LangChain

What: A chain of runnables: retrieve, fill a prompt template, invoke the model, parse JSON. Each hop is a small object you can swap.
Why: A hidden call that “just uses the library” is hard to learn from. The pipe — retriever | prompt | llm | parser — is the lesson. A chain is not a graph; branches live in LangGraph.
What you are seeing: Flow on the left lights as each hop finishes. AWS Architecture on the right is that hop’s box — click the title for the stacked twin. Click a finished hop for passages, the formatted prompt, raw text, or parsed JSON.

Static GitHub Pages demo — sample questions replay recorded runs with paced animations. Vector DB hashes in this tab and does not call localhost. Clone the repo and make dev for live labs.

Flow

AWS Architecture

START
accept the question
END
return parsed JSON

Step output

Run the chain — hops light as each runnable finishes. Click a finished hop to inspect it.