Framework for Developing an Agentic System
Project description
-Plug and Play-
Insert agentic ability into anywhere your existing enterprise code.
-Make Agent Robust-
Instruct agents via code, leading configurable and robust.
-Lite-
Less dependency, more scalability.
Plug-and-Play
Embed the agent's action into any your existing code, transforming your original code into an agentic system
No DSL. No Workflow. Only Python (We understand that you don't like DSL or Workflow)
Make Agent Robust
Autonomous Agents can do tasks by themselves and work in many situations, but only be able to solve very simple problems.
RPA (Robotic Process Automation), can handle complex tasks but isn't very flexible.
We provide a hybrid solution of Agents and RPA.
Code-Driven
An LLM predicts the next token
An agent predicts the next action.
We believe that an LLM-based agent needs to predict the next action; in reality, it predicts the next code. This is the philosophy of being code-driven.
Install
pip install git+https://github.com/PuppyAgent/Puppys.git
Quick Start & User Case
- 📢 Hacker News Reporter
from puppy.pp.mei import Mei
# change the API key to your own
# os.environ["OPENAI_API_KEY"] = ""
def hacker_news_decisiontree(self):
self.do_check("go to https://news.ycombinator.com/ show the HTML", show_response=True)
self.do_check("show the top 10 news @llm, and send it to me", show_response=True)
self.do_check("pick the news that related to Large Language Models, summarize all the news, and send it to me")
hacker_news = Mei(hacker_news_decisiontree)
hacker_news.run()
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.