AB
AiBoss
project

Cerebellum - A browser AI assistant for automating web page tasks and action planning.

Cerebellum is a browser-based AI assistant built on Claude 3.5 Sonnet and Selenium WebDriver. It can understand task intent and perform automated web page tasks, such as data scraping and automated website testing. Cerebellum...

What is Cerebellum?

Cerebellum is a browser-based AI assistant built on Claude 3.5 Sonnet and Selenium WebDriver. It understands task intent and executes automated web tasks such as data scraping and website automation testing. Cerebellum simplifies web browsing as directed graph navigation, using LLM to analyze page content and interactive elements, intelligently planning action paths, and dynamically adjusting strategies based on page status and historical operations. Cerebellum supports multiple browsers, accurately simulates user behavior, and is suitable for complex automation scenarios.

The main functions of Cerebellum

  • Graphical NavigationThis simplifies web browsing into a directed graph navigation, where each webpage is treated as a node, and user actions (such as clicks or input) are treated as edges between nodes.
  • Node discoveryBased on LLM (Large Language Model) analysis, new web page nodes are discovered and identified.
  • Action DecisionBased on the current webpage status and historical user actions, intelligently decide on the next action.
  • Multi-browser supportIt is compatible with multiple browsers, including Chrome, Firefox, Safari, and Edge.
  • User behavior simulationIt accurately simulates user behaviors, such as clicks and input, and is suitable for complex automation scenarios.
  • Dynamic strategy adjustmentIt accepts runtime instructions and dynamically adjusts browsing strategies and operations based on real-time feedback.
  • Form fillingAutomatically fill out web forms using user-provided JSON data.

The technical principles of Cerebellum

  • Web page modelingThe web browsing process is modeled as a directed graph, where each webpage is a node and user actions are the edges connecting the nodes.
  • LLM IntegrationIntegrating large language models (LLMs), such as Claude 3.5 Sonnet, to analyze web page content and interactive elements, and identify and plan new nodes.
  • State AnalysisLLM determines the best course of action based on the current state of the webpage and historical operation data.
  • Action execution and feedbackCerebellum executes the operations planned by the LLM and feeds back the new webpage status after the operation to the LLM for further decision-making.
  • Goal-orientedStarting from the initial webpage, the goal is to reach the target node that represents the completion of the task.
  • Automated processesUsers set goals, and Cerebellum automatically handles the tedious script recording and execution, achieving automated tasks.
  • Environmental IntegrationTightly integrated with Selenium WebDriver, enabling cross-browser automation based on Selenium's browser-driven capabilities.

Cerebellum project address

Applications of Cerebellum

  • Website Automated TestingTo perform functional testing of web pages, automate the execution of test scripts, and ensure that all functions of a website or web application work as expected.
  • Data scraping and analysisAutomated data scraping from web pages for market research, competitor analysis, or data aggregation services.
  • Web content managementManage website content, such as automatically updating articles, publishing blogs, and managing user comments.
  • e-commerceIt automates tasks such as order processing, inventory management, price comparison, and product information updates.
  • Customer service automationImprove customer service efficiency by leveraging automated chatbots or frequently asked questions in customer support systems.