Java + RAG: Create an AI-Powered Financial Advisor using Spring AI 💡

[@billrutherfordmusic]

Great video! 100% step by step showing how to implement some of my current ideas. ❤

[@pejko89]

I'm so excited about this! This is the coolest videos about Java! I've build an accounting software and I was planning to add a help section where users could ask questions and get relevant answers. I have around 30 pdf documents of different laws and regulations. This is the perfect thing at a perfect time ❤

[@JarikKomarik]

Thanks for coherent tutorial, will be extremely thankful for tutorial of stand-alone RAG app with olama model!

[@DanVega]

You're most welcome. What do you mean by stand-alone RAG app?

[@JarikKomarik]

@@DanVega I mean without dependency on OpenAI. With own LLM that generates response based on provided context.

[@renanvaz615]

The best video about RAG I have ever seen

[@MrBoczkas]

That's great video. Going from general picture to code examples is great idea!

[@sergey--24]

Great video, Dan - thanks! It would be very interesting to dive into vector database mechanics as well 👍

[@kevinjia1984]

you are as cool as always, thanks Dan !

[@javadahmadzadeh7129]

Great Video! Thanks. Looking forward to Ollama videos

[@ahmeda.maksoud9637]

Excellent as usual

[@thecote]

This was really great, I appreciate it!

[@DanVega]

Thank you Cote!

[@dineshrajpampati8928]

So the images are not processed right. They also hold some data, how can we include images too?

[@esteban4lv1n02]

Thank you Dan, for touching RAG, if prompting is require may you point out some resources you recommend for the topic, thank you

[@larienasujith7005]

Which is better pdf document parser it is OllamaParser or PDF Document Reader?

[@arnaudpoutieu1331]

Thank you Dan for this one! I have a question regarding how the ChatClient is instantiated using the Context. How it will deal with large dataset from the vectorStore : any performance issue that could be forseen?

[@AnkitLadha-fo8em]

How do we know ho much input token and output token were used in this request so that we can track of the Financials

[@konstantingromov6485]

It's returned in metadata of each call to ChatClient. Also spring-ai team added those data as metrics in newer verrsion (at least in 1.0.0-M3) - so all you need just to add actuator to classpath.

[@Mowgli717]

Wonderful tutorial... but how to use .xlsx file in RAG?

[@NguyenVietHoangHE]

nice video, can u tell me what is your font u are using in your IDE

[@Nick-yd3rc]

Great intro Dan, I always enjoy watching your videos. ❤ As for the model and the retrieval themselves, well, 250 bps priced in? Seriously? Leaving aside the average quality of Tika and the warts of PdfBox, and the fact that it’s by far not enough to just leverage the existing open-source libraries to come close to anything production-grade, that toy sample is fun to showcase that the shape looks right. But I’m still struggling to come up with a justification for pulling in the confines and cruft of Spring into the realm of Python, where you get everything from the papers immediately. I believe, Spring AI has managed to retain the relevance of Spring in the coming couple years. But Spring is unsuitable for the agile and often experimental workflows in data science. An average Spring dev can’t do proper data science, and an average data scientist can’t do proper Java and much less so proper Spring. So I don’t know any other reason why VMware had decided and Broadcom has agreed to keep it up. I wish Spring had focused on improving what it’s good at and reigning in some consistency at last. I know, different people, different projects, but today it’s often just enough to add a Flask or a FastAPI wrapper and be done with it.

[@tranquocthinh7406]

Can you provide guidance on how RAG can work in real-time to read dynamic data? I have a project for an e-commerce website integrated with a chatbot. I want the chatbot to be able to read updated data (such as price changes) in real-time, instead of having to restart everything from scratch. Please, help me

[@indianengineer5802]

Make video on Agentic RAG using Spring AI

[@prashlovessamosa]

Just came from X thanks.

[@ConAim]

Nice, but couldn't posting it to Ollama ...

[@DanVega]

What problems did you have with Ollama? This should work the same just swapping out OpenAI for Ollama and an open source model.

[@glauberandreolli2252]

@@DanVega Hello, I can't post on Ollama llamma3.1 either. I/O error on POST request for "localhost:11434/api/embed": timeout

[@mayikx]

The PG Vector blew my mind.

[@DanVega]

What part blew your mind?

[@ssomasun]

Hello Dan Vega, I'm currently working with Spring AI to retrieve structured data from AI calls. At present, I’m using a method structured as follows: java Copy code private T promptChatClient(Class responseType, Resource loanData, String textData) { return chatClient.prompt() .user(u -> { u.text(loanData); u.param("data", textData); }).call() .entity(responseType); } This method utilizes chatClient.prompt() to send structured data (loanData and textData) and expects a response in the specified responseType. Would you have any suggestions how to get number token used other information that I get when I call chatResponse()? Thank you for your time and expertise!