Connecting Azure SQL DB to Azure OpenAI Service using Cognitive Search | Create Microsoft Teams Bot
HTML-код
- Опубликовано: 16 окт 2024
- In this video, we connect the Azure SQL DB to the Azure OpenAI service using Azure Cognitive Search. We also created a custom application/chatbot using Microsoft Bot Framework that connects to Microsoft Teams Channel. Here, we also used the Semantic Search capabilities of Azure Cognitive Search for better and more accurate results.
Source Code: shop.dewiride....
Chatbot Consultation (Paid): www.calendar.c...
Free Consultation (15 Minutes Only) - www.calendar.c...
Business inquiries: info@dewiride.com
Follow us here:
Company Website - dewiride.com
JD Bots Blog - jd-bots.com
LinkedIn (Trainer) - / jagdishkumawat
LinkedIn (Company) - / dewiride-technologies
Facebook - / dewiride
Twitter - / jdbots
Instagram - / jdbots
Telegram - t.me/JDBots
Thank you for the tutorial; it's been incredibly helpful! I do have one question though: is there a method to include document references or enable citations in chatbot replies
It gets the citation in the response. You can parse JSON and retrieve it and showcase the way you want.
Using Cognitive Search and indexing has costs. Can you do a video on costs. Whether is cheaper to use a vector database rather than Cognitive Search
Sure
Hi! On my case we need to connect some tables from our Data Warehouse to a LLM (Open AI). If there are multiple tables can the model perform "Joins" to query the data based on a prompt? Our Data Warehouse is also hosted on Azure enviroment
Ohk, in that case you would need to generate a sql query in run time using LLM. This is one such example of generating SQL query in runtime, your case requires join and little more advanced - ruclips.net/video/X1cYe9YHHtk/видео.html
Your video about the Azure's OpenAI integration was useful. Is it possible to integrate the above approach in my own application instead of Microsoft Teams ?
Yes, you can. Just call the api from any application.
Your video is very helpful. I want to train my BOT from dynamic data (ongoing chats in various teams channel). How to do that? do I have to load my data from teams channel in to a db or file and link it with cognitive search?
directly send the teams data to azure ai search index. DB need not be needed.
Is there any c# code that will be able to retrieve evaluation metrics data?
This is a very good explanation of configuration and implementation, thank you very much for the video, however not very much to applying the cognitive search on SQL in real-life practical usage. Appreciate if you can apply this implementation in sales and transaction database.
Correct. I am finding a best and accurate way for sales, financial data.
This is so cool, but the cognitive search seems using embedding semantic searching, which does not work when you are heading for some structured query, in this cases tools like AskYourDatabase may be better
in this video you worked with only one table but can we work with multiple tables within a database?
same question
Hello, Thanks for this detailed explanation. Is there a way to integrate Azure Open AI with Postgresql in Azure where in I have a table with a column where I want to search for using vector embeddings effectively.
Thanks in advance.
Currently connector not available for Postgresql. But, we can still build a pipeline to create embeddings and index the data.
When new records got inserted into the source table, Is there a way to automatically index the data in azure cognitive search for the new records ? And same for deletion case as well.
Yes, use indexer and schedule it. And configure deleted columns.
how to implement that in react.js code with our custom data
how to you this when we have multiple relations between different tables and answer need to be retrived from multiple tables in the database
Inbuilt support of Multi Sources is still not available yet. But, as a workaround, you can query multiple indexes and feed it to the Azure OpenAI. It will all depends on your data and relationships.
Thanks for the great video. Subscribed to your channel!
1) What is the best way to train the open AI model based on AzureDevops User Stories and Bugs?
2) Once I add this to Teams, is it possible for others in my organization to use this as well?
3) If added to Teams, can everyone see what others have asked?
4) It is possible to add this to Teams in away that everyone can only see their questions (including history) in the chat window? Just as if you speak to a colleague and no one else gets to see your conversation with them?
1. Export as CSV and Upload to Index
2. Yes, Ask your admin to Upload the Teams App Manifest to Org Catalog
3. No
4. You can maintain each user history through storage and keeping User's Object Id as the primary key.
@@Dewiride Thank you for the quick reply! Do you have a tutorial on how to get the Team App Manifest file from Azure AI dashboard?
Not sure about the Azure AI dashboard, but you can create a manifest from Microsoft Teams Developer Portal application - teams.microsoft.com/l/app/14072831-8a2a-4f76-9294-057bf0b42a68?source=app-details-dialog.
Else, get the manifest from Microsoft Github Teams samples repo and add your app id and modify it based on your needs.
I would recommend second option.
@@Dewiride thank you..can you do a RUclips tutorial?
Creating manifest tutorial is already there on our JD Bots RUclips channel. Check out multicloud series or enterprise bot series.
Do you provide learning/training on real projects for Azure to beginners?
Yes, we do provide.
@@DewirideI'm also interested in some advance projects in azure
Sure. Connect with us on info@dewiride.com
A quick question even before i continue to watch, storing in azure sql db vs storing json,csv,pdf files in blob for the same open ai document search, which will be more efficient and less error pron?
It all depends on the type of data you have in these documents. The structure is documents.
Pdf can have varied structure but other ones have fixed.
All will work, but what will work for your data will only depend until you try out.
@@Dewiride Ok. One more Question, let's say I have few structured data (csv or tables) and few docs with Q/A. am I able to use multiple types of services (sql db , blob) ? and secondly, for table can I still do rbac or control over data to who can see their own data only and who can see everything (list of admins)?
Yes, you can do filtering on the search results retrieved from cognitive search. Just select the fields that you can use to filter and perform RBAC. For example, if I have a field called Email, then while querying the cognitive search, you can pass on the filter as - "Email eq bob@example.com".
@@Dewiride thanks. How about multiple source files, can i use all files stored as json & put them in blob? Or can i use multiple data source such as sql + blob stored json?
@@premsoni0143 If you have multiple data sources, you can create a new index for each one. Else, if you need single index, then, put all files as json in single container.
can you help with connecting using managed identity. can you provide connection string
What service you want to connect using managed identity?
@@Dewiride Azure Database to AI search
Here is the doc for the azure sql - learn.microsoft.com/en-us/azure/app-service/tutorial-connect-msi-sql-database?tabs=windowsclient%2Cefcore%2Cdotnet.
For AI Search - learn.microsoft.com/en-us/azure/search/search-security-enable-roles?tabs=config-svc-portal%2Cdisable-keys-portal
hi, thank you for the video. It is very helpful. I'd like to ask you if you have any sample (or link to view) similar to this one but with more transactional DB style. For example, let's say a sales order SQL table with a couple of columns (order ID, customer ID, total amount, etc.). Do you an idea on how we could use Azure Open AI to query this kind of data? Thanks in advance.
I have already tried this, not a great results. I am refining it to make the model perfect.
is it possible to train to collect some data? like example I want user to provide some data, so if user is typed something all the data we need, so it should ask for full data. and convert it to json.
That is not training, if you want to get all the data. If you want to get the data as the response in json, then do not use Azure openai, directly query the cognitive search.
Cognitive search will respond you all the data based on the user's query
Fantastic
Hey sir...
Agar ek pdf ki information dusri word file mai edit krni ho automatically to wo kar skte hai kya...
Pls help me
Sorry, did not get what you are trying to achieve.
@@Dewiride I want to tell you that I have a 1 pdf file in which data is filled and I want to fill the same data in another file ,in the second file some different data is also there. then after analyzing both the data, that second file Have to fill automatically, like fill in the blanks. can there be some any automation like that?
I short I want to copy data from one file and want that data paste into other file ..
Main Task is that (ai have to analyse data smarty coz both files are not similar , for an example x info is in page no. 1 but in 2nd file ,that x info should be paste in page no. 9 )
@@Dewiride I hope u understand my query
Yes, that can be possible. Based on your first comment, you want to retrieve data from pdf and send it to word file. You can use form recognizer to extract the data from pdf and create variables in your word file to feed the data in word.
Can I connect OData service to Azure OpenAI?
Direct connection will not be there, you might want to send the stream of data to Cognitive search through coding or send the data to SQL db and then connect to Azure OPenAI
Hi. could you point me to the video you mentioned about set up Visual studio for bot dev pls?. Thanks :)
ruclips.net/video/PFDbtYQPIb0/видео.html
ruclips.net/video/Bue40V6BCKg/видео.html