Skip to content

Hugging Face Papers retrieval

Open In Colab

This notebook shows how one can use LaVague to create an agent that can outperform Gemini or ChatGPT on the task of finding the latest hottest papers on HuggingFace Papers!

HuggingFace paper example

We will create the agent and serve it through a Gradio interface.

You can see in the demo below how our agent is able to outperform both Gemini and ChatGPT to answer the query "What is the most trendy recent paper on text-to-video on Hugging Face papers? Provide the date and a summary of the paper".


Note: We use OpenAI's models, for the embedding, LLM and Vision model. You will need to set the OPENAI_API_KEY variable in your local environment with a valid API key for this example to work.

If you don't have an OpenAI API key, please get one here:


We start by downloading LaVague.

pip install lavague


We start here by pulling extra knowledge about Hugging Face's Papers pages to ensure the WorldModel will provide the right reasoning steps.

You can learn more about building Agents with LaVague in our webinar.


We can now define our agent and run it to create a Gradio demo.

from lavague.drivers.selenium import SeleniumDriver
from lavague.core import ActionEngine, WorldModel
from lavague.core.agents import WebAgent

selenium_driver = SeleniumDriver()
action_engine = ActionEngine(selenium_driver)
world_model = WorldModel()

agent = WebAgent(world_model, action_engine)
agent.demo("What is the most trendy recent paper on LLava on Hugging Face papers? Provide the date and a summary of the paper")