Build your own RAG (retrieval augmented generation) AI Chatbot using Python | Simple walkthrough

Поделиться
HTML-код
  • Опубликовано: 4 ноя 2024

Комментарии • 102

  • @Avra_b
    @Avra_b  23 дня назад

    Watch my latest video on Building AI apps with AI :
    ruclips.net/video/BViiyl5D8J0/видео.html

  • @shalempanthagani6034
    @shalempanthagani6034 Месяц назад +1

    Very crisp and on point. Forget about the folks complaining about audio clarity. I have totally understood what you tried to explain. Cheers.

    • @Avra_b
      @Avra_b  Месяц назад

      Cheers !

  • @funmitoadeyemi4647
    @funmitoadeyemi4647 Месяц назад +1

    not even five minutes into the video and the explanation I've gotten is clear and very arranged...thanks!

    • @Avra_b
      @Avra_b  Месяц назад

      Glad to hear ! ☺️

  • @ammarali4697
    @ammarali4697 Год назад +6

    videos keep getting better and better! 👍

    • @Avra_b
      @Avra_b  Год назад

      Thanks Ammar !❤️ So happy to hear such kind words ☺️

  • @Avra_b
    @Avra_b  9 месяцев назад +5

    I hope you guys have enjoyed this video 🫶 If you would like to know more about it , please have look to this article.
    Blog Post - medium.com/databutton/why-your-next-ai-product-needs-rag-implemented-in-it-9ee22f9770c8

  • @caysomusic
    @caysomusic 7 месяцев назад +2

    great video and info, only thing is the audio could be more clear, i think even a cheap USB mic off the internet would help quite a bit. Thanks and best of luck!

    • @Avra_b
      @Avra_b  6 месяцев назад +1

      Thanks ! I do agree. Already upgraded with a new mic - using that nowadays :) unfortunately, in this video - the sound gain was maximized while editing - my bad .

    • @caysomusic
      @caysomusic 6 месяцев назад

      @@Avra_b awesome, no worries and thanks for the reply!

  • @Outr8js
    @Outr8js 11 месяцев назад +7

    I subscribed. Constructive criticism to improve your channel as I think you provide good instructions. Study shows that bad audio quality discourages people to continue watching than bad video. The audio is already low to begin with but when you start moving away from the camera, which I think is the audio in as well, I could barely hear you. My 2cents. Thank you though for the content

    • @Avra_b
      @Avra_b  11 месяцев назад +2

      Really appreciate your feedback ! ☺️ I’m planning to invest more on the video production and it’s quality . Please do keep in touch and always feel free to leave such feedbacks 🫶🏽

  • @SwatiAgarw
    @SwatiAgarw 2 месяца назад +3

    The video content is really good. However pls do something about the audio. Please bring more clarity in audio.

    • @Avra_b
      @Avra_b  2 месяца назад +1

      Agree , noted ! Thanks for the feedback :)

  • @nutkut321
    @nutkut321 9 месяцев назад +3

    hey Avra_b, this is pretty helpful. Do we have any limitation on the size of the file or the number of files that I can upload while using above method?

    • @Avra_b
      @Avra_b  9 месяцев назад +1

      Yes with the file size there is a limitation from the front end side. However not with the number of files .
      But you can dwell with it under the hood with some config file ( check Streamlit documentation)

  • @AIforAll-DOMANA
    @AIforAll-DOMANA 9 месяцев назад +6

    Great explanation, great job

    • @Avra_b
      @Avra_b  9 месяцев назад

      Appreciate your kind words! 😊

  • @sidrakubra3634
    @sidrakubra3634 3 месяца назад +1

    Thank you for sharing this amazing video tutorial; it's really helpful. I have one question. If I use this chat and want to make a personalized adaptive chatbot-adaptive in the sense that the chatbot can adapt to user behavior and respond accordingly-would it be possible to use RAG to create such an adaptive chatbot?
    Please help me understand if I can create a personalized adaptive chatbot through this RAG framework, and if you could share any suggestions or tips for making this kind of chatbot, I would be very grateful.

    • @Avra_b
      @Avra_b  2 месяца назад

      Yes ! Custom prompt is one way to achieve that.

  • @Ash2Tutorial
    @Ash2Tutorial 20 дней назад +1

    Please improve the audio quality. The video is good, but the sound keeps getting inaudible as you move front and back.

    • @Avra_b
      @Avra_b  17 дней назад

      Hey ! Thanks for the feedback. I’ve now upgraded my mic set up . Hopefully that would help in making videos with better audio quality :)

  • @sayanghosh6996
    @sayanghosh6996 6 месяцев назад +1

    Really appreciate this video. If possible can you fix the microphone quality? Sometimes its hard to understand what you are saying due to the poor audio quality.

    • @Avra_b
      @Avra_b  6 месяцев назад

      Thanks ! I do agree. Already upgraded with a new mic - using that nowadays :) unfortunately, in this video - the sound gain was maximized while editing - my bad .

  • @deepanshugoel3790
    @deepanshugoel3790 9 месяцев назад +2

    Hi Avra i want to ask how can i restrict the chatbot to answer the question related to a particular space only? for example if i am building a chatbot for an e-commerce company and someone asks questions related to finance it should not answer that question.

    • @Avra_b
      @Avra_b  9 месяцев назад +2

      One way is to add gaurdrails ! I’ve played around with few .
      Alternatively, easiest workaround - fine tuning the Prompt. Prompt engineer with CAPs and some key words . Remind the LLM. It does work- but you need to burn some cash to make sure that it does work well .

    • @deepanshugoel3790
      @deepanshugoel3790 9 месяцев назад +1

      @@Avra_bthanks for helping out

  • @mangaimsc
    @mangaimsc 8 месяцев назад +1

    How do you update a PDF document stored in Pinecone/Chroma db? Can you do a video explaining how to Update /Delete Content stored in Vector d.

    • @Avra_b
      @Avra_b  8 месяцев назад

      Sure! Pinecone has this method called upsert. Also chroma has something like updating .. but def I can make one :)

  • @jooznagawa3650
    @jooznagawa3650 3 месяца назад +1

    Hi, I can't seem to run the Python codes in Databutton and the front end language for Databutton is React now.
    Can I get some guide on how I can run the codes in your Github? Maybe in vscode since I think Databutton has a different interface now.

    • @Avra_b
      @Avra_b  2 месяца назад

      That’s right! You can run the same code using Streamlit framework.

  • @pritampadhan5977
    @pritampadhan5977 4 месяца назад +4

    You can work a bit regarding audio. Btw very helpful

    • @Avra_b
      @Avra_b  2 месяца назад

      Agreed !

  • @teguhbagussurya7900
    @teguhbagussurya7900 6 месяцев назад +1

    Hi, Thanks for your tutorial!
    Can it be used for a complete beginner ?

    • @Avra_b
      @Avra_b  6 месяцев назад

      I would say intermediate. There are couple of things here . A) front end with Streamlit B) using Python stacks
      C) introduction to the RAG concepts
      Would be great, if someone starts with easy tutorial . Like building a simple chatbot with open ai and streamlit . Then slowly move to other usage from there.

  • @Ovrlordz
    @Ovrlordz 8 месяцев назад +1

    The content is very good, Thanks for the video. You're microphone needs to be updated though.

    • @Avra_b
      @Avra_b  8 месяцев назад +1

      Thanks ! Sorry for the microphone issue . Appreciate that you went through the video despite of it :)

  • @zahraahawie3707
    @zahraahawie3707 Год назад +1

    I have seen the new OpenAI updates for building our custom GPT.. is it easy to be embedded on our website? And what it differs from building RAG chatbot as in the video.. Thanks for your knowledge. 👍👍

    • @Avra_b
      @Avra_b  11 месяцев назад +1

      If I’m not wrong at present , you can create and share GPTs only if you are under paid subscription. Let’s say I create a public GPT and share the link with you . You can only access if you also have paid subscription ( I.e. GPT plus ) .
      This is a great question ! I personally found the current state of GPT is not at all up to the mark of real RAG . You can click and upload files and provide some custom instructions . But you can’t further improve it - since you don’t know what’s happening under the hood ( like chunking / embeddings / system prompts all additional attributes ) . Let me know if this answers your question. And thanks again :)

  • @amethyst1044
    @amethyst1044 10 месяцев назад +1

    Very helpful video, thank you for the effort !
    I am going to try it out and see how it's perform on complex PDFs (have images, tabular Data.. etc.)
    Do you think it'd be performant ? If else, how can I enhance it so that it's retrieve complex data from tables and structures (mostly tables)
    Keep up the good work !!

    • @Avra_b
      @Avra_b  10 месяцев назад +1

      Thanks ☺️ I highly doubt how good it would be with images - but do give it a try and let us know :) Happy to hear 🤗

    • @amethyst1044
      @amethyst1044 10 месяцев назад

      @@Avra_b Thank you.
      I went throught the code, can you please tell me what is this chunk of code responsible of ?
      for chunk in st.session_state.chat.send_message(prompt):
      text = chunk.choices[0].get("delta", {}).get("content")
      if text is not None:
      response.append(text)
      result = "".join(response).strip()
      botmsg.write(result)
      Is it to keep a history of the conversations, I don't really understand it.
      I'd reatly apreciate the help !

  • @nikhilhero9131
    @nikhilhero9131 10 месяцев назад

    Great explanation, can you please tell me how did you learn all these things ??

    • @Avra_b
      @Avra_b  10 месяцев назад

      Hey thanks ! Glad to hear . Staying up to date with the latest happenings. Reading blogs and articles.

    • @nikhilhero9131
      @nikhilhero9131 10 месяцев назад

      @@Avra_b thanks for replying can you suggest me some blogs and articles sites where I can learn all those things

  • @xspydazx
    @xspydazx 6 месяцев назад +1

    Question : once loading a vector store , how can we output a dataset from the store to be used as a fine tuning object ?

    • @Avra_b
      @Avra_b  6 месяцев назад +1

      If I understood you correctly - you would like to fine tune the response obtained from the vector store ?

  • @arupmukherjee3403
    @arupmukherjee3403 Месяц назад

    Can I develop RAG application locally in i7 16 gb ram cpu based laptop? I tried with chroma and ollama llama2 model but getting memory issues

    • @Avra_b
      @Avra_b  Месяц назад

      Should be . Not sure about your issue. Can you elaborate a bit more ?

  • @ahmedmusawir
    @ahmedmusawir 9 месяцев назад +2

    Audio volume goes up and down like crazy ... impossible to focus on the content ...

    • @Avra_b
      @Avra_b  9 месяцев назад

      Thanks for your feedback ! Apologies for this ! Will soon invest on fixing the sound issue :) stay tuned

  • @BackupSte
    @BackupSte 9 месяцев назад +1

    It works only with a pdf document ? if I use an url with a text addon too it's ok or not ?

    • @Avra_b
      @Avra_b  9 месяцев назад

      The implementation shown in the video is meant for PDF files . I’ve another video showing how to work with URL as well . Please check that out :)

  •  9 месяцев назад +2

    Love the content but record the audio offline might help.

    • @Avra_b
      @Avra_b  9 месяцев назад

      Thansk ! I agree . Stay tuned ❤️

  • @swetharangaraj4521
    @swetharangaraj4521 10 месяцев назад +1

    what is the process if i what to query chat from cloud mangoDB using llm and RAG

    • @Avra_b
      @Avra_b  8 месяцев назад

      Very much possible . Extract the data . Pass it to the LLM . You can fetch and have it in form of Dataframe . Then use pandas Agent from Langchain or Lamaindex .

    • @swetharangaraj4521
      @swetharangaraj4521 8 месяцев назад

      @@Avra_b how to do interconnected for multiple collection within the dataset

  • @ChronicleContent
    @ChronicleContent Год назад +1

    the difficult part will be to create a saas app with user sign in, subscription plans, landing page etc.

    • @databutton
      @databutton Год назад +2

      We'd love to talk to you about that. 💜

  • @whynesspower
    @whynesspower 7 месяцев назад +1

    Really outdated. Using openai==0.28 chat completions (outdated)
    Convert st.cache_data to st.cache_resource
    Streamlit doesn't supports that old one anymore

    • @Avra_b
      @Avra_b  6 месяцев назад

      Agree ! But at the same time, you need to consider that this entire niche is very dynamic - endpoints are getting updated and deprecated almost every weeks. I would rather take this idea / workflow and adapt it to the new endpoints .
      Alternatively , I’m really thinking of making a new video on it :)

  • @loncewyse696
    @loncewyse696 9 месяцев назад +1

    Your recording is using some kind of automatic gain control. The sound is terrible - impossible to watch!

    • @Avra_b
      @Avra_b  9 месяцев назад

      Yeah you are totally right here! Extremely sorry for this . Can’t do much now. Will make a new video and would def try to not repeat with this :)

  • @SankarJankoti
    @SankarJankoti 7 месяцев назад

    Hey, can you pleae tell me where i can get a class to learn all these

    • @Avra_b
      @Avra_b  6 месяцев назад

      I'm sure there are lot of online courses covering similar topics. Please make a google search. However, I would encourage you to try to grasp as much as possible from the openly available resources - Documentation of such tools / RUclips / Medium blogs / Github / . Free online resources and lot of scope for learning from there. Best wishes :)

    • @SankarJankoti
      @SankarJankoti 5 месяцев назад

      @@Avra_b thank you

  • @anshgoyal3671
    @anshgoyal3671 Месяц назад

    Can you tell me which font is this ?

    • @Avra_b
      @Avra_b  Месяц назад

      Sure, but Font style of ?

  • @asvyoutube
    @asvyoutube 8 месяцев назад

    How canwe import this model to work in whatsapp ? Or Instagram?

    • @Avra_b
      @Avra_b  8 месяцев назад

      Not sure :(

  • @kethavathaadarsh5604
    @kethavathaadarsh5604 Месяц назад

    what after building this. Where can i sell this yo? I can teach you multiple Ai products better than this but telll me are you able to make money on them.

  • @ArthurThawne
    @ArthurThawne 6 дней назад

    How can I do this with JavaScript

    • @Avra_b
      @Avra_b  6 дней назад

      Langchain has a Typescript library. Probably you can give that a try ?

  • @harishnarrendranr2785
    @harishnarrendranr2785 5 месяцев назад +1

    Try Improving the audio quality

    • @Avra_b
      @Avra_b  5 месяцев назад

      Agreed! 😊

  • @aceLalou
    @aceLalou 5 месяцев назад

    how can i use the latest version of chat gpt?

    • @aceLalou
      @aceLalou 5 месяцев назад +1

      and how can i use storage to read documents? the script only read the file that we load in the conversation but i want to create a real data base wwith lot of pdf that i need him to read them in this data base

    • @aceLalou
      @aceLalou 5 месяцев назад

      sorry to ask you so many questions but how can i add an ocr technique for pdf

    • @Avra_b
      @Avra_b  2 месяца назад

      You mean the api ? Change the model name

  • @sharduljadhav2269
    @sharduljadhav2269 8 месяцев назад +1

    Is there any free alternative for openAI ?

    • @Avra_b
      @Avra_b  8 месяцев назад

      Yes just released a video with Mistral AI. It’s running with the Groq LPU

  • @deepak_kori
    @deepak_kori 7 месяцев назад +1

    voice can be improved but good video

    • @Avra_b
      @Avra_b  7 месяцев назад

      Yep, agreed thanks! But yeah, have upgraded the mic 🎙️ now - so more hopefully the audio will be better in future , stay tuned :)

  • @lishyloshy5859
    @lishyloshy5859 4 месяца назад

    'm facing pydantic.field error

    • @Avra_b
      @Avra_b  2 месяца назад

      Tell us a bit more about the error .

    • @lishyloshy5859
      @lishyloshy5859 2 месяца назад

      @@Avra_b it resolved it was my system's error

    • @lishyloshy5859
      @lishyloshy5859 Месяц назад +1

      @@Avra_b actually I developed RAG app through your and some more different videos but the thing is in one video i came to know i have to subscribe to openai for gpt pre rain chatbot and I was thinking I was missing some libraries which made my laptop out of storage .
      Not gonna lie I like your this video and I'll InshshAllah going to built RAG Chatbot i future cause I'm studying LLM, Generative AI.
      my problem was "not able to get a free pre-trained model for chatbot"

  • @vikassahu6896
    @vikassahu6896 7 месяцев назад

    Why your ping is high? 😂

    • @Avra_b
      @Avra_b  7 месяцев назад +1

      Agreed ! The editing went wrong , increased the sound gain way too much - apologies 🥲

  • @YouToYoub
    @YouToYoub 9 месяцев назад +1

    Very incoherent explanation. Demo is great though.

    • @Avra_b
      @Avra_b  9 месяцев назад

      Hey thanks for your candid feedback. I appreciate it ☺️

  • @sandeepnahta1
    @sandeepnahta1 13 дней назад

    Video sound is pathetic. You need a better mic bro

    • @Avra_b
      @Avra_b  12 дней назад

      I agree. Sorry for the troubles.
      Now we have a new one 🎙️btw , lemme know if it’s better :)

  • @pecke86
    @pecke86 8 месяцев назад +4

    I can't hear you mate. Fail

    • @Avra_b
      @Avra_b  8 месяцев назад

      Ah noo :( I’m aware of this sound issue, apologies for this experience

    • @technologicalvivek7510
      @technologicalvivek7510 8 месяцев назад

      TRUE sir he is mate .What does he want to explain to nobody know 😂

    • @audacityhour3104
      @audacityhour3104 3 месяца назад +1

      @@Avra_bthe audio is bad but I got the gist. Saving it as a reference point. .. maybe u can edit clean up the audio with fl studio or something n then repost it