LangChain Agent | Generative AI | Priyadarsini S AP CSE | SNS Institutions

Added: 2026-05-21

[00:00:05][music] >> Okay, good morning. Today we are going to discuss an advanced and excited topic in generative AI called LangChain. We already know that large language models like LLM like ChatGPT can answer questions and generate content, but modern AI system should do more than just answering.

[00:00:32]They should think, make decision, use tools, access live data, and perform actions automatically. This is where AI agents and frameworks like LangChain become important.

[00:00:50]If AI had to plan for entire day, what it would need?

[00:00:56]I'll give you option. First is it is memory, tool, decision logic, or all of the above.

[00:01:05]Students, imagine you have an AI assistant managing your entire day. You should remember your previous task, access tools like calendar and maps, make decisions, and continuously update the plan. For example, suppose you say, "Schedule my classes, remind me of my meetings, and order food." Then AI should remember remember your schedule, use internet tools, analyze timing, and make smart decisions. This is possible only when AI combines intelligence, memory, tools, and reasoning. Simply, the conclusion is the combination of combination of all these form an AI agent.

[00:01:49]Okay.

[00:01:51]What is the problem we are dealing with traditional LLM?

[00:01:56]It cannot access the real data. No decision-making capability, no memory persistence.

[00:02:03]Traditional data traditional large language model models are powerful, but but they still have limitations. It cannot automatically book tickets, it cannot depend independently perform actions, it may forgot previous conversations, it can cannot always access live time information. This creates a problem. AI answer, but cannot act. But AI agent can search hotel, compare prices, make bookings, and send confirmation automatically. Traditional AI behaves like a knowledgeable student. AI agents behaves like an intelligent assistant who can actually do do tasks.

[00:02:40]What are agents? An agent is an system that act, observe, think, and act. An agent collect input from the user or environment.

[00:02:49]Uh for example, voice command, sensor data, or text. Think is like agent reasons and decides what to do. Act as agent performs action.

[00:03:02]Okay. LangChain is a framework used to build application powered by large language models. It connects models, tools, memory, and external system.

[00:03:11]LangChain helps developers create intelligent AI agent capable of reasoning and performing actions. Suppose you ask ask "Summarize today's AI news and email to me." LangChain can access news website, summarize information using LLM, use email APIs, and automatically send the email.

[00:03:32]LangChain acts like a bridge between AI and external tool. It acts like a tool in between them.

[00:03:42]And components of LangChain agent. LLM is a brain.

[00:03:48]Tools is like API database, search engines, calculators. and your memory stores conversations, trees, conversational histories, and context.

[00:03:59]And we have agent executor executor controls the decision-making loop, which told to use, when to stop, when the action to be performed.

[00:04:10]Next is how LangChain works.

[00:04:14]Step one, receive the input.

[00:04:17]Input is received.

[00:04:20]And it thinks.

[00:04:22]And it choose the suitable API.

[00:04:28]And it execute the task, which is perform the task.

[00:04:37]It repeats until the goal is achieved.

[00:04:41]For example, suppose user says, "Find cheapest flight to Delhi." Understand the request, search flight APIs, compare the prices, suggest pick up options, best options, update result if price changes.

[00:04:54]Just think human just like human thinks step-by-step before just completing the task. Here's the type of LangChains.

[00:05:02]React agents. React as reason plus acting.

[00:05:10]Think step-by-step and perform action.

[00:05:13]Tool calling agents.

[00:05:16]Uh agent specialized in using tools, calculator or search API. Conversational agent, agent designed for chatting and maintaining conversation. Customer support chatbot, okay? Multi-agent systems, multi-agent collaborating together. Example, one agent searches data, another agent summarizes, and another sends the input.

[00:05:37]Next is Here Here we have a case study.

[00:05:41]Uh let us understand LangChain using a real-world case study. User says, "Plan my Goa trip under the 10,000." AI agent search the flight price, check the hotel cost, create itinerary, suggest affordable location. To do this, agent uses APIs, memories, and reasoning.

[00:05:59]Why this is so powerful? The user gives only one instruction, but AI handles multiple subtasking automatically.

[00:06:06]Modern travel applications are slowly moving towards intelligent AI assistants.

[00:06:12]Okay, here are the benefits of benefits and challenges. Benefits is the automation, reduce the manual work, and the smart decision-making AI choose best action, real-time data usage accesses latest information, and challenges cost.

[00:06:26]Advanced AI system use requires powerful hardware and APIs, and it is complexity, and building agent is technically difficult. Reliability, agent may sometimes produce incorrect action.

[00:06:40]Realistic example, if an AI agent misunderstand user instructions, wrong booking may happen.

[00:06:48]Okay, here is the mind map for your reference.

[00:06:53]Here is the cues. What is the main role of an agent? It is obviously we know, observe, think, and act.

[00:07:02]Okay, which component stores stores past past information? Nothing but memory.

[00:07:08]And what makes LangChain more powerful?

[00:07:10]It is a combination of LLM, tools, memory, and reasoning.

[00:07:20]Okay.

[00:07:22]Thank you. Hope you understood this concept. Let's see you in the next video.