Llama3 local RAG | Step by step chat with websites and PDFs

Поділитися
Вставка
  • Опубліковано 31 жов 2024

КОМЕНТАРІ • 50

  • @MrRexzhen
    @MrRexzhen 5 місяців тому +3

    excellent job. this is exactly what I am looking for. Will try it out.

  • @thedatamaster
    @thedatamaster 6 місяців тому +2

    This video is one of the most beautiful ones I've ever seen.❤

    • @phidata
      @phidata  6 місяців тому +2

      Thank you ❣this is motivation to make more videos

  • @orhancavus2229
    @orhancavus2229 6 місяців тому +1

    Exceptional tutorial, thank you very much, it works very well, installation took me several attempts but at the end worked thanks a lot Ashpreet

  • @sannguyen1008
    @sannguyen1008 6 місяців тому +2

    Awesome. Thanks for a great video and building such a great tool. Please keep it up and build a larger community by integrating with popular framework like llama index or dspy.

    • @phidata
      @phidata  6 місяців тому +1

      Thanks San! I'm on it :)

    • @mrrohitjadhav470
      @mrrohitjadhav470 6 місяців тому

      @@phidata Yes, I have been searching for how to use dspy to produce datasets from PDFs. Using ollama (local llm) to facilitate this process would be fantastic.

  • @AlexanderYevseyevich
    @AlexanderYevseyevich 4 місяці тому

    Fantastic stuff and very practical. thank you

  • @BorisDuran
    @BorisDuran 6 місяців тому +3

    Hi @phidata, thanks for sharing such a nice project. I followed your instructions and everything went as in the video. However, when testing it I found out that some URLs are loaded without problems whereas others are not loaded, I just get a "Could not read website" message on the UI and nothing on the console. Are there any restrictions or configurations that we should be aware for reading URLs?
    Another thing, how can I read TXT files also and not only PDFs? something like adding support for other types of files to be ingested... where in the "phidata" repository is this done?

  • @BenjaminTovarCisneros
    @BenjaminTovarCisneros 3 місяці тому

    Excellent work, this tool is fantastic. Do you know if there is a limit on the amount of PDFs that I can use to feed the model? Do I have to upload all the PDFs every time I restart the tool or are all the PDFs stored persistently in the knowledge base? Even if I restart Docker . Again, this tool is amazing, thank you.

  • @chjpiu
    @chjpiu 6 місяців тому +1

    Excellent video. Could you consider supporting another format, such as doc, markdown, or CSV… in the future? Thanks a lot

    • @phidata
      @phidata  5 місяців тому

      absolutely :)

  • @irkedoff
    @irkedoff 6 місяців тому +2

    This is great. Thanks.

  • @mukeshkund4465
    @mukeshkund4465 5 місяців тому

    Without using ollama can we build this ?? If yes can you show us the way

  • @sebacisterna
    @sebacisterna 6 місяців тому +1

    What do you use to record your screen and zoom keeping such good quality?
    thanks!

    • @phidata
      @phidata  6 місяців тому +1

      I use screenstudio, its very very good :)

  • @franckmigone2433
    @franckmigone2433 6 місяців тому +2

    Very great. Since i'm French speaker can I use it with a French Version of LLAMA 3 ?

    • @phidata
      @phidata  6 місяців тому

      Yes you can sir, give it a try and let us know how it goes

  • @deeplearningdummy
    @deeplearningdummy 5 місяців тому

    Thanks for the video. Have you tried AirLLM? Allows you to run Llama 3 70B uncompressed on a 4GB GPU. Can you do a tutorial on that? Also, have you tested any of the quantized versions of Llama 3 70B?

  • @AndresSolar-y3g
    @AndresSolar-y3g 5 місяців тому

    finally a good channel.

  • @saschacontes2305
    @saschacontes2305 11 днів тому

    How is it going? It’s too quite on this channel.

  • @kylewilliams4721
    @kylewilliams4721 4 місяці тому

    I've installed docker desktop, but the code pasted into the terminal does not work. I don't see any instructions on how to link docker to the app. Please help!

  • @six1free
    @six1free 3 місяці тому

    I want to use mysql for both the knowledge base and storage but i don't see an appropriate tool and don't want to hack and slash my own without understanding your underlying structure
    ... prety please? :D

  • @luisfernandomoraes9778
    @luisfernandomoraes9778 6 місяців тому

    Which theme are you using, sir?

  • @alfonsor.6722
    @alfonsor.6722 6 місяців тому

    HI there, Awesome Video. Thanks for it. Question: How can I do, if I want to upload to a hosting service such as pythonanywhere, Azure, etc.? Any thoughts I appreciate it. 👍

  • @JTIAPBN
    @JTIAPBN 6 місяців тому +1

    is the speed of the response faster than groq local rag from your previous video? thanks for the video. :)

    • @phidata
      @phidata  6 місяців тому +2

      Groq is definitely faster, but also here we're running the 8b model whereas on Groq we're running the 70b model. Also on local it can depend on how overloaded my system is, 3-4 pycharms, a couple of google chromes and the models get slower.
      Have you tested it locally vs groq?

    • @JTIAPBN
      @JTIAPBN 6 місяців тому +1

      @@phidata not successfully yet. i got the error"UnicodeEncodeError: 'ascii' codec can't encode character '\xa0' in position 7: ordinal not in range(128)" while using nomic-embed-text for the embedding as shown in your cookbook/llms/groq/rag/

    • @phidata
      @phidata  6 місяців тому

      @@JTIAPBN Please can you try with another document? The error seems to be with a character encoding in your file. Here's more details: stackoverflow.com/a/9942822

    • @JTIAPBN
      @JTIAPBN 6 місяців тому

      @@phidata I did try others documents but no luck

  • @weirobin7523
    @weirobin7523 2 місяці тому +1

    Great sharing. I tried following your steps, but got error when I read the pdf file. 'StatementError: (builtins.ValueError) expected 768 dimensions, not 0', what might be the reason for this? thx.

    • @Tony-cw6om
      @Tony-cw6om Місяць тому

      I got a similar error, have you find a solution for this?

  • @anindabanik208
    @anindabanik208 6 місяців тому

    i follow all the steps,but when i start app,it show only llama3(no user input option are shown) only one option just llama3.

  • @babuinba
    @babuinba 6 місяців тому +2

    If llama installed in different server, how to open the URL

    • @phidata
      @phidata  6 місяців тому

      you can set the `host` value on the llm=Ollama(host="your url"). read more: docs.phidata.com/llms/ollama#params

    • @babuinba
      @babuinba 5 місяців тому

      @@phidata separate video to call URL and get the result should help everyone. By the way loved your videos very much

  • @impactsoft2928
    @impactsoft2928 6 місяців тому

    how to host this app at internet

  • @sfl1986
    @sfl1986 6 місяців тому

    Ive been trying to edit the llama3 temperature cuz it tends to hallucinate a lot, where should I go to edit it?

    • @phidata
      @phidata  6 місяців тому

      You can send it to the LLM like: `llm=Ollama(model="llama3", options={temperature: 0.1}),`
      Docs: docs.phidata.com/llms/ollama#:~:text=Timeout%20for%20requests-,options,-Dict%5Bstr%2C%20Any

    • @sfl1986
      @sfl1986 6 місяців тому

      @@phidata Im using the groq api with export GROQ_API_KEY=***, where can I edit temp?

  • @atozaboutdata
    @atozaboutdata 6 місяців тому +1

    does this work in Windows 10?

    • @phidata
      @phidata  6 місяців тому

      Yes it does :)

  • @shackinternational
    @shackinternational 6 місяців тому +1

    The problem is 200 MB file limit… otherwise it is nice

    • @phidata
      @phidata  6 місяців тому +1

      You can increase the limit :) this is just set on the front-end (streamlit) but you can do as big a file as you like

  • @drsimonpark
    @drsimonpark Місяць тому

    Llama3 similar to chatgpts

  • @aajn88
    @aajn88 6 місяців тому +1

    Hey! Thanks for the video. I'm hitting the following error:
    "ERROR (psycopg.OperationalError) connection failed: FATAL: no pg_hba.conf entry for host
    "", user "ai", database "ai", no encryption "
    Any idea how to solve it? Was anything missing in the readme file? I don't see where the file is being configured in the docker container

    • @aajn88
      @aajn88 6 місяців тому

      I was able to solve the issue by going to `/var/lib/postgresql/data/pgdata/pg_hba.conf` in the Docker container files and adding the following line:
      `host ai ai /32 trust`
      Make sure to replace `` with the IP that is shown in the error.
      Then restart docker and refresh the page. It should be usable now.

    • @phidata
      @phidata  6 місяців тому +1

      Great you were able to solve this!! such an odd error :/