Thank you so much for your feedback, Mark!

You raise some excellent points. It's true that the market is highly dynamic right now. As you mentioned, many production-ready tech stacks are available, and a lot of tools can easily be replaced with others. Cloud-native solutions, for example, work perfectly well in many cases, and I completely agree with that.

That said, I wanted to share the stack that I personally find easier to use and more customizable compared to cloud solutions. For building small to mid-sized applications, these tools cover all the essential needs.

Regarding VectorDBs, you're absolutely right that solutions like Pinecone, Weaviate, or combining GraphDBs with vectors (e.g., Neo4j) offer powerful capabilities, especially when relationships and context are important. I can definitely see the value of more specialized databases for large-scale production environments where performance and scalability are critical. But for smaller projects, these tools are an overkill. With ChromaDB, you can get your VectorDB working in no time. And when you don't have that many vectors/data, ChromaDB may even be more performant, as it's stored locally on the server.

As for GraphDBs, I haven’t had the chance to work with them yet, so I can't speak from personal experience. They do sound promising, but so far, VectorDBs have met all my requirements.

Once again, thank you for your thoughtful feedback. You’ve raised some excellent points, and I may explore some of the other tools you mentioned like KubeFlow and Lamini in future articles.