Build a Private Chat My PDF Data RAG System with LangChain, Ollama, FAISS Vector Store & Llama 3.2

[@manforpeople]

Great explainer video. Thanks a lot for uploading this.

[@joohuynbae5084]

i 've watched a bunch of videos about rag and kept running into problems. but your video is absolutely the best. no bugs at all. thank you so much, sir

[@KGPTalkie]

Thank you so much. I have published my LLM course on Udemy. You can get that here. www.udemy.com/course/ollama-and-langchain/?referralCode=7F4C0C7B8CF223BA9327

[@amfiz9735]

Thankyou sir I had tight deadline of 3 days and I was panicking but your video helped me a lot and now I have successful rag implemnetation.

[@KGPTalkie]

Thank you

[@viebell1440]

Can you give the source code please

[@KGPTalkie]

@viebell1440 please check codes at my github account github.com/laxmimerit

[@slal]

Excellent work on RAG, Laxmi Kant! . Completed in one-go without any single bug.

[@KGPTalkie]

Thank you

[@sreyabasak9677]

Thank you sir. Works smoothly and wonderful explanation.

[@JasonSmith-og5uf]

dude, what a great video! With all the information, you somehow made it easy to follow. Bravo sir!

[@KGPTalkie]

Thank you

[@pratik2617]

@@KGPTalkie I have a laptop of 8gb ram and entry level gtx graphics card. will i able to run llama3.2 ?? @KGPTalkie

[@KGPTalkie]

How much memory your gpu have?

[@yoottananiema-um7071]

the best video no bug, no ploblem

[@KGPTalkie]

Thank you 🥰😍

[@afzal_mansury]

This is one of the best video Available on Yt wrt to RAG, headsoff to you bro.. The way you described is absolutely great... It would be great if you add chapter, that would be more helpful for audience... At the end I want to say big thank you.. And also your channel is underrated... Continue this great work bro ❤

[@mobycrazy99]

good work and explanation

[@AtalUpadhyay]

Excellent and very Informative.

[@KGPTalkie]

Thank you so much

[@alexramos587]

Good video.

[@KGPTalkie]

Thank you

[@jyothishkumar.j3619]

How did you learn ? What resources do you recommend for beginners?

[@MykolaDanylchuk]

Thank you for the instructions. Would the picture in PDF be read by the LLM or not?

[@prasadgenesys6782]

Great work!, please make a video on finetuning llama with our own data as dataset.

[@dps158]

Thanks sir , Very Nice video. Can you please create the one course on Langchain from beginner level to advance level. I am also looking the contents on Agent.

[@KGPTalkie]

Yes sure. Thanks. It's already in development.

[@OKORIEREX]

Hi Coach, I hope you're doing well! I wanted to take a moment to congratulate you on creating your Udemy course. it’s no small feat to turn your expertise into something that can inspire and educate others. Your dedication and effort are truly commendable. I’ve come across your course, and it’s clear that you’ve put a lot of thought into delivering value to your students. It’s always exciting to see instructors who are passionate about sharing their knowledge. I’d love to hear more about your journey! Best regards, Rex

[@tusharshrikhande3263]

Sir is there any library to process tables differently from the pdf if i use this method the retrieval of text inforamation is working properly but the tables are getting messed up and not able to retrieve the information that is stored in the table from the pdfs? For a semi-structured data a pdf that contains(text and tables both)?

[@KGPTalkie]

Hi, Yes you can use docling from IBM or MarkItDown from Microsoft. These libraries are released recently on github.

[@tusharshrikhande3263]

Sir please can you make a video on that I tries using docling by referring through different videos but they are giving different kind of errors

[@tusharshrikhande3263]

Hi sir I have huge pdf for my project if I take the chunk size to be 1000 then the total chunks are close to 1000 for that ids = vector_store.add documents is taking a lot of time the maximum size is for chunks character is 8192 correct so is it okay if I increase my chunk size and decrease the chunk overlap? Or is there another way to reduce the time it took almost 10 to 15 minutes

[@tusharshrikhande3263]

The pages are close to 550 in the pdf

[@KGPTalkie]

@tusharshrikhande3263 hi, Document indexing is one time process and I think it's fine. You can leverage parallel execution to speed it up. You can use python threading for parallel computing.

[@KGPTalkie]

8192 is context size in tokens which is around 25000 chars.

[@tusharshrikhande3263]

@@KGPTalkie Ok thank you

[@joohuynbae5084]

how can i find the base_url in my computer? i see you just look at your screen and type in your base_url. how to do that?

[@KGPTalkie]

It's same for everyone. You just need to run ollama and it automatically starts at that url. Let me know if you still have an issue.

[@joohuynbae5084]

@@KGPTalkie thank you so much 🥰

[@KGPTalkie]

@joohuynbae5084 I have published a udemy course on Langchain today. You can register here www.udemy.com/course/ollama-and-langchain/?referralCode=7F4C0C7B8CF223BA9327

[@robertog3663]

i am watching this video with huge faith. All the others video i've found gave only errors

[@KGPTalkie]

Let me know if it works

[@robertog3663]

@@KGPTalkie thank you. For the moment i switched to the easiest "Build Your Own Chatbot with Langchain, Ollama & LLAMA 3.2 | Local LLM Tutorial"

[@robertog3663]

@@KGPTalkie would be interesting to know all the packages needed to run it, my mystake, you specify this exactly on the first line, after first # IT WORKED! Thank you Sir, amazing job and patience. Is there a chance to transform this in a chat like ollama does with open webui? i mean a kind of a mix between this project and your chat bot, to interact with pdf files with the chat

[@KGPTalkie]

@robertog3663 github.com/laxmimerit/LangGraph-and-Ollama/blob/main/requirements.txt Plus Streamlit

[@Tesla1]

im new to here nd im subscriber and can you give me one reply please and im having student records of 4000 and i wanna ask ai model that i want particular amruth phone number and it should give me number and if there are two amruth then ai model should tell which amruth you want 8 class or 9 class or give me surname of him like that can you give me wt should i learn or this is correct video for me

[@KGPTalkie]

Yes. This can be done. Do you have data in csv format?