Co-Agent: Multi-Agent Conversational Framework

Co-Agent is a sophisticated Multi-Agent Conversational Framework designed to automate the creation of LinkedIn-ready posts from blog content. Leveraging advanced AI technologies such as Google Gemini for natural language processing and multi-agent systems for task delegation, Co-Agent ensures an engaging, professional, and shareable output tailored for social media.

🌟 Key Features

• Intelligent Blog Scraping
  Extracts meaningful content from blogs using a customizable scraper.

• AI-Driven Summarization
  Summarizes blogs into concise LinkedIn posts, maintaining professional tone and format.

• Multi-Agent Collaboration
  Deploys a dynamic agent-based system for iterative content refinement, ensuring quality and coherence.

• Cross-Platform Support
  Features both a Streamlit-based UI and console-based interface for accessibility and user convenience.

• Plug-and-Play Architecture
  Easily integrates with pre-trained LLMs like Google Gemini or other APIs for future scalability.

• Preformatted Outputs
  Produces LinkedIn-ready summaries complete with hashtags, headlines, and call-to-action links.

🚀 Quick Start

Installation

Install Co-Agent from PyPI using the following command:

pip install coagent-framework

Usage

Console Workflow

Here’s a step-by-step example for using Co-Agent via the console:

from co_agent import AssistantAgent, UserProxyAgent, llm_config
from co_agent import scraper

# Setting up the Google Gemini LLM API key
llm_config["api_key"] = "Your_Google_API_Key"

# Initializing the scraper
blog_scraper = scraper.BlogScraper(name="blog_scraper")
blog_scraper.scrape()   

# Initialize agents (Assistant and User Proxy)
print("Multi-Agent Chat started:")

assistant = AssistantAgent(name="assistant", llm_config=llm_config)
user_proxy = UserProxyAgent(name="user_proxy", assistant=assistant)

# Process blog content for LinkedIn
summary = user_proxy.initiate_postmaking_process("blog_1")

Streamlit UI

To use Co-Agent with an interactive interface, visit the Streamlit app here: 👉 https://co-agent.streamlit.app/

Simply enter the blog URL in the provided field, and the app will guide you through the process of generating a LinkedIn-ready post.

🛠️ How It Works

Core Components

1. Scraper Module
   The scraper fetches content from a user-provided blog URL, stripping unnecessary formatting while retaining essential information.

2. Multi-Agent System

   • AssistantAgent: Responsible for generating initial summaries from blog content.
   • UserProxyAgent: Reviews and refines the summary based on user feedback and iterative collaboration.

3. LLM Integration
   Google Gemini, or a similar LLM, is used for understanding context, generating concise summaries, and formatting content.

4. Database Storage
   Approved summaries are stored in a database for later retrieval, ensuring content reusability.

5. Formatter
   Formats the final summary into a LinkedIn-ready post, including a headline, body, hashtags, and call-to-action links.

📂 Directory Structure

The following is the structure of the project:

coagent_framework/co_agent/
├── agents.py          # Multi-agent system implementation
├── scraper.py         # Blog scraping functionality
├── database.py        # Database utilities
├── config.py          # Configuration settings
app.py                 # main file
pyproject.toml         # Poetry configuration for dependencies
README.md              # Project documentation

📊 Example Output

Input

Blog URL: Viola-Jones Algorithm Blog

Output

Headline:
Viola-Jones Object Detection: A Revolutionary Leap in Computer Vision

Body:
Remember the early 2000s when real-time object detection felt like science fiction? That all changed thanks to the groundbreaking work of Viola and Jones! Their 2001 algorithm, a marvel of machine learning, used a boosted cascade of simple features (Haar-like features and AdaBoost) to achieve incredibly efficient object detection. This clever approach prioritized relevant features and quickly discarded irrelevant ones, making real-time face detection a reality – a true game-changer!

This algorithm's impact is still felt today. Easily accessible via OpenCV, it continues to serve as a foundational element in many computer vision applications. Want to delve deeper into the magic behind this revolutionary technique?

⚖️ License

This project is licensed under the MIT License. See the LICENSE file on our github for more details.