Link​

• Building with Generative AI: Lessons from 5 Projects Part 1: RAG - DEV Community

What to Expect​

I understand people have different opinions about the current state of AI or AI in general. This article isn’t about hyping it up or fear-mongering. It's for developers who are interested in building exciting projects using Generative AI tools. I’ve spent the past few months exploring what we can build using these tools, diving into concepts like * RAG*, embeddings, and agents.

I’d like to share everything I learned while building 4 projects(lessons, code, and links) to hopefully give you a clearer roadmap and the resources you need to use these tools in your next project.

In the first project, we’ll look at the most common use case: RAG. We’ll use it to build a document search system, which will naturally lead us to embeddings, my favorite tool of all. Then we’ll move into agents, which help us build more specific and tailored solutions. Finally, I’ll share thoughts on moving to production: how to build secure, real-world applications and which emerging tools are worth watching.

Link​

• Building with Generative AI: Lessons from 5 Projects Part 2: Embedding - DEV Community

What to Expect​

In the last part of this series (optional but worth reading), we explored basic RAG and noticed how embeddings play a key role.

In this article, I’ll introduce what embeddings are and what kinds of solutions can be built using them.

Specifically, we’ll look at how embeddings can power podcast segment search, recommendations, image search, and classification systems.

What to Expect​

What are agents and how AI companies define them, and how tools give agents access to external environments.

How to build agents with the Google ADK library and how to connect small agents to build comprehensive agentic solutions.

What to Expect​

• What are the weak points in building agents and how can we build guardrails in front of them
• Learn two practices and some patterns for securing agents with Google's ADK library

Problem​

Many top AI services that work with PDFs require users to upload their documents and interact with the document's data through a chat interface. I wanted a solution that allows users to chat with their PDFs while reading them, enabling questions exclusively about the current page or chapter.