Azure OpenAI BYOD: ChatGPT with Your Own Data!

[@kjvasquez]

I have learned more from you in 15 minutes than following the entire AI-102 Microsoft Training, new subscriber here

[@DanWahlin]

Glad to hear it was helpful! Thanks.

[@maciejnowosielski714]

Awsome tutorial. I uploaded all my sofrware manuals into OpenAI and now OpenAI opens questions from my customers. Great work

[@ayesha365]

If you use ai search it will cost you huge and also I believe you need to pay for storage as well

[@DevTango]

My POC was completed in less than 30 minutes, thanks to you.

[@JohnnysaidWhat]

You just made me look like a rockstar at my job today. Thank you so goddamn much.

[@DanWahlin]

That's awesome! So cool to hear that.

[@bharatruparel9424]

Well done Dan. Please keep them coming. Have a great thanksgiving.

[@DanWahlin]

Thanks!

[@RyanPuffer-r1n]

Dan, you're the man. Answered some questions I had about the semantic search error perfectly. Subbed and liked!

[@Woodborn]

Awesome video, lots of condensed value, thanks!

[@kaisquared90]

Great info. The fast forward with music is so good!

[@lembatz12]

Great explanation and pretty straightforward, that's all I need. Thanks man

[@AndreasNeumann-l8j]

Thanks, Dan. This was exactly the information I was looking for.

[@JSmiley-i4c]

A+ Sir!!! Saved me lots of time... Thanks

[@sumitbawa1662]

Excellent Tutorial, Thanks

[@antondoit]

Fantastic. You describe it so well.

[@Kalicall]

Really helpful video how we handle ai to learn from our data and give us what we expected

[@zhang20244]

So cool, Thanks for introduce !

[@dinamohamed13600]

It's a great video very helpful, Thanks for sharing.

[@aifresh]

Thank you for this video. 🤝

[@oscargarin]

Great video!

[@Mohitsingh-cq7op]

Dan, thanks for the great video. My question is around data security - my organisation treats it every bit of data as sensitive. So I would like to know more how is my organisation data is being treated. Thanks in advance 🙂

[@benpreston4766]

Great video thanks a lot

[@NadeemAshraf]

Thanks, Dan! Your explanation was very clear. I have a follow-up question: How can we configure AI search to utilize the text data stored in our relational database, rather than relying on individual document files?

[@DanWahlin]

It doesn't support that as of today unfortunately. Hopefully in the future.

[@ryanolson1594]

Do you have any resources for indexing table data rather than documents?

[@raiobg]

hi thanks for the video! you explain very good !, in the end you say its getting better over time, you mean the chatbot, which is learning from the interaction? can that gpt be used as a LLM to answer more in a human way? thanks!

[@conexaomotorizada]

Is it possible to use files in csv or xlsx format to produce the chatbot? What would be the necessary requirements?

[@danielburley4632]

Do you have a guide on how to get the API Management to discover your BYOD?

[@EliteHeadshots]

How much would this roughly cost per month? thinking whether to use it personally in the meantime for work purposes, then once its refined, to introduce it to the workplace?

Example Monthly Cost: Assuming the company has 10 GB of data to store and generates 1,000,000 tokens per month: Azure OpenAI Resource: GPT-3.5 Turbo: 1,000,000 tokens x $0.0024 = $2.40 Ada: 1,000,000 tokens x $0.0008 = $0.80 Data Storage: 10 GB x $0.0184 = $0.184 Azure Cognitive Search: $100 per month Vector Search: 1,000,000 tokens x $0.0008 = $0.80 Generating and Hosting the Web Application: Estimated $50 per month Total Monthly Cost: $2.40 (GPT-3.5 Turbo) + $0.80 (Ada) + $0.184 (Blob Storage) + $100 (Cognitive Search) + $0.80 (Vector Search) + $50 (Application Hosting) = $154.184

[@Qamarkhan007]

Thanks for sharing.

[@Kryabenok]

Great video. Did anyone find out how do I add another document?

[@mikel9292]

Great video. Is there any possibilty of adding more documents via API? It's not very useful if the documents are the same all the time. The goal is that the user can provide their own documents and ask them questions.

[@riochip]

Did you find solution for that ? I wanna know. It would be silly if we train or deploy another model, right ? 😂

[@mikel9292]

@@riochip yeah, you have to populate an index with an indexer with the content of the document. You have to use some kind of a searcher. I'm doing this using azure ai sesrch service. It's a little bit complex to set it up but it works really nice.

[@dereinzigwahreJules]

Is there a way to like not use the azure ai interfaxe at the end of your demo but like create your own name and such and use your own interface design

[@babla3157]

Thank you so much Dan for the video. When I use the chat feature it gives me an error - "The extensions chat completions operation must have at least one extension." Any suggestion what may be happening?

[@naveenkumarjainbtp3901]

Can we train this using API data, instead of Docs?

[@reginaldoassuncao]

I'm new to programming. I did the step by step, but how do I create my own app?

[@RandyKrakowski]

How can you get back to that chatbot with your index/files if you leave that screen and didn't deploy it? Say for instance that I want to tweak it without having to re-index all of my PDFs each time.

[@somjadek]

🎉 Great video, how about document size greater than 16MB?

[@rajgopal87]

Thanks , Nice Video, Few questions , what is the limitation of the data i can upload ? Also if i have to upload documents continuously with an interval of every 10 hours , shall i remove the resources after injestion. ?

[@andonii46]

it is amazing! is there any way to add some azure ad sso authentication to the chat that you show in the end?

[@moritzwilhelm9658]

Hello, nice video! I haven't had any contact with Azure until now. If I want to test the same as you in the video, is it completely free if it is under 1000 massages? And do you know how many pdfs for example could be uploaded? Thanks!

[@martinnoah9716]

So simple. How do I add another document? Doesn’t seem you can add additional documents from the playground UI.

[@semosemo3827]

thank you very much for this video, I have question: what did you use the gpt3 for? I mean I saw you use ada model only, right?

[@DanWahlin]

GPT is still used behind the scenes for the prompts. When you ask a question to a document GPT3 kicks in.

[@pkk1363]

Hi Dan, can I use Azure OpenAI BYOD without uploading any of my data to a cloud. I have sensitive data that needs to remain on my computer. Will Azue OpenAI work just on my laptop or desktop and do its search? Please let me know.

[@Z4rnii]

can you show a video of how to deploy a web app for an enterprise and not in the web?

[@ifr-ed]

How do I change or add some data after deployment? Looks like once deployed there´s no chance to edit the data source anymore, correct?

[@kakanisantosh]

Suggestion: We have so many videos like this Can't you try showing how to get both video and text as answer for questions based on pdf document (which have mixed pics and text like websites)

[@43SunSon]

good video~

[@edwinyulianto3660]

Most of my documents are on Sharepoint. Instead of manually uploading them, is it possible to just link them?

[@andrewherrick879]

I'd love to see how to do all this without the GUI just using command line. Allow me to create all these items, upload data from a blob storage, and generate the API. The Chat UI is weird in that it only allows you to interact with 1 DataSource at a time?

[@AntonioOlander]

I have been watching this upload your own files, but what I would like to accomplish is to be able to use the api's to upload documents from my application uI into he storage blob and reindex or ingest the files so openAI will be able to use the latest information. I have not figured that out yet. The use case is to let my clients be able to upload company documents.

[@DataSpook]

I found there is a vetting process. Is this required for everyone to use OpenAI?

[@subhankargharami4893]

Hi, everything worked fine for me, but when I came next day my uploaded file was missing, do you know anything about it? Thanks.

[@RandyKrakowski]

Have you done any versions/videos where you layer Python in for context and to help the model understand your data better?

[@DanWahlin]

I haven't but that's something I'll keep in mind for the future. Thanks for the suggestion.

[@oscargarin]

I tried to follow your steps but Azure OpenAI Services request to send a form to Microsoft to have access to it. Did you have to do the same?

[@TechJarves]

Yes, you need to fill form first to gain the access.

[@royalh13]

How about searching for something in the web app (after turning off "Limit responses to your data") that's not in you're uploaded files... something like "what color is the sky"... doesn't work, does it...

[@xex-one]

Dan, at thge Cognitive Search Resource your selection creates $200+ monthly cost - is that correct?

Example Monthly Cost (not so small): Assuming the company has 10 GB of data to store and generates 1,000,000 tokens per month: Azure OpenAI Resource: GPT-3.5 Turbo: 1,000,000 tokens x $0.0024 = $2.40 Ada: 1,000,000 tokens x $0.0008 = $0.80 Data Storage: 10 GB x $0.0184 = $0.184 Azure Cognitive Search: $100 per month Vector Search: 1,000,000 tokens x $0.0008 = $0.80 Generating and Hosting the Web Application: Estimated $50 per month Total Monthly Cost: $2.40 (GPT-3.5 Turbo) + $0.80 (Ada) + $0.184 (Blob Storage) + $100 (Cognitive Search) + $0.80 (Vector Search) + $50 (Application Hosting) = $154.184

[@hsrinivas797]

I have built an app based on that, but used a blob container instaed of manual uploading documents. After that I have added a few more documents to the blob container, tried re-indexing, still serach is not showing latest added documents. Do you know why it is not working, any methods to follow? Thanks

[@zedcodinacademibychinvia9481]

Sir what's the max size of the pdf i can add and is they a way to call this through an api lets say i want the chat feature to my website

[@amitnilajkar6272]

Thank you, I have a question, how safe is my data which is related complex engineering drawings. How do i secure it as its on azure tenant.

[@kushsharma4375]

Hi Dan, I appreciate the video, but I have a question. What if I want to modify the theme, add a title to the web app (chatbot), or make some cosmetic changes? Is that possible?

[@jirigrof]

Hello, Can you help me how can I publish my web app to ms teams?

[@daveloh1329]

Hi Dan, how do i enable the GPT able to answer even the information is not in the uploaded document? for example, my document is about World War 2 history, but the GTP can answer Math question based on it general knowledge?

[@rajns8643]

Hi. I am facing the same issue. If you have solved it, please let me know how you achieved it? Thanks!

[@akshaybholee3263]

That is a very good question. Also I wanted to add that like he mentioned in his example of how to do a refund and the openAI responded with what is there already. Now if I ask it how to improve the refund process would it be able to suggest improvements? To pay more than 200USD just to ask what is there in a document does not make sense, instead it should be of use to us as a way like using our data and try to improve our processes or reporting.

[@NakkaAbhishek-u5l]

how can we export thhis rag application?

[@yusufkemaldemir9393]

@Dan Wahlin, Thanks for the video. Does OpenAI back up and able to see what document I upload? Who is able to see/read the document I upload? How to deal with confidential data and not willing to let OpenAI or Azure to store/read it? Thanks

[@MrAndrew535]

I have been engaged in a writing project with Chat GPT4o which stopped three days ago. Whenever I add new text, it disappears, which means my work, of over 700 pages, has come to an end. Do you have any idea how I might retrieve our missing work, on the same thread, so that we can continue?

[@Justin-ff9gm]

I'd like to chat against 12 separate SQL databases. 😁😁

[@DanWahlin]

That may involve a bit more code. :-)

[@BlindrHaur]

So without deploying the chat as web app, I can still use it via Api if I want to build my own Chat UI, correct Dan ?

[@titchfau6181]

I have the same question, did you find out how to create an API endpoint connecting to the chat model ?

[@luisavitoria6862]

Hi! I found this example using api: kzbin.info/www/bejne/n6XFhZtriNagmNksi=_ykWyC2nMxngnEzJ

[@tmccart1]

How about going against a good ole SQL database?

[@AudraCarpenter]

Hey Dan, new to your channel and have to say... thank you so much for simplifying this process for beginners. Question though, when I go to generate the query I'm getting an unsupported model error. Saying that gpt-35-turbo-0613 isn't supported. So I change it to gpt-35-turbo and it never completes the query. Returning a 'requested information is not found in the retrieved data'. But its there. Not that you can trouble shoot for me, just looking for a suggestion as to where the issue could be. Thanks again!

[@AudraCarpenter]

Actually found that that model has a known issue. I was able to add the 16k and it worked. Guess this comment can serve incase anyone else runs into this issue. Still love the content... now we just need more:)

[@DanWahlin]

@@AudraCarpenterSome of the models don’t work with the playground unfortunately (as you discovered). Glad you got it going though!

[@robbert.s6414]

While trying this out, I have found that most of the features you see are only available for the east (US) region. Is it true that Microsoft first releases everything for that region and then for others?

[@DanWahlin]

Hi Robbert. I'm not involved with the rollout to different regions, so I don't have any information about that process unfortunately.