Agents
An AI agent refers to a system or program that is capable of autonomously performing tasks on behalf of a user or another system by designing its workflow and utilizing available tools.
See:
Resources
- LangChain State of AI Agents Report
- There is an admiration for these capabilities of AI agents:
- Managing multistep tasks
- Automating repetitive tasks
- Task routing & collaboration
- Human-like reasoning
- There is an admiration for these capabilities of AI agents:
- ChatGPT’s next level is Agent AI: Auto-GPT, BabyAGI, AgentGPT, Microsoft Jarvis & friends
- Agents @ Work: Dust.tt - Latent Space
- Auto-GPT - An experimental open-source attempt to make GPT-4 fully autonomous
- AgentGPT
- JARVIS / HuggingGPT
- Building a Local Voice AI Assistant with Llama 3.2 & OpenAI Whisper Turbo 3 - YouTube
- Agents in AutoGen | AutoGen (microsoft.github.io)
- Agent - an entity that can act on behalf of human intent
- we can incorporate both very simple agents that can solve simple tasks with a single backend, but also we can have agents that are composed of multiple simpler agents
- When we use the word 'multi-agent' and 'single-agent', I think there are at least two different dimensions:
- Interface
- From the user's point of view, do they interact with the system in a single interaction point or do they see explicitly multiple agents working and need to interact with multiple of them?
- A single interaction point can make many applications' user experience more straightforward
- Architecture
- Are there multiple agents underneath running at the backend?
- The multi-agent design of the architecture is easier to maintain, understand and extend than a single agent system
- It's very important to recognize that the multi-agent architecture is a good way to build a single agent
- Interface
- A particular system can have a single-agent interface and a multi-agent architecture
- Ways of combining models and agents:
- Different conversation patterns or chats: sequential, group, constrained group, customized group, nested
- Different prompting and reasoning techniques such as reflection, ReAct
- Tool use and code execution
- Planning and task decomposition
- Retrieve augmentation
- Integrating multiple models, APIs, modalities and memories
- Stateflow - enhancing LLM task-solving through state-driven workflows
- #PAPER StateFlow: Enhancing LLM Task-Solving through State-Driven Workflows
- yiranwu0/StateFlow
- StateFlow - Build State-Driven Workflows with Customized Speaker Selection in GroupChat | AutoGen
- Customize Speaker Selection | AutoGen
- FSM - User can input speaker transition constraints | AutoGen
- StateFlow: Build Workflows through State-Oriented Actions | AutoGen
- EcoAssistant - Using LLM Assistants More Accurately and Affordably | AutoGen (microsoft.github.io)
- Agent for text-to-SQL with automatic error correction - Hugging Face Open-Source AI Cookbook
- Agent for text-to-SQL with automatic error correction
- Uses ReactCodeAgent in transformers.agents
- Natural Language Queries Across Databases with LLMs, AutoGen, and Langchain (PostgreSQL) | by Manojpn | Medium
- Autogen for SQL queries
- Step by Step guide to develop AI Multi-Agent system using Microsoft Semantic Kernel and GPT-4o | by Akshay Kokane | Jun, 2024 | Medium
- Multi-agent system with semantic kernel and openai assistant. Using .NET
- https://numbersstation.ai/text-to-sql-that-isnt/
- Agent for text-to-SQL with automatic error correction - Hugging Face Open-Source AI Cookbook
- Natural Language Queries Across Databases with LLMs, AutoGen, and Langchain (PostgreSQL) | by Manojpn | Medium
Courses
- #COURSE AI Agentic Design Patterns with AutoGen
- #COURSE DLAI - AI Agentic Design Patterns with AutoGen (deeplearning.ai)
- #COURSE AI Agents in LangGraph - DeepLearning.AI
- #COURSE Multi AI Agent Systems with crewAI - DeepLearning.AI
Code
- #CODE Autogen
- GitHub - autogen-ai/autogen: A programming framework for agentic AI. Discord: https://discord.gg/pAbnFJrkgZ
- GitHub - microsoft/autogen: A programming framework for agentic AI 🤖
- Resuming a GroupChat | AutoGen
- Group Chat with Coder and Visualization Critic | AutoGen
- https://microsoft.github.io/autogen/0.2/docs/notebooks/agentchat_logging
- FSM Group Chat -- User-specified agent transitions | AutoGen
- FSM - User can input speaker transition constraints | AutoGen
- agentchat_groupchat_customized.ipynb - Colab
- Group Chat with Retrieval Augmented Generation | AutoGen
- Chatbots:
- #CODE Semantic-kernel
- Introduction to Semantic Kernel | Microsoft Learn
- Microsoft’s Agentic AI Frameworks: AutoGen and Semantic Kernel | Semantic Kernel
- Microsoft’s Agentic AI Frameworks: AutoGen and Semantic Kernel | Semantic Kernel
- Step by Step guide to develop AI Multi-Agent system using Microsoft Semantic Kernel and GPT-4o | Semantic Kernel
- #CODE Letta (MemGPT)
- #CODE Mem0
- #CODE CrewAI - Framework for orchestrating role-playing, autonomous AI agents. By fostering collaborative intelligence, CrewAI empowers agents to work together seamlessly, tackling complex tasks
- #CODE LangGraph
- #CODE Llama-agents
- #CODE Transformers-agents: License to Call: Introducing Transformers Agents 2.0 (huggingface.co)
References
- #PAPER Reflexion: an autonomous agent with dynamic memory and self-reflection (2023)
- #PAPER HuggingGPT: Solving AI Tasks with ChatGPT and its Friends in HuggingFace (Shen 2023)
- #PAPER MLE-bench: Evaluating Machine Learning Agents on Machine Learning Engineering
- #CODE https://github.com/openai/mle-bench
- MLE-bench is a benchmark for measuring how well AI agents perform at machine learning engineering
- #PAPER MLAgentBench: Evaluating Language Agents on Machine Learning Experimentation
- #WHITEPAPER AIDE: Human-Level Performance in Data Science Competitions
- #PAPER OpenHands: An Open Platform for AI Software Developers as Generalist Agents (2024)
- #PAPER DSBench: How Far Are Data Science Agents to Becoming Data Science Experts? (2024)
- #PAPER Executable Code Actions Elicit Better LLM Agents (2024)