AI Agents¶

Contents¶

AI Agent Types¶

An "agent" is an automated reasoning and decision engine. The key agent components can include, but not limited to: - Break down of a complex question into smaller ones - Choosing an external tool to use and coming up with parameters for calling the Tool - Planning out a set of tasks - Storing previosuly completed tasks in a memory module.

When to Use Agents¶

Agents are highly beneficial when tasks require complex decision-making, autonomy, and adaptability. They excel in environments where the workflow is dynamic and involves multiple steps or interactions that can benefit from automation.

When Not to Use Agents¶

• Tasks that are straightforward, infrequent, or require minimal automation.
• Tasks that require deep domain-specific knowledge and expertise, e.g., legal or medical advice.
• Tasks that require a high level of human empathy, creativity, or subjective judgement, e.g., psychotherapy.

AI Agent Frameworks¶

Three prominent frameworks for building AI agents are: - AutoGen - CrewAI - LangGraph

Comparison summary¶

Open Source Agents¶

• AutoGPT
• BabyAGI
• SuperAGI
• ShortGPT
• ChatDev
• AutoGen
• MetaGPT
• Camel
• LoopGPT
• JARVIS
• OpenAGI

Papers/Books¶

• My own literature review on AI agents
• Agent AI: Surveying The Horizons of Multimodal Interaction. Nice overview (book) on AI agents.

Online Articles¶

1. Navigating the New Types of LLM Agents and Architectures
2. AI Agents — From Concepts to Practical Implementation in Python. This uses CrewAI framework for implementation.4
3. Top 11 Open-Source Autonomous Agents & Frameworks: The Future of Self-Running AI
4. Mastering Agents: LangGraph Vs Autogen Vs Crew AI