{
  "video_id": "d2-R_o5I5FQ",
  "channel_slug": "jamesbriggs",
  "channel_handle": "jamesbriggs",
  "title": "When Should You Use LangChain?",
  "duration_seconds": 560.0,
  "url": "https://www.youtube.com/watch?v=d2-R_o5I5FQ",
  "upload_date": "",
  "transcript": "Okay, so the first chapter of the\ncourse, we're going to focus on when\nshould we actually use lang chain and\nwhen should we use something else. Now\nthrough this chapter, we're not really\ngoing to focus too much on the code.\nWell, you know, every other chapter is\nvery code focused, but this one is a\nlittle more just theoretical. What is\nline chain? Where's it fit in? When\nshould I use it? When should I not? So,\nI want to just start by framing this.\nLine Chain is one of, if not the most\npopular open-source framework within the\nPython ecosystem, at least for AI. It\nworks pretty well for a lot of things\nand also works terribly for a lot of\nthings as well. To be completely honest,\nthere are massive pros, massive cons to\nusing chain. Here, we're just going to\ndiscuss a few of those and see how line\nchain maybe compares a little bit\nagainst other frameworks. So the very\nfirst question we should be asking\nourselves is do we even need a\nframework? I is a framework actually\nneeded when we can just hit an API. You\nhave the open AI API other APIs mistrol\nso on and we can get a response from an\nLLM in five lines of code on average for\nthose is incredibly incredibly simple.\nHowever, that can change very quickly\nwhen we start talking about agents or\nretrieval, augmented generation,\nresearch assistance, all this sort of\nstuff.\nThose use cases, those methods can\nsuddenly get quite complicated when\nwe're outside of frameworks. And that's\nnot necessarily a bad thing, right? It\ncan be incredibly useful to be able to\nuh just understand everything that is\ngoing on and build it yourself. But the\nproblem is that to do that you need time\nlike you need to learn all the\nintricacies of building these things.\nThe intricacies of these methods\nthemselves like what you know how do\nthey even work? And that kind of runs in\nthe opposite direction of what we see\nwith AI at the moment, which is AI is\nbeing integrated into the world at an\nincredibly fast rate. And because of\nthis, most engineers coming into the\nspace are not from a machine learning or\nAI background. Most people don't\nnecessarily have any experience with\nthese systems. a lot of engineers coming\nin that could be DevOps engineers,\ngeneric backend Python engineers, even\nyou know front- end engineers coming in\nand building all these things which is\nis great but they don't necessarily have\nthe experience and that you know that\nmight be you as well and that's not a\nbad thing because the idea is that\nobviously you're going to learn and\nyou're going to pick up a lot of these\nthings and in this scenario there's\nquite a good argument for using a\nframework because a framework means that\nyou can get started faster and a\nframework like Langchain, it abstracts\naway a lot of stuff. And that's a m\nthat's a big complaint that a lot of\npeople will have with lang chain. But\nthat abstracting away of many things is\nalso what made slime chain popular\nbecause it means that you can come in\nnot really knowing okay what you know\nrag is for example and you can implement\na rag pipeline get the benefits of it\nwithout really needing to understand it.\nAnd yes, there's an argument against\nthat as well, just implementing\nsomething without really understanding\nit. But as we'll see throughout the\ncourse, it is possible to work with line\nchain in a way as we will in this course\nwhere you can implement these things in\nan abstract way and then break them\napart and start understanding the\nintricacies at least a little bit. So\nthat can actually be pretty good.\nHowever, again, circling back to what we\nsaid at the start, if the idea or your\napplication is just a very simple, you\nknow, you need to generate some text\nbased on some basic input, maybe you\nshould just use an API, that's\ncompletely valid as well. Now, we just\nsaid, okay, a lot of people coming to\nLchain might not be from an AI\nbackground.\nSo another question for a lot of these\nengineers might be okay if I want to\nlearn about you know rag agents all\nthese things should I skip line chain\nand just try and build it uh from\nscratch myself well line chain can help\na lot with that learning journey so you\ncan start very abstract and as you\ngradually begin to understand the\nframework better you can strip away more\nand more of those abstractions and get\nmore into the detail. And in my opinion,\nthis gradual shift towards more explicit\ncode with less abstraction is a really\nnice feature and it's also what we focus\non right throughout this course. That's\nwhat we're going to be doing. Going to\nbe signing abstract, stripping away the\nabstractions and getting more explicit\nwith what we're building. So for\nexample, building an agent in lang chain\nthere's in very simple and incredibly\nabstract create tools agent method that\nwe can use and like it it creates a tool\nagent for you. It's it doesn't tell you\nanything. So you can you can use that\nright and we we will use that initially\nin the course but then you can actually\ngo from that to defining your full agent\nexecution logic which is basically a\ntools call to open AAI. You're going to\nbe getting that tool information back\nbut then you got to figure out okay how\nam I going to execute that? How am I\ngoing to store this information? And\nthen how am I going to iterate through\nthis? So we're going to be seeing that\nyou stripping away abstractions as we\nwork through as we build agents as we do\nas we build like our streaming use case\namong the many other things even chat\nmemory we'll see there as well so line\nchain can act as the on-ramp to your AI\nlearning experience then what you might\nfind and I do think this is quite true\nfor most people is that if you if you're\nreally serious about AI engineering and\nthat's what you want to do like that's\nyour focus right which isn't for\neveryone for certain a lot of people\njust want to understand a bit of AI and\nthey want to continue doing what they're\ndoing and just integrate AI here and\nthere and maybe those you know if that's\nyour focus you might stick with line\nchain you know there's not necessarily a\nreason to move on but in the other\nscenario where you're thinking okay I\nwant to get really good at this I want\nto just learn as much as I can and I'm\ngoing to dedicate basically my you know\nmy short-term future of my career on\nbecoming AI\nengineer then line chain might be the\non-ramp it might be your initial\nlearning curve but then after you've\nbecome competent with line chain you\nmight actually find that you want to\nmove on to other frameworks and that\ndoesn't necessarily mean that you're\ngoing to have wasted your time with lang\nchain because one lang chain is a thing\nhelping you learn and two one of the\nmain frameworks that I recommend a lot\nof people to move on to is actually a\nlang graph which is still within the\nlang chain ecosystem and it still uses a\nlot of lang chain objects and methods\nand of course concepts as well. So even\nif you do move on from lang chain, you\nmay move on to something like langraph\nuh which you kind of to know lang chain\nfor anyway and let's say you do move on\nto another framework instead. In that\nscenario, the concepts that you learn\nfrom lang chain are still pretty\nimportant. So to just finish up this\nchapter, I just want to summarize on\nthat question of should you be using\nlang chain. What's important to remember\nis that lang chain does abstract a lot.\nNow this abstraction of line chain is\nboth a a strength and a weakness. With\nmore experience those abstractions can\nfeel like a limitation.\nAnd that is why we sort of go with the\nidea that lang chain is really good to\nget started with. As a project grows in\ncomplexity or the engineers get more\nexperience, they might move on to\nsomething like Langraph, which in any\ncase is going to be using lang chain to\nsome degree. So in either one of those\nscenarios, lang chain is going to be a\ncore tool in an AI engineer's toolkit.\nSo it's worth learning in our opinion,\nbut of course it comes with its, you\nknow, it comes with its weaknesses and\nit's just good to be aware of that. It's\nnot a perfect framework, but for the\nmost part, you will learn a lot from it\nand you will be able to build a lot with\nit. So, with all of that, we'll move on\nto our first sort of hands-on chapter\nwith Lang Chain where we'll just\nintroduce Lang Chain, some of the\nessential concepts. We're not going to\ndive too much into the syntax, but we're\njust going to understand a little bit of\nwhat we can do with it.",
  "transcript_chars": 8425,
  "ingested_at": "2026-05-15T04:38:13.901226+00:00",
  "source": "channel",
  "yt_meta": {
    "view_count": 7091,
    "like_count": 126,
    "channel_id": "UCv83tO5cePwHMt1952IVVHw",
    "categories": [
      "Science & Technology"
    ],
    "tags": [
      "python",
      "artificial intelligence",
      "langchain",
      "langchain course",
      "langchain tutorial",
      "langchain agent",
      "langchain chatbot",
      "langchain crash course",
      "james briggs",
      "james briggs langchain",
      "langchain project",
      "langchain agent tutorial",
      "langchain agent executor",
      "langchain agent example",
      "langgraph",
      "langchain langgraph",
      "langchain langsmith",
      "langchain langsmith langgraph",
      "langsmith",
      "langsmith tutorial"
    ]
  }
}