run in docker, and work through more than one account

Both taken from TheNetsky/Microsoft-Rewards-Script, which packages a container
and handles several accounts. Approach only: that project is GPL-3.0 and this
one MIT, so no code crosses over.

**Accounts.** Rewards is per Microsoft account and the browser profile holds
the sign-in, so an account here is a profile directory. REWARDS_ACCOUNTS takes
a comma separated list and gives each its own directory under the configured
one. They run in sequence, and a profile that will not start is reported and
skipped rather than ending the run. Left unset, a run uses the single profile
exactly as before.

Names are validated rather than trusted: they become directory names, so
"../escape" is refused instead of quietly writing outside data-dir.

**Docker.** The image carries only what main.py actually reaches, selenium and
numpy. pygetwindow, keyboard, matplotlib and pygame are used solely by the
recording and visualisation scripts, and two of those are Windows-only, so
none of them belong in a container. msedgedriver is pinned at build time to
the Edge the image installed rather than to latest, which drifts from it
between releases.

QUERY_SOURCE defaults to trends in the image, so a container needs no Ollama
account and no model download at all.

That default turned out to require a fix. queries.py imported llm_utils at
module scope, which imports ollama, so a trends-only install still needed the
ollama package: exactly what running in a minimal image is good at exposing.
The import is now made inside the llm branch, and the wordlist fallback reads
nouns.txt directly rather than borrowing llm_utils.get_random_noun.

REWARDS_HEADLESS drives the headless flags. The window size is set explicitly
because the pointer code works in viewport coordinates and the default
headless window is small enough to put cards out of reach, which is the
MoveTargetOutOfBoundsException from #19. Verified on the host that
move_to_element and human_like_click both work headless before relying on it.

Verified in the built image: Edge 151.0.4129.107 with a driver of exactly the
same build, the trends feed reachable from inside, Edge driven to bing.com and
rewards.bing.com at 1920x1080, and REWARDS_ACCOUNTS producing separate profile
directories with traversal refused.

src/accounts.py

@@ -0,0 +1,85 @@
+"""Which accounts a run works through.
+
+Rewards is per Microsoft account, and the browser profile is what holds the
+sign-in, so an account here is just a profile directory. One directory per
+account keeps their cookies apart, which is the whole requirement.
+
+    REWARDS_ACCOUNTS=personal,spare python src/main.py
+
+Unset, the run uses the single profile in constants.py exactly as before, so
+nothing about an existing setup changes.
+"""
+
+import os
+import re
+from dataclasses import dataclass
+
+from constants import USER_DATA_DIR, PROFILE_NAME
+
+ENV_VAR = "REWARDS_ACCOUNTS"
+
+# Names become directory names, so keep them to something a filesystem and a
+# command line both handle without quoting.
+SAFE_NAME = re.compile(r"^[A-Za-z0-9._-]+$")
+
+
+@dataclass(frozen=True)
+class Account:
+	"""A named browser profile to run the tasks against."""
+
+	name: str
+	user_data_dir: str
+	profile_name: str
+
+	@property
+	def is_default(self) -> bool:
+		return self.user_data_dir == USER_DATA_DIR
+
+
+def _named(name: str) -> Account:
+	# Each account gets its own directory under the configured one, so the
+	# existing data-dir stays where it is and the new ones sit beside the
+	# profile it already holds.
+	return Account(
+		name=name,
+		user_data_dir=os.path.join(USER_DATA_DIR, name),
+		profile_name=PROFILE_NAME,
+	)
+
+
+def configured() -> list[Account]:
+	"""Accounts for this run, in order.
+
+	Raises ValueError on a name that cannot be a directory, rather than
+	silently creating something surprising next to the real profiles.
+	"""
+	raw = os.environ.get(ENV_VAR, "").strip()
+
+	if not raw:
+		return [Account(name="default", user_data_dir=USER_DATA_DIR, profile_name=PROFILE_NAME)]
+
+	names = [part.strip() for part in raw.split(",")]
+	names = [name for name in names if name]
+
+	if not names:
+		return [Account(name="default", user_data_dir=USER_DATA_DIR, profile_name=PROFILE_NAME)]
+
+	seen: set[str] = set()
+	accounts: list[Account] = []
+
+	for name in names:
+		if not SAFE_NAME.match(name):
+			raise ValueError(
+				f"{ENV_VAR} entry {name!r} is not usable as a directory name; "
+				"use letters, digits, dot, dash or underscore"
+			)
+
+		# Duplicates would run the same profile twice, which earns nothing the
+		# second time and doubles the run length.
+		if name.lower() in seen:
+			continue
+
+		seen.add(name.lower())
+		accounts.append(_named(name))
+
+	return accounts

src/main.py

@@ -1,26 +1,92 @@
+import os
+import sys
+
+import accounts
 import rewards_tasks
 import mouse_trajectory
 import mimic_typing
 from selenium import webdriver
-from constants import USER_DATA_DIR, PROFILE_NAME
+from selenium.common.exceptions import SessionNotCreatedException
 
-options = webdriver.EdgeOptions()
+HEADLESS = os.environ.get("REWARDS_HEADLESS", "").strip().lower() in ("1", "true", "yes")
 
-options.add_experimental_option("excludeSwitches", ["enable-automation"])
-options.add_experimental_option('useAutomationExtension', False)
-options.add_argument("--disable-blink-features=AutomationControlled")
-options.add_argument(f"--user-data-dir={USER_DATA_DIR}")
-options.add_argument(f"--profile-directory={PROFILE_NAME}")
 
-driver = webdriver.Edge(options=options)
+def build_options(account: accounts.Account) -> webdriver.EdgeOptions:
+	options = webdriver.EdgeOptions()
 
-mouse = mouse_trajectory.MouseUtils(driver)
-keyboard = mimic_typing.KeyboardUtils(driver)
+	options.add_experimental_option("excludeSwitches", ["enable-automation"])
+	options.add_experimental_option('useAutomationExtension', False)
+	options.add_argument("--disable-blink-features=AutomationControlled")
+	options.add_argument(f"--user-data-dir={account.user_data_dir}")
+	options.add_argument(f"--profile-directory={account.profile_name}")
 
-rewards = rewards_tasks.RewardsTaskUtils(driver)
+	if HEADLESS:
+		# A container has no display. The window size is set explicitly because
+		# the pointer code works in viewport coordinates, and the default
+		# headless window is small enough to put cards out of reach.
+		options.add_argument("--headless=new")
+		options.add_argument("--window-size=1920,1080")
+		options.add_argument("--no-sandbox")
+		options.add_argument("--disable-dev-shm-usage")
 
-rewards.complete_all_tasks()
+	return options
 
-input("Press Enter to exit...")
 
-driver.quit()
+def run_account(account: accounts.Account) -> bool:
+	"""Work one account. Returns whether the browser started."""
+	try:
+		driver = webdriver.Edge(options=build_options(account))
+	except SessionNotCreatedException as exc:
+		# Chromium allows one process per user data directory. When the profile
+		# is already open the driver's copy exits during startup, and selenium
+		# reports it as the browser crashing with a message that names neither
+		# the profile nor the other window.
+		print(f"[FAIL] {account.name}: could not start Edge with this profile.")
+		print(f"       profile directory: {account.user_data_dir}")
+		print("       The usual cause is that this profile is already open in another")
+		print("       Edge window, including one left over from a previous run.")
+		print(f"       driver said: {str(exc).strip().splitlines()[0]}")
+
+		return False
+
+	try:
+		mouse = mouse_trajectory.MouseUtils(driver)
+		keyboard = mimic_typing.KeyboardUtils(driver)
+
+		rewards = rewards_tasks.RewardsTaskUtils(driver)
+		rewards.complete_all_tasks()
+	finally:
+		driver.quit()
+
+	return True
+
+
+def main() -> int:
+	try:
+		configured = accounts.configured()
+	except ValueError as exc:
+		print(f"[FAIL] {exc}")
+
+		return 2
+
+	started = 0
+
+	for account in configured:
+		if len(configured) > 1:
+			print(f"\n=== account: {account.name} ===")
+
+		if run_account(account):
+			started += 1
+
+	if len(configured) > 1:
+		print(f"\n{started}/{len(configured)} accounts ran")
+
+	# Nothing is watching a container, and stdin is not a terminal there.
+	if not HEADLESS:
+		input("Press Enter to exit...")
+
+	return 0 if started else 1
+
+
+if __name__ == "__main__":
+	sys.exit(main())

src/queries.py

@@ -13,9 +13,14 @@ Bing, and Bing's own autosuggest answers that question directly.
 
 import os
 
-import llm_utils
 import query_sources
 
+# llm_utils is imported inside the llm branch rather than here. It imports
+# ollama at module scope, so importing it eagerly would make the ollama package
+# a hard requirement even for a run that never touches a model, which is the
+# opposite of the point. A trends-only install, the Docker image for instance,
+# does not ship it.
+
 LLM = "llm"
 TRENDS = "trends"
 
@@ -45,6 +50,8 @@ def search_query_for_task(task_description: str) -> str:
 
 		return task_description.lower()
 
+	import llm_utils
+
 	return llm_utils.get_search_query_from_task_description(task_description)
 
 
@@ -59,6 +66,8 @@ def related_queries(count: int):
 		print("[WARNING] No query source reachable, falling back to the wordlist.")
 
 		# nouns.txt is already in the repo for exactly this kind of seed.
-		return [llm_utils.get_random_noun() for _ in range(count)]
+		return query_sources.wordlist_queries(count)
+
+	import llm_utils
 
 	return llm_utils.get_related_search_queries(llm_utils.get_random_noun(), num_queries=count)

src/query_sources.py

@@ -130,6 +130,24 @@ def suggestions(seed: str) -> list[str]:
 	return [_clean(s) for s in payload[1] if _clean(s)]
 
 
+def wordlist_queries(count: int) -> list[str]:
+	"""Seeds from nouns.txt, the last resort when nothing is reachable.
+
+	Read here rather than borrowed from llm_utils so that a trends-only install
+	never has to import the model client.
+	"""
+	try:
+		with open("nouns.txt", encoding="utf-8") as handle:
+			nouns = [line.strip().lower() for line in handle if len(line.strip()) >= 3]
+	except OSError:
+		return []
+
+	if not nouns:
+		return []
+
+	return random.sample(nouns, min(count, len(nouns)))
+
+
 def _clean(text: str) -> str:
 	"""Strip markup, collapse whitespace and drop punctuation Bing does not need."""
 	text = re.sub(r"<[^>]+>", " ", text or "")