Recently, I came across this Hugging Face AI Agents Course, where you can study AI Agents in theory, design and practice. For this course, you will be required to have a computer system and a Hugging Face account (bare minimum). The course is quite detailed and it will help you build a solid foundation in the fundamentals of AI Agents. I have completed the first module and created simple agents using SmolAgent Framework. Talking about the course in the first module, you will find a simple to understand AI Agents definition, the role of LLMs in Agents, how Agents use external tools to interact with the environment, the agent workflow (Think \u2192 Act \u2192 Observe.) and more.<\/p>\n
Here in this article, I will be sharing what I learnt from the course, let\u2019s dig in!!!<\/p>\n To understand what an AI Agent is, let\u2019s give some order to the agent \u2013 Sapphire (Name of the Agent). You can assume this agent to be your assistant that helps you with the everyday tasks like cooking, brewing and more.<\/p>\n Here we are ordering or giving a task to the agent \u201cHey Sapphire, can you make a good tea for me?\u201d<\/em><\/p>\n Now, Sapphire can understand this language and process the request of the user easily. But what happens internally? Sapphire Reason and Plan to execute the steps he needs to follow to make a good cup of tea.<\/p>\n These are planning steps and after this Sapphire will execute his plan using different tools like \u2013 Kettle, Teapot or Mug, Tea Infuser or Strainer (for loose-leaf tea), Teaspoon.\u00a0<\/p>\n Once the task is completed, you\u2019ll have a cup of tea to energize your day\u2014a practical example of how an Agent operates.<\/p>\n Here\u2019s the technical definition:<\/p>\n An Agent refers to an artificial intelligence system designed to analyze, strategize, and engage with its surroundings autonomously. The term \u201cAgent\u201d stems from its agency\u2014the power to independently perceive, decide, and act within a given environment to achieve goals (like crafting your ideal morning beverage).<\/p>\n Read this to know more: What are AI Agents?<\/a><\/em><\/p>\n An agent can be conceptualized as having two interconnected components:<\/p>\n This component serves as the agent\u2019s \u201cintelligence hub.\u201d It processes information, analyzes contexts, and generates strategic plans. Using algorithms or learned patterns, it dynamically selects appropriate actions to achieve goals based on real-time inputs or environmental conditions.<\/p>\n But how can this AI Model be used as a brain for the Agent? The most common AI model found in Agents is an LLM (Large Language Model), which takes Text as an input and outputs Text as well. Such as GPT-4, Llama, Gemini and more.<\/p>\n Similarly, LLMs like ChatGPT can also generate images but how? Aren\u2019t these text generation models? You are absolutely right, by nature, are but these are integrated with additional functionality (called Tools), that the LLM can use to create images. This is how AI takes action on its environment.\u00a0<\/p>\n This component represents the agent\u2019s tangible abilities and resources. It encompasses tools, sensors, and physical or digital actuators that translate decisions into outcomes. The range of feasible actions is inherently constrained by the agent\u2019s design\u2014for instance, a human agent cannot execute a \u201cfly\u201d action due to biological limits but can perform \u201csprint,\u201d \u201clift,\u201d or \u201cthrow\u201d using their musculoskeletal system. Similarly, a robot\u2019s actions depend on its programmed hardware (e.g., grippers, wheels).<\/p>\n An agent\u2019s effectiveness hinges on the synergy between its Cognitive Core<\/em> (strategic adaptability) and Operational Interface<\/em> (practical capacity). Limitations in either domain directly impact its functional scope.<\/p>\n As mentioned above, an agent can accomplish tasks by leveraging specialized tools programmed to execute specific actions. These tools act as building blocks, enabling the agent to interact with its environment and solve problems.<\/p>\n Example Scenario:<\/p>\n Imagine designing an agent to manage your calendar (e.g., a virtual assistant). If you request, \u201cReschedule today\u2019s team meeting to 3 PM,\u201d the agent could use a custom tool like a reschedule_meeting function. Here\u2019s how it might work in Python:<\/p>\n When prompted, the agent\u2019s LLM (Large Language Model) would autonomously generate code to invoke this tool:<\/p>\n Key Concepts:<\/strong><\/p>\n By focusing on strategic tool creation, agents evolve from simple scripts into dynamic systems capable of complex, real-world problem-solving. For instance, Personal Virtual Assistants, Customer Service Chatbots and others are good examples of AI Agents.<\/p>\n An LLM (Large Language Model)<\/a> is an advanced AI system that reads, interprets, and creates human-like text. These models learn by analyzing massive amounts of written content\u2014like books, articles, and websites\u2014to grasp language rules, context, and even subtle meanings. The more data they process, the better they become at tasks like writing or answering questions. Most modern LLMs rely on a structure called the Transformer<\/a>, a design introduced since the release of BERT from Google in 2018.<\/p>\n Transformers use a clever method called \u201cattention\u201d<\/strong> to focus on the most important parts of a sentence or phrase. This helps them understand relationships between words, even if they\u2019re far apart. There are three main types of Transformers:<\/p>\n Most famous LLMs today, like ChatGPT or Claude, use decoder-based Transformers. These models excel at creative tasks because they\u2019re built to predict and generate text step-by-step. Their enormous size (billions of parameters) allows them to handle complex language patterns.<\/p>\n Popular LLMs You Might Know<\/strong>:<\/p>\n In short, LLMs are powerful tools that mimic human language skills, and their Transformer \u201cbrain\u201d helps them adapt to everything from answering questions to writing poetry! Here are some popular decoder-based models:<\/p>\n A large language model (LLM) operates on a simple yet effective principle: it predicts the next token in a sequence based on the ones that came before. A \u201ctoken\u201d is the smallest unit of text the model processes. While it may resemble a word, tokens are often smaller segments, making them more efficient for language processing.<\/p>\n Rather than using full words, LLMs rely on a limited vocabulary of tokens. For instance, although the English language has around 600,000 words, an LLM like Llama 2 typically works with about 32,000 tokens. This is because tokenization breaks words into smaller components that can be combined in different ways.<\/p>\n For example, the word \u201cplayground\u201d might be split into \u201cplay\u201d and \u201cground\u201d, while \u201cplaying\u201d could be divided into \u201cplay\u201d and \u201cing\u201d. This allows LLMs to efficiently process variations of words while maintaining flexibility in understanding language.<\/p>\n Here\u2019s the tokenizer playground for you to experiment with the tokens for a particular word or sentence:<\/p>\n![\"Hugging](\"https:\/\/cdn.analyticsvidhya.com\/wp-content\/uploads\/2025\/03\/unnamed-2025-03-10T122018.420.webp\")
<\/figure>\nWhat is an AI Agent?<\/h2>\n
![\"What](\"https:\/\/cdn.analyticsvidhya.com\/wp-content\/uploads\/2025\/03\/unnamed-2025-03-10T122133.284.webp\")
\n
![\"Reason,](\"https:\/\/cdn.analyticsvidhya.com\/wp-content\/uploads\/2025\/03\/unnamed-2025-03-10T122159.355.webp\")
1. Cognitive Core (Decision-Making System \u2013 AI Model, the Brain)<\/h3>\n
2. Operational Interface (Action Execution System)<\/h3>\n
def reschedule_meeting(participant, new_time, agenda):\u00a0\u00a0\n\u00a0\u00a0\u00a0\u00a0\"\"\"Reschedules a meeting with a participant to a specified time and updates the agenda.\"\"\"\u00a0\u00a0\n\u00a0\u00a0\u00a0\u00a0# Code to integrate with calendar APIs (e.g., Google Calendar)\u00a0\u00a0\n\u00a0\u00a0\u00a0\u00a0...<\/code><\/pre>\nreschedule_meeting(\"project_team\", \"3:00 PM\", \"Q3 deadlines discussion\")\u00a0<\/code><\/pre>\n\n
\n
\n
\n
\n
Technical Explanation of the Use of LLMs<\/h2>\n
How Transformers Work?<\/h2>\n
\n
\n
\n
\n
Why Decoders Dominate Modern LLMs?<\/h3>\n
\n
\n\n
\n \nModel<\/th>\n Provider<\/th>\n<\/tr>\n<\/thead>\n \n Deepseek-R1<\/td>\n DeepSeek<\/td>\n<\/tr>\n \n GPT4<\/td>\n OpenAI<\/td>\n<\/tr>\n \n LLaMA 3<\/td>\n Meta (Facebook AI Research)<\/td>\n<\/tr>\n \n SmolLM2<\/td>\n Hugging Face<\/td>\n<\/tr>\n \n Gemma<\/td>\n Google<\/td>\n<\/tr>\n \n Mistral<\/td>\n Mistral<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/figure>\n LLM\u2019s Prediction of Next Token<\/h2>\n
![\"chunking\"](\"https:\/\/cdn.analyticsvidhya.com\/wp-content\/uploads\/2025\/03\/image-11-2.webp\")