Improve your Generative AI Application with RAG

Поделиться
HTML-код
  • Опубликовано: 23 ноя 2024

Комментарии • 11

  • @LiebsterFeind
    @LiebsterFeind 8 месяцев назад +6

    Summary: Embed your text or documents in a vector database, create an embed representation of the query, pre-filter the content fed to the LLM in prompt by doing an N-best vector database query first.
    Excellent video, more like these please. Do some Node.js ones, don't be a Python elitist! (kidding).

  • @YUVRAJVAGHELA-zi6dj
    @YUVRAJVAGHELA-zi6dj 8 месяцев назад +6

    Very insightful, Thank you for the explanation.

  • @andreseungyeobsong4899
    @andreseungyeobsong4899 7 месяцев назад +1

    I have a question!(And I'm a newby to AI stuff. I'm very sorry) I'm currently using Lambda. If I put in lots of lots of information in S3 bucket, how long does it take to vectorize the stuff in S3 bucket and put it into the memory? (Because she mentioned that she put in the AWS documentation to S3 bucket, and I think that should be a pretty hefty amount of data!)

    • @mikegchambers
      @mikegchambers 7 месяцев назад +3

      Hey! You might have guessed this, but the amount of time will depend on the amount of data you have. When the knowledge base is finished being created, the Status of the knowledge base changes to Ready. But I should mention... you said "vectorize the stuff in S3 bucket and put it into the memory"... just to be clear knowledge base puts the data in to a vector database for you and you have some options there. You then perform queries on that data, and or use a RAG architecture. You can call that from a Lambda function if you like.

    • @TheDrWolfram
      @TheDrWolfram 4 месяца назад

      I'm not completely sure of how AWS Bedrock handles this behind the scenes, but just for the information: you don't need to necessarily load all the vectors into the memory. He was doing that in the demo because it was faster in that small scale context. If you are dealing with large amounts of data, it could be more cost-efficient to persist the vectorized data somewhere and just access or update the information as needed. One native AWS option for that is Kendra, but you can use other third-party cloud-based or open-source solutions.

  • @miguelhermar
    @miguelhermar 4 месяца назад

    Hey! Where can we find the code explained here?

  • @JoshCrosby1234
    @JoshCrosby1234 5 месяцев назад

    what does the 1536 represent, where did you get that?

  • @ManishYadav0719
    @ManishYadav0719 Месяц назад

    ❤❤

  • @Vinay-bt8ug
    @Vinay-bt8ug 6 месяцев назад

    Hey I am a newbie here
    Trying to develop a Chatbot with the data that we have sql db. What’s the best approach here !
    Thanks

    • @awssupport
      @awssupport 6 месяцев назад

      Hi there! 👋 Our scope for tech assistance is limited on this platform, but you can reach out to our awesome community of industry gurus here: go.aws/aws-repost. 🤓 Check out the other channels on this page if you still need further assistance: go.aws/get-help. ℹ️ ^RW

  • @java-aws-techie
    @java-aws-techie 3 месяца назад

    Is this code on Github? If yes please ping the URL.