Automate AI chat workflows through browser automation.

PlayPi provides simple, high-level functions for automating browser-based AI chat workflows, starting with Google Gemini Deep Research. The package delegates all browser lifecycle management to the playwrightauthor library, so Chrome for Testing installation and profile handling happen automatically.

pip install playpi

Or for development:

git clone https://github.com/twardoch/playpi
cd playpi
uv sync

import asyncio
from playpi import google_gemini_deep_research

async def main():
    # Perform deep research on a topic
    result = await google_gemini_deep_research(
        "Analyze the impact of quantum computing on cryptography",
        headless=True,  # Retained for compatibility; playwrightauthor runs headed
        timeout=600,    # 10 minute timeout
        verbose=True    # Enable detailed logging
    )

    print(result)  # Markdown-formatted research results

asyncio.run(main())

# Standard Gemini prompt with optional file prompt + Deep Think toggle
playpi gemi --file_prompt prompt.md --prompt "Add a concise summary" --deep --output_file response.md

# Single or multi Deep Research via JSON piped on stdin
printf '{"prompt": "What are the latest developments in renewable energy?"}' | playpi gemi_dr
cat jobs.json | playpi gemi_dr

# Test browser session
playpi test

Headless flags are currently ignored because playwrightauthor always operates Chrome in headed mode.

Before using Google Deep Research, you need to:

1. Login to Google: Open https://gemini.google.com in your browser and sign in
2. Browser profiles: PlayPi relies on playwrightauthor-managed profiles, so you only need to login once in the selected profile

• Python 3.12+
• Chrome for Testing (downloaded automatically by playwrightauthor)
• 2GB+ RAM for browser automation
• Internet connection

• ✅ Google Gemini Deep Research - Automated research with source compilation
• ✅ Session Management - Persistent authentication across multiple queries
• ✅ Cross-Platform - Works on Windows, macOS, and Linux
• ✅ Concurrent Processing - Run up to 3 research tasks simultaneously
• ✅ Rich Output - Clean Markdown formatting with links and structure
• ✅ Error Handling - Graceful handling of timeouts and UI changes

Perform Google Gemini Deep Research on a given prompt.

Parameters:

• prompt (str): Research query or question
• headless (bool): Retained for compatibility; playwrightauthor currently launches Chrome in headed mode (default: True)
• timeout (int): Maximum wait time in seconds (default: 600)
• verbose (bool): Enable detailed logging (default: False)
• profile (str): Browser profile name managed by playwrightauthor (default: "default")

Returns:

• str: Research results formatted as Markdown

Raises:

• AuthenticationError: Not logged in to Google/Gemini
• PlayPiTimeoutError: Research exceeded timeout limit
• ProviderError: UI elements not found or changed

# Run all tests without hitting live Gemini login
PLAYPI_FORCE_AUTH_FAILURE=1 uvx hatch run test:python -m pytest

# Run with coverage (same auth guard applies)
PLAYPI_FORCE_AUTH_FAILURE=1 uvx hatch run test:python -m pytest --cov=src/playpi --cov-report=html

# Run specific test categories
PLAYPI_FORCE_AUTH_FAILURE=1 uvx hatch run test:python -m pytest tests/test_session.py -v

PLAYPI_FORCE_AUTH_FAILURE=1 short-circuits the authentication check during testing so suites do not wait for manual Google sign-in.

# Format code
uv run ruff format

# Lint code
uv run ruff check

# Type checking
uv run mypy src/playpi

MIT License. See LICENSE for details.