How we cut a healthcare partner's LLM spend to $0 per chart
TOPIC AREA

RAG

Retrieval-augmented generation in practice: chunking, overlap, retrieval and reranking, and how to tell whether any of it helped.

Retrieval-augmented generation (RAG) answers a question by first finding the relevant passages in your documents and then asking a language model to answer from them. Most of the answer's quality is decided before the model writes a word: did the right passage make it into the context?

These notes cover the parts you control, such as how documents are split, how much passages overlap, how many are retrieved and whether a reranker re-scores them, and how to tell whether any change helped. Every topic links to a demo that runs in your browser, so you can try the idea on your own text.

Each topic carries a stance: use it, try it, watch it, or avoid it. We do not publish accuracy numbers we have not measured, and we say so when a claim comes from a demo rather than a benchmark.

Try

Promising and worth a trial, but measure it on your own data first.

Try

RAG reranking: when a second pass pays off

Reranking is worth trying when your retrieval is close but not reliable. It is not free accuracy, and the only way to know is to measure it on your own questions.

READY TO BUILD WHAT'S NEXT?

Not sure where to start? Start with the problem.

Working on something in this area? Tell us the problem and we will tell you what we would do.

View All
Contact us

Have a problem worth getting right?

Send us the paragraph you'd normally write to a colleague. We'll get back to you within 24 hours.

hello@tathastha.com