "
"to ~/smd-test-reports/downloads/ (off by default)")
ap.add_argument("--js", default=str(DEFAULT_JS),
help=f"path to social-media-downloader.js (default: {DEFAULT_JS.name})")
args = ap.parse_args()
if args.setup:
setup_login(); return
js_path = Path(args.js).resolve()
if not js_path.exists():
print(f"ERROR: SMD script not found at {js_path}", file=sys.stderr)
sys.exit(2)
results = run_all(js_path, headless=args.headless, site_filter=args.site,
download=args.download)
if args.update_baseline:
update_baseline(results); return
report = write_report(results)
print(f"\nReport: {report}")
print(f"Latest: {REPORT_DIR / 'latest.md'}")
failed = [r for r in results if not r.passed and not r.needs_login]
if failed:
print(f"\n{len(failed)} test(s) failed - see the report.")
sys.exit(1)
if __name__ == "__main__":
main()