LaVague
LaVague is an open-source framework for building AI Web Agents that automate web tasks by turning natural language objectives into actions. It uses a World Model and an Action Engine to generate and execute code, and is aimed at developers creating automation for end users.
✨ Key features
- Large Action Model framework for AI Web Agents
- World Model and Action Engine architecture
- Supports Selenium, Playwright, and Chrome extension drivers
- Built-in contexts and customizable configuration
- Includes test runner, token counter, and logging tools
- Optional Gradio interface and debugging tools
🎯 Use cases
- Automate web tasks like navigating to a page and extracting info
- Generate and execute actions from natural language objectives
- Create AI agents for end-user process automation
- Write automated tests from Gherkin specs with LaVague QA
📦 Installation
🧰 Requirements: Requires Python, pip install lavague, and an OpenAI API key (OPENAI_API_KEY) for default configuration.
pip install lavague
🚀 Usage
from lavague.core import WorldModel, ActionEngine
from lavague.core.agents import WebAgent
from lavague.drivers.selenium import SeleniumDriver
selenium_driver = SeleniumDriver(headless=False)
world_model = WorldModel()
action_engine = ActionEngine(selenium_driver)
agent = WebAgent(world_model, action_engine)
agent.get("https://huggingface.co/docs")
agent.run("Go on the quicktour of PEFT")
# Launch Gradio Agent Demo
agent.demo("Go on the quicktour of PEFT")
⚠️ Good to know
Not all drivers support all features; for example, Chrome extension does not handle iframes, and Playwright headless and multi-tab support are coming soon.
❓ FAQ
What drivers does LaVague support?
LaVague supports Selenium, Playwright, and a Chrome extension driver, each with different feature support.
How do I turn off telemetry?
Set the LAVAGUE_TELEMETRY environment variable to "NONE" to disable all telemetry.
What is the default LLM used?
By default, LaVague uses OpenAI's gpt4-o, but it is customizable.
Can I test and benchmark LaVague agents?
Yes, LaVague includes a test runner for testing and benchmarking agent performance.
📊 Repository
🤖 Overview, features, install steps and FAQ were generated from the project's README on Sep 4, 2026. Always check the original source before running commands.