Other topics: - [[AI Protocols]] - [[AI Agent Architectures]] # Overview [[Large Language Models (LLMs)]] static machines that take an input and spit out an output. The result is nondeterministic, but the process doesn't change. This works great in some formats, like question and answer chatbots, but doesn't allow for acheiving autonomy. Agents aim to give LLMs autonomy by equipping them with the ability to execute code. This single ability opens up the door for autonous actions like: the LLM can trigger special prompts that may help it think or plan deeper, it can create code and execute it to verify output, it can reach out to other data sources to research an issue. The possibilities are endless and only bounded by context. ## An interesting formal definition from [[AI Agent Systems- Architectures, Applications, and Evaluation.pdf]] ![[AI Agents-1773687328722.webp]] # Key Considerations ## Challenges - Interactions & Long Horizon Deployment - Safety and Security Pressures ## Reasoning Paradigms - ReAct (Reasoning and Action) - ReWOO (Reasoning WithOut Observation) ## Types of Agents - Simple Reflex Agents - Model-based Reflex Agents - Goal-based Agents - Utility-based Agents - Learning Agents ## Multi-Agent Frameworks - [[AutoGen by Microsoft]] - [[CrewAI]] - [[Langchain]] # Implementation Details [[Agents.markdown]] # Useful Links # Related Topics - [[Deep Agents]] - [[AI Protocols]] ## Reference #### Working Notes #### Sources - [GitHub - NirDiamant/GenAI\_Agents: This repository provides tutorials and implementations for various Generative AI Agent techniques, from basic to advanced. It serves as a comprehensive guide for building intelligent, interactive AI systems.](https://github.com/NirDiamant/GenAI_Agents)