LaVague Weekly Update #4
LaVague: A Large Action Model framework for building Web Agents

LaVague Weekly Update #4

Welcome to our fourth dev update, where we announce the revival of our Gradio interface, a Playwright integration, a memory component & new documentation

Laura Yie

Welcome to our fourth weekly dev update!

Here are our dev updates for this week:

  • The interactive Gradio interface is back: Check out the our guide for more details.
  • Agent-compatible Playwright Driver: We’ve added an agent-compatible Playwright Driver.
  • Adding new module guides in the docs:
  • Memory Module: We've added a Memory module which enables the World Model to access past instructions and user-supplied data. (Belated announcement from last week)
  • Agent run method update: The agent.run() method now returns an ActionResult object containing:
    • instruction: The user objective
    • code: All the code of the successful actions run (for Navigation Engine only)
    • success: Boolean indicating if the objective was achieved
    • output: The output, if any (e.g., information retrieved from a website)
  • Adding a new web QA use cases: See the new web QA use case in our documentation

📢 Help Wanted

  • Community Poll: Please vote on our poll on Discord to help us make the project more contribution-friendly.
  • General Contributions:
    • Keep testing LaVague, opening bug reports, and suggesting new features!
    • Check out our GitHub issues or ping us if you want to contribute to the project.