LaVague Weekly Update #3
LaVague: A Large Action Model framework for building Web Agents

LaVague Weekly Update #3

Welcome to our weekly updates, where we announce the latest news on our open-source framework for Web Agents & share what's coming next and how you can get involved!

Laura Yie

Welcome to our third weekly dev update!

Here’s what we’ve been up to this week...

📑 Added a local logger
We've added a local logger, which provides information about your agent run. This can be very handy for debugging. We will add more info on this in the docs soon, but for now - here is a quick notebook to show you how it works.

🚒 Wrapping all engines into the Action Engine
To avoid needing to pass a growing number of engines to your Web Agent, you now just need to initialize an Action Engine which will contain the:

  • navigation_engine (formerly the ActionEngine - generates and executes Selenium code to perform actions on a web page)
  • python_engine (generates and executes code for tasks that do not involve navigating a web page)
  • navigation_control (performs frequently required navigation tasks without needing to make any extra LLM calls: scroll up, scroll down & wait)

🤗 Our First Webinar
We did our first webinar introducing the LaVague framework. If you missed it, you can catch it on our YouTube channel here.

🛤️ Coming soon

  • A web Q&A end-to-end example which creates a pytest file based on LaVague-generated actions
  • Documentation covering LaVague's architecture, key modules, a debugging guide & FAQ
  • The launch of our Gradio interface for Web Agents
  • An eval module for evaluating Action Engine performance
  • Adding memory capabilities
  • And more events!

📢 Help wanted

This week, we defined some key areas in which we'd love your help. Please check them out here and get in touch if you'd like to become a contributor!