This repository contains the explanation, links, and videos about these AI Agents and agentic systems:

According to OpenAI, swarm focuses on “making agent coordination and execution lightweight, highly controllable, and easily testable.” Swarm uses two subsystems: ‘agents’ or ‘routines’ and ‘hand-offs’. In these systems, an agent is responsible for instructions and tools and can, at any point during the process, hand off the responsibility to another agent.

Google introduced its latest model, DataGemma. Unlike previous RAG systems, it works on a RIG (Retrieval Interleaved Generation) technique, merging Language Models with Data Commons, an open-source large-scale database of public data.

Adala offers a robust framework for implementing agents specialized in data processing, emphasising diverse data labelling tasks. These agents are autonomous, meaning they can independently acquire one or more skills through iterative learning. This learning process is influenced by their operating environment, observations, and reflections. Users define the environment by providing a ground truth dataset. Every agent learns and applies its skills in what we call a "runtime", synonymous with LLM.

CrewAI agentic framework offers multi-agent collaboration with different LLMs for complex tasks, including RAG systems.

AutoGen is a Multi-agent multi-task agentic framework. AutoGen is the collaboration between humans and multiple agentic systems. Agents communicate via asynchronous messages, supporting both event-driven and request/response interaction patterns. Users can interoperate agents across different programming languages.

The developers (Shinn et al in 2023) showed that using linguistic feedback and asking the model to self-refine itself is more effective than the traditional reinforcement learning methods e.g., by updating weights.

HuggingGPT uses ChatGPT as the controller to receive the task prompt from a user and then divide it into sub-tasks. The controller then uses any relevant model available in the Hugging Face repositories to perform the sub-tasks and retain the results, and even combine the results of multiple AI models for very complex tasks that involve multiple modalities such as text, speech, and vision.

Open Interpreter lets LLMs run code (Python, Javascript, Shell, and more) locally. You can chat with Open Interpreter through a ChatGPT-like interface in your terminal by running $ interpreter after installing. This provides a natural-language interface to your computer's general-purpose capabilities.

The Internet of Agents or IOA system for large language models is inspired by how the Internet works, more specifically decentralised collaborative projects such as Wikipedia and Linux, to build an Internet-like system for autonomous agents. Below is a video explaining the full method: The Internet of Agents or IOA system for large language models

AgentForge is a low-code framework designed for the rapid development, testing, and iteration of AI-powered autonomous agents and cognitive architectures. Compatible with a range of LLM models—including OpenAI, Google's Gemini, Anthropic's Claude, and local models via Ollama or LMStudio.

AgentVerse is designed to facilitate the deployment of multiple LLM-based agents in various applications. AgentVerse primarily provides two frameworks: task-solving and simulation.

This is an agentic system for Data discovery, data cleaning, analysis & visualization; AI Data Analyst that works with your CSV, Excel, Google Sheets and SQL Databases. The system has a limited free tier too.

This is a general-purpose agentic system based on the paper "Improving Factuality and Reasoning in Language Models through Multiagent Debate", where individual language models generate and critique other models' outputs, to contribute to factually accurate responses.

GPT Pilot is an AI agent that codes the entire app while you can observe the code being written as a development tool for scalable apps. Only around 5% of intervention is needed by a developer.

Vanna is an Open-Source Python-based AI SQL agent trained on your schema that writes complex SQL queries for your databases using a RAG system based on your training data, i.e., the databse information.

Multi-agent-building frameworks and platform using the best LLMs from OpenAI, Google, Mistral, and Anthropic to deploy agents where your team is (Zendesk, Slack, Discord) and your agents can collaborate with each other as a team to complete complex workflows.