Thanks sooo much for this video. You covered everything perfectly from beginning to end. Most other tutorials will skip a step or two and hard to follow but this was perfect.
Hi, this is a great video. I hope you can answer a few of my questions. 1. are you using Azure Form Recognizer anywhere? 2. what happens if you add a new document to your storage? Does it get automatically indexed? If not, how can this be done automatically? 3. can this app be run without deployment, like locally in the browser? 4. can we change the frontend UI?if so how? 5. how would you describe the architecture for this? Thanks a lot. If possible, could you please answer these questions? Again, really great video. Thanks.
1. No, Form recognizer is not used here. For large documents, yes, it can be used if working on Azure Stack. 2. If indexer is scheduled to run then, depending on your schedule, it will update the index. 3. Yes, source code already mentioned in the description 4. Yes, source code available 5. Simple and no coding but for better accuracy and more enhancements, it can be improved with the help of langchain.
@@Dewiridehi, thanks for quick reply, what classifies as large document and if possible can i avoid using form recognizer completely even tho i might have large number of documents and can langchain be implemented without code, thanks
As of now, using Azure OPenAI studio to index documents directly does not chunk all docs. So, we need help from langchain or semantic kernel to chunk it down and then upload to index. Reading document -> can use form recognizer or other open source libraries. There is one tool in the market to work with langchain without code - FlowiseAI. Did not check the capabilities of this tool, you can check if it works for you without code.
Great video! At one point you mentioned updating documents that the chat is based on, but didn't expand on it. Is there a way to update, add, or remove documents and trigger a retraining?
your video is so awesome. Thank you very much. Some queries and some possible next video suggestions 1. Instead of Azure Web App, is it possible to use Power Apps? 2. Can SharePoint Online search be used to Azure OpenAI search be used? Thanks again for your video!
Very informative video sir. I have a doubt regarding the citation part like how it is generated and can we customize the citations like adding more details like some url of that doc something like that. Thanks
Hello, thank you for the perfect explanation. How can I directly contact you? I have a few more questions about this topic that will help me with my thesis
Thanks a lot for this great video! A question: I seletected the option "Limit responses to your data content" but when I made the deploy in the Azure AI Studio Portal to a new web application, the web application is searching not limit the responses to our data content. Do you know why? When calling through the Open API what is the parameter that sets the option "Limit responses to your data content"?
If I refresh the browser page in the Chat playground, my data source disappears and I have to add all of my data again and allow it to ingest & index all over again. Is there a way around that?
This was very informative, Thanks for that. Additionally do u know how to integrate this deployment to an Avatar chatbot, Where avatar will talk instead of string output
For insights on the financial data, I would recommend to go with Power BI Q&A feature. Else, you can still try this by writing a script that uses cognitive search api/library to index your excel data.
Hi Thanks Great Demo to start the OpenaI with our own data .My query is about API/SDK , Is any way above mentioned steps can we do through API/SDK ,to integrate with our custom application?
@@Dewiride Hi I am not sure for API level this chucking & indexing we have to do ,I am not able to find proper API's to do so .Can you please help on this
Yes, scripts are available that use api and client libraries. Go to the GitHub repo mentioned in video description and click on scripts. For uploading through website, we have a video on it using qna embedding
Hi thanks for the great video. I am getting this error after filling the 'Add Data' form. I have selected blob storage and added a simple word document into that. Is it possible to help with this? Thanks Error: "We couldn't connect your data Skillset related operations are not enabled in this region."
bro i have a doubt, let's say i want to train my model using org data. does azureopenai provide data security?if we want to pretain a model using custom data, we have concerns about data privacy right. does azureopenai provides data security
great video very helpful . I am following the same process but i cant able to deploy my webapp the location place is immutable there please help me out with this
Title seems to be wrong. It's is Using your Custom data for Chat on Azure Open AI studio. It confused me with training the model. Is there anything like training your data? I don't think it is even a valid statement.
i used unstructured dataset but it is not giving me good response. saying Based on the retrieved document 1 , I don't find any information related to converting data to structured data. Is there anything else I can help you with?
It was working fine. However, today I have started getting the issue while creating the data source. I took the blob search. After giving all the values on the add data source page. The error is as follows: We couldn't connect your data Indexing failed. Container "name" doesn't exist. Please adjust your data source definition in order to proceed.
Yes, I checked. Somehow it is not pointing to the right container I guess. With File upload it is working fine. But I want to use it with blob search as well. I will check for the container thing once again. Also I tried Azure APIs for Java. However it is only considering the conversation history while running in intelij. It is not able to interact with the data that I provided in the document(data source). I am using ChatCompletions class.
Hi, thanks for this great video! Is it possible to request this "custom chat gpt" with an API in order to make multiple requests at once? And is it possible to import a large CSV file with 500 000 rows for instance ?
Yes, API can be called for this custom chatgpt. Regarding CSV, currently, the format is not supported yet - learn.microsoft.com/en-gb/azure/cognitive-services/openai/concepts/use-your-data#data-formats-and-file-types. You can support it by doing some custom implementation.
@@Dewiride how does the model search through them if they are not vectorized? does that mean i can just upload a bunch of documents to the storage and not have to use an embeddings model first?
They are vectorized, not in blob storage but in cognitive search. Correct, for this implementation, just upload documents and disconnect and connect the data source, you are good to go.
@@Dewiride thanks! would blob storage be better or worse than cognitive search if i had like hundreds of pdfs that i was trying to search for? trying to figure out which solution perfomrs better and then decide which one to implement based on the cost.
Blob storage is only for storing your documents. The options showed in video, all uses cognitive search. If selecting directly the cognitive search, you must create an index manually and point the index to a data source. If selecting blob storage, cognitive search will automatically create the index and connect to your blob storage as the data source.
Thanks. this is very useful demo. Do you know how can we integrate this web app and interact via Teams? also, for inserting our own data as table in form of .csv?
This is great! One question. Is it possible to integrate this into Microsoft Teams? So instead of using the web app to chat with the bot, users will just use Teams to chat. How can this be done if it is possible?
Hi, Thanks for this video. I am able to add my own data to OpenAI through Azure Blob storage worked properly. But that persists only for that session, when I login again the data is missing and I tried to re-upload the same data but the model isn't working anymore.
@@Dewiride I found my problem..one file size was too big. However, I am noticing for example that if I upload a PDF that has an address in it, and I question with the chat to print the address in the file, it says that it can't find the address in my file when I know it is in there. what am i doing wrong? if i type the request for the address again though, sometimes it will return it..it is not consistent
Try changing the prompt. Sometimes, it does that even though the content which is there, it says did not find. When trying with different prompts, it should work.
@@Dewiride I notice that but I wish the prompts can be standard, especially if I intend to sell this service to a business..also, why when turning off the check box for just using the data, if i ask a general question not in my document, it still cites the document?
since it is doing keyword search with help of cognitive search, it will try its best to answer from the docs. May be you wanna try semantic search. Create a config for semantic search and select that while connecting to your data. Semantic search will have better results.
Thank you for your wonderful video. This was very helpful. I replicated every step and I was able to successfully generate responses. But I did end up with few problems. to start with, I added 10 different Word documents to the container, but engine was only reading the 3 files, Not beyond that and when I asked engine how many files that could retrieve from the blog? It was responding 3 repeatedly. I don't know why any hint any suggestion ?I really appreciate your response
@@Dewiride I uploaded 6 single page pdfs in storage account, now when i trying to convert those pdfs into JSON format using a prompt it is not converting each pdf. Taking only 4 pdfs at once. My goal is to convert every pdf into JSON. I used chunks 1024 and gpt-4o model
When i try to add it to my custom website using API call, it pops up an CORS error. Could you please create a video showing you adding a small pop up bot on your website calling the API and having conversation with it.
Thanks for replying. Can we also get a video on how can we make it more responsive because sometimes it gets stuck on a query. Is there a way we can scale up this bot and make it handle multiple users at the same time without delaying or crashing. How to scale this up for almost enterprise level?
How can we consume this api after adding data in datasource its work on playground screen when we consume this api our application is its give me random answer
@@Dewiride friend, I'm totally new to git hub. I understand that when I click on deploy app it uses this git repository. But how would I make Azure use my edited repository with the images I want in the layout? Thanks
Exceeded quota is given if you try to create resources more than the available quota. For example - Azure openai resource currently can only be created once per subscription. If you already have 1 azure openai resource and try to create another one. You will get quota error. Also, you can only have 1 free cognitive search resource per subscription. If tried creating more, it will ask you for the paid resource and give you the error. Hope you don't end up in this scenarios.
I think the majority of people here that say perfectly explained have not tried implementing it
Still waiting for acceptance to use OpenAI service.
i wish it was more clearly explained step by step rather than jumping from one tab to another. Great work, thanks
Thanks sooo much for this video. You covered everything perfectly from beginning to end. Most other tutorials will skip a step or two and hard to follow but this was perfect.
Glad it was helpful!
You need to tell people before they waste their time following this that you need an Open AI account approved by Microsoft
If you don’t know this, you shouldn’t be watching this video. It’s too advanced. Try Azure fundamentals , AZ-900
Very simple and best use of 20 minutes. Thanks a lot!
Thank you.
Hi, this is a great video. I hope you can answer a few of my questions.
1. are you using Azure Form Recognizer anywhere?
2. what happens if you add a new document to your storage? Does it get automatically indexed? If not, how can this be done automatically?
3. can this app be run without deployment, like locally in the browser?
4. can we change the frontend UI?if so how?
5. how would you describe the architecture for this?
Thanks a lot. If possible, could you please answer these questions? Again, really great video. Thanks.
1. No, Form recognizer is not used here. For large documents, yes, it can be used if working on Azure Stack.
2. If indexer is scheduled to run then, depending on your schedule, it will update the index.
3. Yes, source code already mentioned in the description
4. Yes, source code available
5. Simple and no coding but for better accuracy and more enhancements, it can be improved with the help of langchain.
@@Dewiridehi, thanks for quick reply, what classifies as large document and if possible can i avoid using form recognizer completely even tho i might have large number of documents and can langchain be implemented without code, thanks
As of now, using Azure OPenAI studio to index documents directly does not chunk all docs. So, we need help from langchain or semantic kernel to chunk it down and then upload to index.
Reading document -> can use form recognizer or other open source libraries.
There is one tool in the market to work with langchain without code - FlowiseAI. Did not check the capabilities of this tool, you can check if it works for you without code.
@@Dewiride Again Thanks for prompt reply, i'll look it FlowiseAI
Great video! Can you make a video on how to call the API from our own custom application?
Sure thing!
@@Dewiride from where can i get the api from azure portal i want to paste that in my custom app?
I will create a detailed video in this weekend
Here's the video - ua-cam.com/video/itbTh5Sqmbw/v-deo.html
Thats great! One question,What azure subscription did you use to implement this and can i do it on a sandbox environment??
I have used my free credits as part of my MSDN subscription.
Great video! At one point you mentioned updating documents that the chat is based on, but didn't expand on it. Is there a way to update, add, or remove documents and trigger a retraining?
Yes, update the documents. Disconnect source and connect again.
Thanks, a great video . Covers the basics to add your data and build an application out of it.
Glad it was helpful!
Thank you for your great work. This helps me a lot to understand the concept of the preview. Will try it today on my own. Have a great day!
Glad it was helpful!
your video is so awesome. Thank you very much. Some queries and some possible next video suggestions
1. Instead of Azure Web App, is it possible to use Power Apps?
2. Can SharePoint Online search be used to Azure OpenAI search be used?
Thanks again for your video!
Thanks for the idea!
Power Apps Demo: ua-cam.com/video/tGeiKIdRO7Y/v-deo.html
Very informative video sir. I have a doubt regarding the citation part like how it is generated and can we customize the citations like adding more details like some url of that doc something like that. Thanks
You can add fields as part of the request body.
Thank you for the video! If I upload more data to the blob will it reindex automatically?
Disconnect source and connect again. It will reindex.
Hello, thank you for the perfect explanation. How can I directly contact you? I have a few more questions about this topic that will help me with my thesis
Thanks a lot for this great video! A question: I seletected the option "Limit responses to your data content" but when I made the deploy in the Azure AI Studio Portal to a new web application, the web application is searching not limit the responses to our data content. Do you know why? When calling through the Open API what is the parameter that sets the option "Limit responses to your data content"?
Property name : inScope
If I refresh the browser page in the Chat playground, my data source disappears and I have to add all of my data again and allow it to ingest & index all over again. Is there a way around that?
No, your index will still be there. Next time, choose azure search as data source and select your index.
Brilliant. Only video worth watching on this topic.
Thanks.
Great Video, Thanks for informative session. i have a CSV data of my enterprise, how can i use it as an input file.
Currently, this implementation do not support .csv files. To support CSV, you would need your custom cognitive search index to point to your csv.
Thank you for such a wonderful video. Very well-explained!
Glad it was helpful!
very well explained with live demo - many aspects covered 👍
Glad it was helpful!
This was very informative, Thanks for that. Additionally do u know how to integrate this deployment to an Avatar chatbot, Where avatar will talk instead of string output
Yes, you can integrate it with a digital human. You can use any 3rd party video provider such as Synthesis or DiD
How to upload an Excel and ask for insights? E.g., which product is selling the most?
For insights on the financial data, I would recommend to go with Power BI Q&A feature.
Else, you can still try this by writing a script that uses cognitive search api/library to index your excel data.
Hi Thanks Great Demo to start the OpenaI with our own data .My query is about API/SDK ,
Is any way above mentioned steps can we do through API/SDK ,to integrate with our custom application?
Yes, both SDK and API are available to perform these steps from your application.
Refer to the official doc of cognitive search.
@@Dewiride Hi I am not sure for API level this chucking & indexing we have to do ,I am not able to find proper API's to do so .Can you please help on this
How to send pdf to train model through api and can we train the model using website link
Yes, scripts are available that use api and client libraries. Go to the GitHub repo mentioned in video description and click on scripts.
For uploading through website, we have a video on it using qna embedding
Thank you so much. This is very well explained.
Glad it was helpful!
Hi thanks for the great video. I am getting this error after filling the 'Add Data' form. I have selected blob storage and added a simple word document into that. Is it possible to help with this? Thanks
Error:
"We couldn't connect your data
Skillset related operations are not enabled in this region."
Try changing the region, certain regions don't support few skillset.
Hi, can you please guide on limit your responses , I tried the InScope property ite not working for me
If that does not work, try the system prompt.
May be, the feature is still in preview, we cannot expect that it works completely.
bro i have a doubt, let's say i want to train my model using org data. does azureopenai provide data security?if we want to pretain a model using custom data, we have concerns about data privacy right. does azureopenai provides data security
Yes, Azure OpenAI provides enterprise level security.
bro would you mind sharing resources or posts where i can read about data security for custom data, Thanks@@Dewiride
learn.microsoft.com/en-us/azure/ai-services/openai/faq
thanks a lot bro
@@Dewiride
Can we have the data stored somewhere else instead azure blob storage? Like for example aws cloud etc.
Yes, you can. Use cognitive search sdk or api to index it.
How do I do that? 😅@@Dewiride
I will create a tutorial on it. For now, refer this - learn.microsoft.com/en-us/azure/search/tutorial-optimize-indexing-push-api
Hello, I wanted to integrate my application to consume trained model service(or webapp), how we can achieve that? please guide, thanks
You can use any of the options we have in our videos and create a custom web app to call the chat apis. Calling APIs is also discussed in our videos.
great video very helpful . I am following the same process but i cant able to deploy my webapp the location place is immutable there please help me out with this
Please schedule a meeting. Calendar link already in the description.
Title seems to be wrong.
It's is Using your Custom data for Chat on Azure Open AI studio.
It confused me with training the model.
Is there anything like training your data? I don't think it is even a valid statement.
Custom Data is your data itself. There are two types of data in ChatGPT - Pretrained and the one which is not pretrained will be used as custom data.
If i upload multiple files in blob does it fetch the details
No, it has limitations of 12000 characters. For indexing all, use custom indexing using APIs or client libraries.
i used unstructured dataset but it is not giving me good response. saying Based on the retrieved document 1 , I don't find any information related to converting data to structured data. Is there anything else I can help you with?
Yeah, sometimes, even though context is there, it still replies with that. Try changing prompt
@@Dewiride okay, i will change it. thanks
Many thanks, this is useful for me. many thanks again
Glad it was helpful!
It was working fine. However, today I have started getting the issue while creating the data source. I took the blob search. After giving all the values on the add data source page. The error is as follows:
We couldn't connect your data
Indexing failed. Container "name" doesn't exist. Please adjust your data source definition in order to proceed.
The error is clear. Did you manually check your container if present?
Yes, I checked. Somehow it is not pointing to the right container I guess. With File upload it is working fine. But I want to use it with blob search as well. I will check for the container thing once again. Also I tried Azure APIs for Java. However it is only considering the conversation history while running in intelij. It is not able to interact with the data that I provided in the document(data source). I am using ChatCompletions class.
Thank you for tutorial video, very nicely explained.
Glad it was helpful!
Thanks for the video, but I have a doubt, is it possible to train structured data, JSON at least?
Yes, cognitive search has the ability to index JSON data as well. The official doc of cognitive search uses JSON data in their demos.
@@Dewiride but when I contacted azure forum they said it is not possible
Here's the official doc - learn.microsoft.com/en-us/azure/search/search-semi-structured-data
@@Dewiride thank you, let me look into it…
Hi, thanks for this great video! Is it possible to request this "custom chat gpt" with an API in order to make multiple requests at once? And is it possible to import a large CSV file with 500 000 rows for instance ?
Yes, API can be called for this custom chatgpt. Regarding CSV, currently, the format is not supported yet - learn.microsoft.com/en-gb/azure/cognitive-services/openai/concepts/use-your-data#data-formats-and-file-types.
You can support it by doing some custom implementation.
Calling the API - ua-cam.com/video/itbTh5Sqmbw/v-deo.html
so are the pdfs being vectorized in the blob data storage?
No, they are just stored there.
@@Dewiride how does the model search through them if they are not vectorized? does that mean i can just upload a bunch of documents to the storage and not have to use an embeddings model first?
They are vectorized, not in blob storage but in cognitive search. Correct, for this implementation, just upload documents and disconnect and connect the data source, you are good to go.
@@Dewiride thanks! would blob storage be better or worse than cognitive search if i had like hundreds of pdfs that i was trying to search for? trying to figure out which solution perfomrs better and then decide which one to implement based on the cost.
Blob storage is only for storing your documents. The options showed in video, all uses cognitive search.
If selecting directly the cognitive search, you must create an index manually and point the index to a data source.
If selecting blob storage, cognitive search will automatically create the index and connect to your blob storage as the data source.
Thanks. this is very useful demo. Do you know how can we integrate this web app and interact via Teams? also, for inserting our own data as table in form of .csv?
This video shows chatbot integrated with teams - ua-cam.com/video/mtXxtSZsIpI/v-deo.html
@@Dewiride thanks a lot for doing that
This is great! One question. Is it possible to integrate this into Microsoft Teams? So instead of using the web app to chat with the bot, users will just use Teams to chat. How can this be done if it is possible?
By just calling an API in Microsoft bot framework contacted to teams
@@Dewiride Thanks. What do you think about Power Virtual Agent? Is it better to use that to connect everything together?
Since the bot is not complex, yes, power virtual agents can be used too.
Here's the video that connects to MS Teams - ua-cam.com/video/mtXxtSZsIpI/v-deo.html
Thank you!! I will take a look.@@Dewiride
great training! would be fine to get another one where we make a call from custom app
Working on it
Here's the video for this - ua-cam.com/video/itbTh5Sqmbw/v-deo.html
@@Dewiride thanks
Hi, Thanks for this video. I am able to add my own data to OpenAI through Azure Blob storage worked properly. But that persists only for that session, when I login again the data is missing and I tried to re-upload the same data but the model isn't working anymore.
Next time, you need not reupload, just connect your Cognitive Search index
@@Dewiride it works. thanks again.
I created this according to your video and loaded 2 pdf docs....however, it only indexes 1 doc and I can never get any info from the second doc.
Please test directly the index through search explorer in Azure portal in cognitive search resource.
@@Dewiride I found my problem..one file size was too big. However, I am noticing for example that if I upload a PDF that has an address in it, and I question with the chat to print the address in the file, it says that it can't find the address in my file when I know it is in there. what am i doing wrong? if i type the request for the address again though, sometimes it will return it..it is not consistent
Try changing the prompt. Sometimes, it does that even though the content which is there, it says did not find. When trying with different prompts, it should work.
@@Dewiride I notice that but I wish the prompts can be standard, especially if I intend to sell this service to a business..also, why when turning off the check box for just using the data, if i ask a general question not in my document, it still cites the document?
since it is doing keyword search with help of cognitive search, it will try its best to answer from the docs. May be you wanna try semantic search. Create a config for semantic search and select that while connecting to your data. Semantic search will have better results.
Excel files are not supported in blob storage ???
They can be uploaded but for indexing, you might need to manually index using client libraries.
@@Dewiride how can we do that would you pls explain or any video you suggest
You can watch my latest video. It is just the start/intro. I am planning to do excel one as well soon.
this is awesome content thank yo so much
Glad it was helpful.
Thank you for your wonderful video. This was very helpful. I replicated every step and I was able to successfully generate responses. But I did end up with few problems. to start with, I added 10 different Word documents to the container, but engine was only reading the 3 files, Not beyond that and when I asked engine how many files that could retrieve from the blog? It was responding 3 repeatedly. I don't know why
any hint any suggestion ?I really appreciate your response
What is the pricing tier used for cognitive search?
@@Dewiride Pricing Tier is Basic
check if some docs do not have large size. Even, similar problem was found by another commentor. He figured out, it was due to large size.
I am not able to Add data, it is saying '
Ingestion in progress
Preprocessing is in progress (0 of 2 files)
Indexing has not started' Pls help
If it does not show error, that means, it is in progress.
@@Dewiride I uploaded 6 single page pdfs in storage account, now when i trying to convert those pdfs into JSON format using a prompt it is not converting each pdf. Taking only 4 pdfs at once. My goal is to convert every pdf into JSON. I used chunks 1024 and gpt-4o model
how are you converting to JSON? Use custom coding using Azure Search and Form Recognizer APIs.
@@Dewiride I did try it but not able to understand how to do it. Now form recognizer is document intelligence right?
correct, document intelligence. Responded on LinkedIn
When i try to add it to my custom website using API call, it pops up an CORS error. Could you please create a video showing you adding a small pop up bot on your website calling the API and having conversation with it.
Ok, sure
Thanks for replying. Can we also get a video on how can we make it more responsive because sometimes it gets stuck on a query. Is there a way we can scale up this bot and make it handle multiple users at the same time without delaying or crashing. How to scale this up for almost enterprise level?
Used Bot Framework SDK to call custom API - ua-cam.com/video/itbTh5Sqmbw/v-deo.html. Next video, can connect this bot framework to web chat.
How can we consume this api after adding data in datasource its work on playground screen when we consume this api our application is its give me random answer
You can get the API details by clicking on code and inspect tool.
Here's the detailed video for this - ua-cam.com/video/itbTh5Sqmbw/v-deo.html
@@Dewiride thanks 😊
Great video to understand :)
Glad it was helpful!
Keep it up
Thank you
Can we change the ui of the site ?
Yes, you can. Let me send you the link of the source code. I got from the deployed zip
Yes, you can. Let me send you the link of the source code. I got from the deployed zip
@@Dewiride okay please send me .
I also found the original repository - github.com/microsoft/sample-app-aoai-chatGPT
@@Dewiride friend, I'm totally new to git hub. I understand that when I click on deploy app it uses this git repository. But how would I make Azure use my edited repository with the images I want in the layout? Thanks
I don't know why it gives me a 429 error
Exceeded quota is given if you try to create resources more than the available quota.
For example - Azure openai resource currently can only be created once per subscription. If you already have 1 azure openai resource and try to create another one. You will get quota error.
Also, you can only have 1 free cognitive search resource per subscription. If tried creating more, it will ask you for the paid resource and give you the error.
Hope you don't end up in this scenarios.
@@Dewiride Why do I get a 424 error?
What a confusing process.