LangChain vs. LlamaIndex - What Framework to use for RAG?
In this video we will explore how to perform basic RAG with LangChain and LlamaIndex. How different are they? Which is better? In this video we cover the BASICS.
Code: github.com/Coding-Crashkurse/...
Timestamps:
0:00 What is RAG?
01:57 Loaders - Langchain
4:23 Loaders - LlamaIndex
5:16 Splitters - LangChain
6:08 Splitters - LlamaIndex
7:29 Indexing - LangChain
8:40 Indexing - LlamaIndex
10:27 Chains - LangChain
13:22 Chains - LlamaIndex
Пікірлер: 41
I am subscribing and turning on notifications because of how well you explain stuff in this video! Kuddos!
@codingcrashcourses8533
21 сағат бұрын
@@maxtayebwa8987 thank you :)
The best, always.
This is a good video. Thank you.
Amazing video. It would be even better if you can compare agents in both frameworks.
felt like langchain are more straight forward , but ofcourse perhaps there are some value for how LlamaIndex are implemented
good
Hello, thank you so much for your comparison in the video! I also have a small request - could you possibly make a prompt engineering video tutorial for both LangChain and Llama-Index, especially the use of LCEL in LangChain?
@codingcrashcourses8533
4 ай бұрын
One or the other :). I would rather make a Video ok LCEL
@limjuroy7078
4 ай бұрын
@@codingcrashcourses8533 great 👍
Great video. Subbed. I have a question , recently i was trying RAG application for german language using mistral , llama cpp , chroma and local embedding . However i found out that using llama index i couldnt even retrieve any documents and returned no answer . Has it happened to you that it didnt recover any chunks so as to say ?
@codingcrashcourses8533
3 ай бұрын
No - the standard is that your retrieve the top-k documents. You sure you indexed correctly? Normally you always retrieve documents, no matter what.
Amazing content as always , can you make a video on RAPTOR, it is getting popular for long context llms as an soft alternative for RAG
@codingcrashcourses8533
4 ай бұрын
Thanks for making me aware of that
2 questions. 1st: How is Hyastack compared to these two?, and 2nd: Which one is more asked in Job openings?
@codingcrashcourses8533
4 ай бұрын
I tried Haystack and ditched it because it forced you to install the whole transformer stack. That was ~1 year ago. Regarding the second question: No idea about job markets.
@sanjayojha1
4 ай бұрын
@@codingcrashcourses8533 I have tried their beta 2.0, so far it looks good.
@yinxing418
4 ай бұрын
Switched to Haystack and never looked back. Repo is way easier for me to understand, and you can write your own components. Just compare all the frameworks yourself and see what makes sense.
Yep, love it. We need more Udemy courses! Best teacher ever.
@codingcrashcourses8533
4 ай бұрын
Haha thank you. I am already thinking of a new course. Probably an advanced course. But not Sure about the way i want to Set it up
@swiftmindai
4 ай бұрын
I suggest you do an advance course with a specific end to end use case or project like you do in real world (for eg. specific chatbot for retails or ecommerce or anything) that can be implemented in production with very minimum change. That will level up skill to professional level for many of us who follows you here. Coursera and some udemy as well offers similar which they call it capstone project if I'm not mistaken. But since most of us here are comfortable with the way you explain, I'll be the first one to purchase your course :)
@codingcrashcourses8533
4 ай бұрын
@@swiftmindai I will probably make an advanced RAG course, since I already got a course on Prod with Azure. I don´t like to cover multiple complex areas in one course/video :)
Very cool. But how can me make it “learn on the go”. Meaning if someone asks a specific question and there is no answer. Like where is the bathroom in the restaurant. How to setup your notify a human to help, and then to add that Q and A to the RAG or Langchain, and then how to rebuild the model so it knows?
@codingcrashcourses8533
3 ай бұрын
It depends. Add a "contact human" and then probably collect the answers and index them again. There are many different ways to add upon that
@DataMind-xd7xm
2 ай бұрын
@@codingcrashcourses8533 how about using a RLHF system from the feedbacks instead?
i dont want to use chromedb how can i avoid it i want to store it in json or pickle or something else
@codingcrashcourses8533
2 ай бұрын
Use FAISS for that. FAISS allows you to pkl your vectorstore
Langchain also has high level chains
@codingcrashcourses8533
3 ай бұрын
Yes, but these legacy chains and the chain interface is not the default.
create a chatbot for HR to filter the best candidate among the resumes
@codingcrashcourses8533
4 ай бұрын
Interesting idea ;)
@aamirsec8305
4 ай бұрын
@@codingcrashcourses8533 can i have a chat with you
@aamirsec8305
4 ай бұрын
@@codingcrashcourses8533 whats your email sir
Very interesting video...we have just landed a big client that has thousand of documents. Would it be possible for you to help us with this project man, or at least offer your expertise on some consulting session ?
@codingcrashcourses8533
3 ай бұрын
Thanks for the request, but I have to decline since I am not allowed to done any consulting related work in my side business.
collection "xyu"😶🌫
@codingcrashcourses8533
3 ай бұрын
Hm?
@coursachGPT
Ай бұрын
😂😂
@coursachGPT
Ай бұрын
@@codingcrashcourses8533that's a curse word in Slavic languages
@codingcrashcourses8533
Ай бұрын
@@coursachGPT thanks for clarification :D
@coursachGPT
Ай бұрын
@@codingcrashcourses8533 good video btw! tried langchain with chromadb but couldn't install langchain-chroma lib, some problems with Microsoft visual runtime. Tried lancedb but couldn't figure out how to use it with local embeddings and llm😂 not my day, need to spend some time to learn all that. Thank you for your content