Self-reflective RAG with LangGraph: Self-RAG and CRAG

Поділитися
Вставка
  • Опубліковано 27 лис 2024

КОМЕНТАРІ • 43

  • @girijeshthodupunuri1300
    @girijeshthodupunuri1300 9 місяців тому +5

    Excellent breakdown! This video really helped me grasp the concepts, and it's one of the first Langchain videos that clicked for me. While Harrison's brilliance is undeniable, Lance, you're truly a great teacher.I love your approach of using diagrams to break down complex code into easily understandable logic. It really enhances the learning experience!

  • @preston_is_on_youtube
    @preston_is_on_youtube 9 місяців тому +5

    This is sooooo cool 🤯
    And I love that you guys are putting out all these educational videos - thank you!

  • @donb5521
    @donb5521 9 місяців тому +2

    great video Lance! The way you diagrammed the flow made it easy to understand the concepts.

  • @UGMurthy
    @UGMurthy 9 місяців тому +1

    Thank you. This video was informative. Your explanation was clear. For me personally the examples that us local LLMs are interesting. Thanks once again.

  • @advfuk
    @advfuk 6 місяців тому

    Thanks Lance for more one great tutorial! Really useful and easy to follow.

  • @ОлександрМельник-ь1у
    @ОлександрМельник-ь1у 7 місяців тому

    Lance, you are great! Thank you for your splendid video!

  • @jim02377
    @jim02377 9 місяців тому +1

    Great tutorial. Keep up the great work!

  • @hamzafarouk1538
    @hamzafarouk1538 9 місяців тому +1

    wow. Excellent demonstration.

  • @ramzirebai3661
    @ramzirebai3661 4 місяці тому

    Very Clear and informative . Thank you

  • @JoshuaMcQueen
    @JoshuaMcQueen 9 місяців тому +1

    Really great video! Keep 'em coming

  • @SimonMariusGalyan
    @SimonMariusGalyan 9 місяців тому

    Thank you for the great explanation!

  • @neonnftz
    @neonnftz 9 місяців тому +11

    Please record your videos in 1080p

    • @r.lancemartin7992
      @r.lancemartin7992 9 місяців тому +3

      (This is Lance, guy in the video.) Good feedback. I record w/ Loom. It was a UX issue w/ Loom where vids were not defaulted recorded to 1080p. Apologies!

  • @RossClaytor
    @RossClaytor 9 місяців тому +3

    Lance - isn't the workflow you outline here a DAG? The LangGraph docs are very explicit about not using LangGraph for DAGs - can you help us understand this nuance?
    "The main use is for adding cycles to your LLM application. Crucially, this is NOT a DAG framework. If you want to build a DAG, you should just use LangChain Expression Language."

  • @carloszelabueno432
    @carloszelabueno432 6 місяців тому

    Great video! I have a question about recent RAG papers which talking about RAG paradigms such as advanced RAG and modular RAG. Could we say that using LangGraph have we applying modular rag? I’m not sure in which paradigm fall self rag and crag.
    Thanks!!

  • @urd4651
    @urd4651 6 місяців тому

    very helpful! thank you.

  • @jzam5426
    @jzam5426 7 місяців тому

    Thank you for breaking this down!
    Is there a way to do cyclical/iterative agents when one produces an answer and the other checks the answer, and if not correct/satisfied with it, sends it back to the first agent to produce a better answer?

  • @anshubhatia6532
    @anshubhatia6532 7 місяців тому

    Hi, I have one question here related to the Retrieval Evaluator. Let's suppose we are not allowed to do a web search. Can we again play around with chunking and different retrieval methods?

  • @MLA263
    @MLA263 8 місяців тому

    Thanks for that ! Do you have a video that helps build an ui associated for prompting ?

  • @corybilyeu4490
    @corybilyeu4490 7 місяців тому

    Do you have any resources for deploying this type of solution on AWS/GCP?

  • @VaibhavPatil-rx7pc
    @VaibhavPatil-rx7pc 9 місяців тому

    Excellent !

  • @LordPBA
    @LordPBA 8 місяців тому

    What are other possibly ways to utilize this on a local database? I mean, if there are no relevant docs what it can do?

  • @GREENDAYROKS1
    @GREENDAYROKS1 9 місяців тому +1

    Very interesting, I assume this would have very long response times?

    • @r.lancemartin7992
      @r.lancemartin7992 9 місяців тому +1

      (This is Lance, guy in the video.) I'm running on a Mac M2, 32GB. Latency is ~5-10 sec for final generations.

    • @Slimshady68356
      @Slimshady68356 9 місяців тому

      ​​@@r.lancemartin7992 lance from langchain has a nice ring to it😅😂

  • @seansullivan6263
    @seansullivan6263 9 місяців тому

    is thsi still effective with conversational memory?

  • @siddharthchauhan3404
    @siddharthchauhan3404 7 місяців тому

    How to deploy langgraph using langserve, can you please help me do it?

  • @dex001
    @dex001 9 місяців тому +1

    I'm working on developing RAG Application on CSV File. It is working fine for some queries, but queries like
    1. Get total count of persons
    2. Get average salary
    These type of questions that include to search all the rows of the given document Is not working fine. Is there any fix I can do or Implementing RAG is not a correct option.

    • @anuragmishra-yu2yx
      @anuragmishra-yu2yx 9 місяців тому +2

      Are you using csv agent for the solution you built? if not, then you can try either csv agent or pandasAI.

    • @Tushii
      @Tushii 9 місяців тому +2

      What worked for me was converting my CSV / multi page xlsx into a minimal database
      And then query that database, SQL queries worked much better for me

    • @sridevigogusetty8371
      @sridevigogusetty8371 9 місяців тому

      It works for csv's but if it's pdf @tushii

    • @Tushii
      @Tushii 9 місяців тому

      @@sridevigogusetty8371 what would you like to do with your pdf ?

    • @sridevigogusetty8371
      @sridevigogusetty8371 9 місяців тому

      @@Tushii for example if I have some pdf which has financial transactions of i want to query how I will convert that to db

  • @janwillemaltink2216
    @janwillemaltink2216 9 місяців тому

    at around 13 minutes, based on your story i would assume that it wouldnt run web search as there are at least a few documents that are relevant, but as i understand even if 1 of the retrieved documents isnt relevant it will do a web search?

    • @r.lancemartin7992
      @r.lancemartin7992 9 місяців тому +1

      (This is Lance, guy in the video.) Yes, with this logic it will do web search if *any* documents are irrelevant.

  • @StuartAtkinson4467
    @StuartAtkinson4467 9 місяців тому

    Wooooo Skynet wooooooo