scitex_browser.collaboration

scitex_browser.collaboration - Interactive browser automation for AI-human teams.

This is a NEW module that does NOT affect existing scitex.browser functionality.

Features: - Persistent shared browser sessions - Multi-agent coordination - AI-human collaboration - Authentication handling - Content extraction

Version: 0.1.0-alpha (experimental)

class scitex_browser.collaboration.SharedBrowserSession(config=None)

Bases: object

Persistent browser session shared between AI agents and humans.

Simple, incremental start - just the basics.

Example

# Start session session = SharedBrowserSession() await session.start()

# Navigate await session.navigate(”http://127.0.0.1:8000”)

# Take screenshot screenshot = await session.screenshot()

# Keep running await session.wait()

# Close when done await session.close()

__init__(config=None)

Initialize shared session.

async start()

Start the persistent browser session.

Uses persistent context pattern from ScholarBrowserManager.

async navigate(url, wait_until='load', timeout=60000)

Navigate to URL.

Parameters:

• url (str) – URL to navigate to

• wait_until (str) – “load”, “domcontentloaded”, or “networkidle” (default: “load”)

• timeout (int) – Timeout in milliseconds (default: 60000)

Return type:

str

Returns:

Current URL after navigation

async screenshot(message='')

Take screenshot and save to scitex.capture directory.

Return type:

str

Returns:

Path to screenshot file

async wait(duration=None)

Wait for duration or until stopped.

Parameters:

duration (Optional[float]) – Seconds to wait, or None to wait indefinitely

async close()

Close the session.

async __aenter__()

Context manager entry.

async __aexit__(*args)

Context manager exit.

_log_event(event_type, data)

Simple event logging.

async _auto_screenshot_loop()

Automatically take screenshots at intervals.

get_info()

Get session information.

Return type:

dict

async type(selector, text, delay=50)

Type text like a human (with delay between keys).

async click(selector)

Click element.

async hover(selector)

Hover over element.

async press(key)

Press keyboard key (Enter, Tab, Escape, etc.).

async scroll_down(amount=500)

Scroll down.

async scroll_to(selector)

Scroll element into view.

async wait_for(selector, timeout=5000)

Wait for element to appear.

async wait_for_text(text, timeout=5000)

Wait for text to appear.

async wait_for_url(pattern, timeout=5000)

Wait for URL to match pattern.

async get_text(selector)

Get text content of element.

Return type:

str

async get_value(selector)

Get value of input element.

Return type:

str

async is_visible(selector)

Check if element is visible.

Return type:

bool

async ask(question, options=None)

Ask user a question in the browser.

Returns user’s response.

Example

response = await session.ask(“What username to use?”) response = await session.ask(“Choose:”, [“Option A”, “Option B”])

Return type:

str

async confirm(message)

Ask user for confirmation in browser.

Returns True if user clicks OK, False if Cancel.

Return type:

bool

class scitex_browser.collaboration.SessionConfig(session_id='default', browser_type='chromium', headless=False, viewport=<factory>, user_data_dir=None, enable_screenshots=True, screenshot_interval=None)

Bases: object

Configuration for shared browser session.

session_id: str = 'default'

browser_type: str = 'chromium'

headless: bool = False

viewport: dict

user_data_dir: str | None = None

enable_screenshots: bool = True

screenshot_interval: float | None = None

__init__(session_id='default', browser_type='chromium', headless=False, viewport=<factory>, user_data_dir=None, enable_screenshots=True, screenshot_interval=None)

class scitex_browser.collaboration.VisualFeedback(page)

Bases: object

Simple visual feedback overlay on browser.

Shows: - Who’s in control - What action is being performed - Status messages

__init__(page)

async initialize()

Initialize visual feedback overlay.

async show_message(message, type='info', duration=3000)

Show feedback message.

async show_participant(name, type='agent')

Show participant indicator.

async show_action(participant_name, action)

Show what participant is doing.

async clear()

Clear all feedback.

class scitex_browser.collaboration.CredentialManager(verbose=True)

Bases: object

Flexible credential retrieval.

Tries multiple sources in order: 1. Explicitly provided credentials 2. Environment variables 3. Terminal prompt (if terminal available) 4. Browser prompt (if browser available)

Always clearly communicates what it’s doing!

__init__(verbose=True)

async get_credential(name, env_var=None, prompt_text=None, page=None, mask=False)

Get credential from best available source.

Parameters:

• name (str) – Credential name (for caching)

• env_var (Optional[str]) – Environment variable to check

• prompt_text (Optional[str]) – Text to show in prompt

• page (Optional[Page]) – Playwright page (for browser prompts)

• mask (bool) – Whether to mask input (for passwords)

Return type:

str

Returns:

Credential value

Example

username = await creds.get_credential(

name=”username”, env_var=”SCITEX_CLOUD_USERNAME”, prompt_text=”Django username”, page=page,

)

_is_terminal_available()

Check if we can prompt in terminal.

Return type:

bool

async _prompt_terminal(name, prompt_text, mask)

Prompt user in terminal.

Return type:

Optional[str]

async _prompt_browser(page, name, prompt_text)

Prompt user in browser window.

Return type:

Optional[str]

async get_login_credentials(page=None, username_env='SCITEX_CLOUD_USERNAME', password_env='SCITEX_CLOUD_PASSWORD')

Get both username and password.

Convenient helper for login flows.

Returns:

‘…’, ‘password’: ‘…’}

Return type:

{‘username’

clear_cache()

Clear credential cache.

class scitex_browser.collaboration.GoogleAuthHelper(email=None, password=None, debug=False)

Bases: object

Google OAuth authentication helper.

Handles the popup-based Google OAuth flow used by many services.

Environment Variables:

GOOGLE_EMAIL: Default email if not provided GOOGLE_PASSWORD: Default password if not provided

__init__(email=None, password=None, debug=False)

Initialize GoogleAuthHelper.

Parameters:

• email (Optional[str]) – Google account email

• password (Optional[str]) – Google account password

• debug (bool) – Print debug messages to stderr

_log(msg)

Print debug message if debug mode is enabled.

async login_via_google_button(page, google_button_selector='button:has-text("Continue with Google")', timeout=60000)

Perform Google OAuth login via a “Continue with Google” button.

This handles the popup-based OAuth flow: 1. Click the Google button on the main page 2. Handle the Google popup for email/password entry 3. Wait for redirect back to the original service

Parameters:

• page (Page) – Playwright Page object (the main page with the Google button)

• google_button_selector (str) – CSS selector for the Google login button

• timeout (int) – Maximum time to wait for login (ms)

Return type:

bool

Returns:

True if login successful, False otherwise

async _handle_google_popup(popup, timeout=60000)

Handle the Google OAuth popup flow.

Parameters:

• popup (Page) – The Google OAuth popup page

• timeout (int) – Maximum time to wait (ms)

Return type:

bool

Returns:

True if authentication successful, False otherwise

async _fill_email(popup)

Fill email on Google login page.

Return type:

bool

async _fill_password(popup)

Fill password on Google login page.

Return type:

bool

async _handle_consent_screens(popup)

Handle OAuth consent or ‘Continue’ screens that may appear.

Return type:

None

async _wait_for_2fa(popup, timeout=60000)

Wait for 2FA verification to complete.

Detects 2FA screens and waits for user to approve on their device.

Parameters:

• popup (Page) – The Google OAuth popup page

• timeout (int) – Maximum time to wait for 2FA (ms)

Return type:

bool

Returns:

True if 2FA completed, False if timed out

async is_logged_in(page, login_indicators=None)

Check if user appears to be logged in.

Parameters:

• page (Page) – Page to check

• login_indicators (list) – List of URL substrings that indicate NOT logged in (default: [“login”, “signin”, “oauth”])

Return type:

bool

Returns:

True if appears logged in, False otherwise

async scitex_browser.collaboration.google_login(page, email, password, button_selector='button:has-text("Continue with Google")', debug=False)

Quick Google OAuth login.

Parameters:

• page (Page) – Playwright Page with Google login button

• email (str) – Google account email

• password (str) – Google account password

• button_selector (str) – CSS selector for Google button

• debug (bool) – Print debug messages

Return type:

bool

Returns:

True if login successful, False otherwise

Example

success = await google_login(page, “user@gmail.com”, “password”)