Enterprise Chat App using Azure Cognitive Search and Azure OpenAI: End-to-End Tutorial
ฝัง
- เผยแพร่เมื่อ 15 ก.ย. 2024
- Welcome to my TH-cam tutorial! In this video, I'll show you how to create an Enterprise Search and Chat App using two powerful tools: Azure Cognitive Search and Azure OpenAI.
I'll guide you through the entire process, step by step, so you can build an intelligent search application with a smart chatbot. We'll use Azure Cognitive Search to make the search super fast and accurate, and Azure OpenAI will help us understand language better.
Even if you're new to this, don't worry! I'll explain everything clearly, and you'll be able to follow along easily. By the end of this tutorial, you'll have the skills to create your own cool projects.
Don't forget to subscribe and turn on notifications for more tutorials and updates. Happy coding! 🤖
AI Anytime's GitHub: github.com/AIA...
The Sample App GitHub: github.com/mic...
LLM Playlist: • Large Language Models
Azure OpenAI: learn.microsof...
#azure #ai #python
how come we missed your channel bro! you are a true genius and more importantly thanks for helping out with every tiny detail possible. Great kudos to your efforts man. With that saying, lets learn together and stay updated!!!
Happy to help! Thanks. Please watch other videos that might help in your projects.
Thank you, your video is beneficial for my work! I have watched over 5 building AI chatbot videos, yours is the most comprehensive and easy to understand!!
Kubh Bhalo informative ... Learned many points from your content... Thank You :)
Thank you sir
Was really looking forward to this video
Hope it helped ....
Thanks for the valuable videos. Waiting for llama2 deployment on aws sagemaker....& fine tuning peft llama2/deployment....
Coming soon! Thanks for your comment.
im glad ,finally i found right channel ..kudos to you boss it was great video and looking forward for more contents from you ..Thanks a ton
thank you for this tutorial, love your channel
You are so welcome!
Very Useful content. Great Work
Howdy! great video. Have few questions -- a) Can we add multiple cognitive search indexes to the same Open AI model, because I could see only 1 index selection in drop down. My use case is data is sourced from multiple systems and have different schemas. b) Can there be mix of organization and internet data getting returned by the Open AI model, say data about share price or competitor news from internet search and sales/finance data from cognitive search result?
Have you found any solution on this?
Thanks for such a wonderful video 👍
Glad you liked it
Awesome 🔥 thank you so much for your video. I waited so long form TH-cam community to see enterprise app integration. Finally it’s here thanks once again.😊
Most of the fintech companies not moving AI because of cloud solutions. If you know how to bring in on premises deployment & data security & performance if you speak about this it will be useful everyone or if you know links share to me
Glad it was helpful! Yes will share something similar soon. Pls stay tuned.
Hi quick question, why do we have to write the python code and run thru Langchain to create the Indexer ? if we choose to create the Indexer thru Azure Portal wont that create the documents for us? Im curious to know what is the rational behind this approach? Thank you
Of course you can do that. Even I recommend that you try it out. Chains anyway make it slower.... you can try that approach!
Hi, thanks for. Can I use a database that has thousands of pdf files?
How can we integrate the chatbot into ms teams?
I would also like to know how I could do it
It’s possible to deploy the same bot on teams using copilot studio (PVA), where you need to call an action and use HTTP POST method.
create a chat bot for teams and copy the code from chat playground from view code section on top right. Now in that code you will have to put your azure open ai endpoint and key which i think will already be there. Once done that deploy code anywhere you wnat and then using chat bot framework accept the user query and put it in the request body in the code which you copied from chat playground and start sending request. Fetch the response returned by azure open ai and again send it back to teams using the chatbot
Nice video! Could you tell about how the data exchange between Azure ChatGPT and Cognitive Search does happen, in what format? is it in JSON structures or some other format? Are those temporary data structures are stored in log files, and where, do you know?
Nice Video. I have one question. Where are we providing embedding model? without embedding it can not perform semantic search. Please correct me If I am wrong.
as per my undertanding, Azure Cognitive Search knows the embeddings coordinates because they are generated during the data processing phase by the indexer and then stored in the specified fields in the index. These embeddings are then used in search queries to improve the relevance and accuracy of the search results.
Thanks for nice tutorial. What resources are deployed in azure with what configuration, when you deployed this chat app to a web app in the playground
Thank you for a great tutorial, have you come across the issue when select AI Studio -> Chat Playground -> Assistant setup -> Add your Data -> Advanced Settings -> "Limit the responses to your data content," that you can retrieve a maximum of 20 documents? I have indexed 500 and would like it to search across all documents.
Did u figure anything out for this
Very nice video.. Can you please explain the meaning of the below line inside the data.append() list: "@search.action": "upload" ??
why didnt used embeddiing model? does the cognitive take care that?
Hi, seems to me that he doesn’t need an embedding model as he is searching the database by key word. Cognitive search is pulling out the tickets via keyword search and these tickets are fed to the llm. An embedding model is necessary if you are searching a vector database
How to integrate this with a Power BI dashboard, would you have any video on it?
Im assuming the python is to import the data, but cant azure handle that natively?
Yes it can. If you have data already on Azure for example, Blob storage, SQL, Cosmos, etc.
at 7:50 what would you recommend for a hobby or paid projects?
What pre-processing steps are required for a csv file?
Could you help me on this, i follow the code exactly and configure the key as edm.string, now i hit this issue,
Message: The request is invalid. Details: parameters : Cannot convert the literal '1' to the expected type 'Edm.String'
It's ok is just need add str infront of all the rows, hv solved it. btw thanks for ur sharing!
Hello, great video! I have a question. If I have a small local business, how viable is it to use Azure Cognitive Search in terms of price? Or is it more convenient to use Pinecone, even if it means losing the advantage of having everything integrated in Azure?
I will recommend Azure thingy. You can easily scale it up in the future also. There are trade-offs but you can setup on Azure. You can also plug in or add many other services.
Can you show example of changing logo and redeployment
Hey nice video,
but is it possible to run this all locally, to keep the data private?
Yes surely possible. Please watch my other LLM videos in the LLM playlist.
The index is stored in cognitive search. But where are the 2000 documents stored?
Thank you for the video :)
Hi bro, how can i train our own csv data using llama2 any model
how to create a document upload feature on it?
How to make changes in the UI ?
I think csv file format is not supported , for that we need to use langchain agent .... and ofc thank you for this video very useful, specially that indexing part 🚀🔥
You can work with CSV also. Just need preprocessing... Thanks for liking it, Akshay.
thank you very much for this eye opening video. Is it possible to use other languages in Azure Ai and understand you well?
Thanks for your comment, Jorge! Yes it is possible though machine translation is a complex problem but you can achieve good enough accuracy with it. Give it a try....
Hi , Great video man, this is a really good option for enterprises who are concerned about safety. Can you please share the dataset file as well?
Thanks Tushaar. It's available on Kaggle. I have shown in the video. Pls search the dataset name and you will get it.
Excellent brother.
Thank you! Cheers!
Great video again! Which do you think is more cost-effective: deploying Llama 2 on AWS or utilizing the Azure OpenAI chatbot option?
Very complex topic to discuss here. Would you mind pinging me on WhatsApp. I will be able to answer there. It depends on many variables. Unit economics is a complex topic.
@@AIAnytime I'm also interested in this. May I ping you on WhatsApp to discuss it further?
@@AIAnytime how can I ping you?
Look at about channel section or TH-cam banner. You will find the social channels links.
How to add index of pdf files. There is no fixed structure of pdf files. How to use cognitive search for pdf files??
I also have this same question @AIAnytime ?
Ok. Only watched 33 minutes of the video yet (so I might have to delete this later after watching the entirety)). You are here searching the Azure Cognitive Index, BUT this by default does not have and sementic/cognitive knowledge. So you are here writing a question in a plain english sentence, but the service is "only" searching word by word (broken down to bits of words actually). This is a bit missleading that you in this stage can write questions in a natural speaking way... Ok, I guess the openAI/GPT stuff later in the video makes sense for that, but right here at 33 min its not really right.
Ok, same missunderstanding at 47 min.
Not: You CAN enable semantic search in Azure Cognitive Search service, but you have not done so in this examplvideo.
Thanku bhaiyaa ❤