remorses/playwriter
Playwriter
Playwriter is a browser automation framework and remote controller that manages stateful sessions and executes programmatic commands via the Chrome DevTools Protocol. It provides a system for controlling web browsers to interact with pages and extract data through both programmatic APIs and a command-line interface.
The project features a visual element selector that generates screenshots with accessibility labels, mapping visual interface elements to programmatic selectors to help agents navigate. It supports remote browser control through WebSocket tunneling, allowing users to manage browser instances running on external machines.
The framework covers isolated session management using sandboxed contexts and persistent state management to maintain continuity between separate commands. It also includes tools for headless browser testing and internal debugging of page scripts.
Features
- Browser Automation Frameworks - A comprehensive framework for managing stateful browser sessions and interacting with web elements via programmatic APIs.
- Agentic Browser Controllers - Acts as a bridge enabling AI agents to visually identify and interact with web browser elements.
- Persistent State Management - Implements a persistent object to store interaction history and state across separate command executions.
- CDP WebSocket Tunnels - Establishes WebSocket tunnels to connect to remote Chrome DevTools Protocol endpoints for browser control.
- Stateful Session Persistence - Maintains browser context and authentication state across multiple API calls to ensure continuity of interactions.
- Browser Session Managers - Manages isolated browser process sessions and data directories to prevent interference between automation agents.
- Context Isolation Managers - Isolates browsing contexts and profiles using independent session identifiers to prevent data leakage between tasks.
- Browser Session Persistence - Maintains and reuses browser sessions and isolated contexts to ensure a consistent state across automation commands.
- Remote Browser Controllers - Manages and controls remote browser instances running on different machines via standardized messaging.
- Chrome DevTools Protocols - Uses the Chrome DevTools Protocol to execute low-level page manipulation and browser control via JSON-RPC messages.
- Remote Browser Connections - Allows connecting to browser instances running on remote machines via local networks or secure tunnels.
- Command Line Interfaces - Ships a command-line tool for executing discrete JavaScript snippets against a remote browser without persistent local connections.
- Browser Interaction CLIs - Provides a CLI for executing JavaScript commands to navigate pages and interact with elements in a running browser.
- Accessibility-Based Selector Mappings - Generates screenshots with accessibility labels to map visual interface elements to programmatic selectors for AI agents.
- Programmatic API Interfaces - Provides a programmatic interface to connect and automate a browser server using the Chrome DevTools Protocol.
- Headless Browser Testing Frameworks - Automates web interactions and provides debugging for internal page scripts within headless browser environments.
- Visual Element Locators - Provides a visual element locator that generates accessibility-labeled screenshots to help automation agents identify UI elements.
- Automation Element Inspectors - Provides a tool to visually inspect elements and generate accessibility-labeled screenshots for automation agents.
- DevTools Protocol Clients - Ships a client and CLI that communicates with browsers through the Chrome DevTools Protocol to control page behavior.
