Agents in AI and Computing

Agents are autonomous or semi-autonomous software entities that can perceive their environment, make decisions, and take actions to achieve specific goals.

Types of Agents

Based on Capabilities

1. Simple Reflex Agents
2. Model Based Agents
3. Goal Based Agents
4. Utility Based Agents
5. Learning Agents

Based on Application Domain

1. Conversational Agents
2. Coding Agents
3. Research Agents
4. Task Automation Agents
5. Multi-Agent Systems

Agentic Frameworks

1. SmoLAgent(Hugging Face Library)

SmoLAgent (Small Language Model Agent) is a open-source agent framework designed to create efficient AI agents that use smaller language models while maintaining strong reasoning capabilities.

Features:

• Simplicity: Minimal code complexity and abstractions, to make the framework easy to understand, adopt and extend.
• Flexible LLM Support: Works with any LLM through integration with Hugging Face tools and external APIs.
• Code-First Approach: First-class support for Code Agents that write their actions directly in code, removing the need for parsing and simplifying tool calling.
• HF Hub Integration: Seamless integration with the Hugging Face Hub, allowing the use of Gradio Spaces as tools.

2. LlamaIndex

LlamaIndex is a data framework designed to connect custom data sources with Large Language Models (LLMs). While not exclusively an agent framework, it provides critical components for building knowledge-enhanced agents.

Features:

• Data Connectors: Integrates with various data sources (PDFs, APIs, databases, websites).
• Indexing: Creates optimized indexes from raw data for efficient retrieval.
• Query Engines: Processes natural language queries to retrieve relevant information.
• RAG (Retrieval-Augmented Generation): Enhances LLM responses with context from indexed data.
• Agent Memory: Provides persistent storage solutions for agent knowledge.
• Tool Integration: Allows agents to access external tools and data sources.

LlamaIndex is particularly valuable for building agents that need to work with domain-specific or proprietary data that isn’t in the LLM’s training data.

3. LangGraph

LangGraph is a framework for building stateful, multi-actor applications with LLMs, developed by the LangChain team. It extends LangChain with a graph-based architecture for creating more complex agent behaviors.

Features:

• State Management: Maintains conversational state and memory across interactions.
• Graph-Based Workflows: Models agent behavior as a directed graph with nodes and edges.
• Multi-Agent Orchestration: Coordinates communication between multiple specialized agents.
• Cyclical Workflows: Supports loops and recursive reasoning processes.
• Conditional Branching: Enables different paths based on agent decisions.
• Human-in-the-Loop: Allows human intervention at critical decision points.