← 开源
theredsix

cerebellum

Browser automation system that uses AI-driven planning to navigate web pages and perform goals.

ApplicationsPersonalPython
在 GitHub 打开
增长势头
+024 小时新增 Star0.0%
865
Star
57
Fork
+0
本周
3
贡献者
创建于 2024-09-16 · 更新于 2026-09-17 · 今日第 13080 名
主要开发者
theredsixscm-aimlmscheidegger

theredsix/cerebellum 排名第 13,080,有 865 个星标。

同类别比较

theredsix/cerebellum 排名第 13,080,此处有 8 个同类条目;OpenBMB/ProAgent 排名第 13,084。

同类别比较
名称排名指标
ekkinox/yai13,059
语言:
Go
星标数:
867
分叉数:
60
简介:
Your AI powered terminal assistant.
youshandefeiyang/feiyangdigital-bot13,061
语言:
Java
星标数:
867
分叉数:
118
简介:
这是一个基于SpringBoot和Telegrambot-Api的多功能Telegram群管机器人,Powered By DeepSeek And Google Cloud Vision。
akshata29/entaoai13,071
语言:
TypeScript
星标数:
865
分叉数:
242
简介:
Chat and Ask on your own data. Accelerator to quickly upload your own enterprise data and use OpenAI services to chat to that uploaded data and ask questions
kazuki-sf/YouTube_Summary_with_ChatGPT13,074
语言:
JavaScript
星标数:
865
分叉数:
144
简介:
YouTube Summary with ChatGPT is a simple Chrome Extension (manifest v3) that allows you to get both YouTube video transcripts and summary of the video with OpenAI's ChatGPT AI technology.
OpenBMB/ProAgent13,084
语言:
Python
星标数:
864
分叉数:
93
简介:
An LLM-based Agent for the New Automation Paradigm - Agentic Process Automation
aws-samples/sample-mobile-ai-assistant13,117
语言:
TypeScript
星标数:
860
分叉数:
135
简介:
A lightning-fast, cross-platform mobile AI Assistant App built with React Native.
shibing624/ChatPDF13,122
语言:
Python
星标数:
860
分叉数:
146
简介:
RAG for Local LLM, chat with PDF/doc/txt files, ChatPDF. 纯原生实现RAG功能,基于本地LLM、embedding模型、reranker模型实现,支持GraphRAG,无须安装任何第三方agent库。
rsasaki0109/lidar_slam_ros213,125
语言:
Python
星标数:
859
分叉数:
174
简介:
ROS 2 LiDAR SLAM for pointcloud-map authoring, benchmarking, and Autoware-compatible map workflows.
关键参数与规格
关键参数与规格
参数值
语言Python
类别Applications
子类别Personal
关键词Computer use
星标数865
分叉数57
贡献者3
创建日期2024-09-16
更新日期2026-09-17
1 天星标数0
7 天星标数0
排名13,080
上次排名12,973
排名变化-107
7 天星标排名13,754
上次 7 天星标排名13,769
7 天星标排名变化15
总星标排名12,506
上次总星标排名12,487
总星标排名变化-19
国家USA
README 摘录

For a better agent browser tool that scores 90% on Online Mind2Web, see Agent Browser Protocol Cerebellum A lightweight browser using agent that accomplishes user-defined goals on webpages using keyboard and mouse actions. See It In Action Goal: Setup Please see setup directions for your language: Python Typescript How It Works 1. Web browsing is simplified to navigating a directed graph. Each webpage is a node with visible elements and data. User actions, such as clicking or typing, are

edges that move between nodes. Cerebellum starts at a webpage and aims to reach a target node that embodies the completed goal. It uses a LLM to finds new nodes by analyzing page content and interactive elements. The LLM decides the next action based on the current state and past actions. Cerebellum executes the LLM's planned action and feeds the new state back into the LLM for next step. The process ends when the LLM decides the goal has been

reached or is unachieveable. 5 Sonnet is the only supported LLM* Features Compatible with any Selenium-supported browser. Fills forms using user-provided JSON data. Accepts runtime instructions to dynamically adjust browsing strategies and actions. 5 Sonnet as a [x] Demonstrate successful across a variety of tasks [x] Create Python SDK [x] Handle tabbed browsing [ ] Handle data extraction from website [x] Improve vertical scrolling behavior [ ] Improve horizontal scrolling behavior [x] Improve system prompt performance [x] Improve mouse position marking on

screenshots [ ] Add ability for converting browser sessions into training datasets [ ] Support for additional LLMs as an [ ] Train a local model [ ] Integrate local model as a Known Issues Claude 3. 5 safety refusals Refuses to solve CAPTCHAs Refuses to navigate when political content is on the page Contributing Contributions to Cerebellum are welcome. For details on how to get involved, please refer to our CONTRIBUTING. We appreciate all contributions, whether they're bug reports, feature requests, or code changes. License This project

is licensed under the MIT License. Maintainer Han Wang Collaborators Han Wang Darwin Lo Michael Shuffett Shane Moran

常见问题

记录了多少星标和分叉?

theredsix/cerebellum 有 865 个星标和 57 个分叉。

仓库排名是多少?

theredsix/cerebellum 排名第 13,080。

记录的语言和类别是什么?

theredsix/cerebellum 有 865 个星标,语言为 Python,类别为 Applications。

哪个仓库排名最接近?

theredsix/cerebellum 与 OpenBMB/ProAgent 比较,后者排名第 13,084,有 864 个星标。

README

This package is depreciated. For a better agent browser tool that scores 90% on Online Mind2Web, see Agent Browser Protocol

Cerebellum

A lightweight browser using agent that accomplishes user-defined goals on webpages using keyboard and mouse actions.

See It In Action

Goal: Find a USB C to C cable that is 10 feet long and add it to cart

Setup

Please see setup directions for your language:

How It Works

  1. Web browsing is simplified to navigating a directed graph.
  2. Each webpage is a node with visible elements and data.
  3. User actions, such as clicking or typing, are edges that move between nodes.
  4. Cerebellum starts at a webpage and aims to reach a target node that embodies the completed goal.
  5. It uses a LLM to finds new nodes by analyzing page content and interactive elements.
  6. The LLM decides the next action based on the current state and past actions.
  7. Cerebellum executes the LLM's planned action and feeds the new state back into the LLM for next step.
  8. The process ends when the LLM decides the goal has been reached or is unachieveable.

Currently, Claude 3.5 Sonnet is the only supported LLM

Features

  • Compatible with any Selenium-supported browser.
  • Fills forms using user-provided JSON data.
  • Accepts runtime instructions to dynamically adjust browsing strategies and actions.
  • TODO: Create training datasets from browsing sessions

Roadmap

  • [x] Integrate Claude 3.5 Sonnet as a ActionPlanner
  • [x] Demonstrate successful BrowserAgent across a variety of tasks
  • [x] Create Python SDK
  • [x] Handle tabbed browsing
  • [ ] Handle data extraction from website
  • [x] Improve vertical scrolling behavior
  • [ ] Improve horizontal scrolling behavior
  • [x] Improve system prompt performance
  • [x] Improve mouse position marking on screenshots
  • [ ] Add ability for converting browser sessions into training datasets
  • [ ] Support for additional LLMs as an ActionPlanner
  • [ ] Train a local model
  • [ ] Integrate local model as a ActionPlanner

Known Issues

  • Claude 3.5 safety refusals
    • Refuses to solve CAPTCHAs
    • Refuses to navigate when political content is on the page

Contributing

Contributions to Cerebellum are welcome. For details on how to get involved, please refer to our CONTRIBUTING.md.

We appreciate all contributions, whether they're bug reports, feature requests, or code changes.

License

This project is licensed under the MIT License.

Maintainer

Collaborators

排名相邻
同类项目