RAG (Retrieval Augmented Generation) cookbook

15 recipes · 100 steps · about 81 minutes of reading

Ground RAG answers in your own documents
Every tutorial wires up a different framework, and I just want my chatbot to answer from my own files.
BeginnerRead 6 minDoing: a couple of hours, once you already have working access to a chat model and an embedding model7 steps
Decide whether your RAG project needs a dedicated vector database
Every roundup insists you need a dedicated vector database. Your RAG project is still a folder of a few dozen files.
BeginnerRead 5 minDoing: half an hour to an hour, mostly arithmetic on your own numbers rather than new code6 steps
Split your documents into chunks that help RAG give better answers
My RAG answers keep citing the right document but the wrong paragraph.
Beginner to intermediateRead 5 minDoing: an afternoon6 steps
Make your RAG system say it does not know instead of guessing
When nothing relevant turns up, my RAG system still makes something up instead of saying so.
Beginner to intermediateRead 5 minDoing: an hour or two, once you already have a RAG pipeline that retrieves and answers from your own documents6 steps
Metadata filters that stop your RAG search returning the wrong department's file
Your RAG system keeps citing a plausible-looking chunk from the wrong department, year, or customer.
Beginner to intermediateRead 4 minDoing: an afternoon6 steps
Add citations to every RAG answer
A teammate asked where an answer came from and I had no way to show them.
IntermediateRead 5 minDoing: a couple of hours, once your pipeline already retrieves chunks and can enforce a JSON schema7 steps
RAG's vector search misses exact terms that hybrid search catches
Searching for an exact error code or model number, my RAG system returns five results about something else entirely.
IntermediateRead 5 minDoing: an afternoon5 steps
Add a reranker so RAG stops burying the right chunk
My RAG system retrieves twenty chunks and stuffs all of them into the prompt, and the model gets lost in the noise.
IntermediateRead 5 minDoing: an hour or two, once you already have a batch of retrieved candidates and a way to call a reranker6 steps
Add a router that skips retrieval when your documents won't help
Why does typing 'hi' to your RAG chatbot still trigger a full document search?
IntermediateRead 5 minDoing: an hour or two, once your RAG pipeline already answers questions end to end7 steps
Your RAG's retrieval hit rate has never been measured
I keep tweaking my RAG pipeline and have no idea if retrieval actually got better or worse.
Intermediate to advancedRead 5 minDoing: an afternoon to build the labeled set and measure once, a few minutes each time after that8 steps
Rewrite a vague question into a sharper RAG search query
Users ask short, messy questions, and my RAG system searches with exactly that messy wording.
Intermediate to advancedRead 5 minDoing: an afternoon, once you have a handful of your own real vague questions to test against6 steps
Your RAG system cites a document you already replaced
Support changed the return policy last week and my RAG bot is still quoting the old one.
Intermediate to advancedRead 5 minDoing: an afternoon, once your ingestion script can already add and delete records by id7 steps
Split a multi-hop question into sub-questions RAG can answer
One RAG answer needs facts scattered across two different documents, and a single search only turns up one of them.
Intermediate to advancedRead 6 minDoing: a few hours8 steps
Defend your RAG documents against indirect prompt injection
Someone could hide an instruction inside a PDF my RAG system indexes, and I have no idea what it would do with that.
AdvancedRead 7 minDoing: an afternoon, with edit access to your system prompt and the code that assembles retrieved chunks7 steps
A parent document retriever for chunks too small to hold their own context
Your chunks are small enough to embed well. The model still misses half a fact because the number landed in the next chunk.
AdvancedRead 8 minDoing: an afternoon, with edit access to your indexing code and wherever the retriever's lookup lives8 steps

Cookbook badge

Finish all 15 recipes to earn this.

RAG (Retrieval Augmented Generation) cookbook

Ground RAG answers in your own documents

Decide whether your RAG project needs a dedicated vector database

Split your documents into chunks that help RAG give better answers

Make your RAG system say it does not know instead of guessing

Metadata filters that stop your RAG search returning the wrong department's file

Add citations to every RAG answer

RAG's vector search misses exact terms that hybrid search catches

Add a reranker so RAG stops burying the right chunk

Add a router that skips retrieval when your documents won't help

Your RAG's retrieval hit rate has never been measured

Rewrite a vague question into a sharper RAG search query

Your RAG system cites a document you already replaced

Split a multi-hop question into sub-questions RAG can answer

Defend your RAG documents against indirect prompt injection

A parent document retriever for chunks too small to hold their own context

In these roadmaps

RAG (Retrieval Augmented Generation) cookbook

Ground RAG answers in your own documents

Decide whether your RAG project needs a dedicated vector database

Split your documents into chunks that help RAG give better answers

Make your RAG system say it does not know instead of guessing

Metadata filters that stop your RAG search returning the wrong department's file

Add citations to every RAG answer

RAG's vector search misses exact terms that hybrid search catches

Add a reranker so RAG stops burying the right chunk

Add a router that skips retrieval when your documents won't help

Your RAG's retrieval hit rate has never been measured

Rewrite a vague question into a sharper RAG search query

Your RAG system cites a document you already replaced

Split a multi-hop question into sub-questions RAG can answer

Defend your RAG documents against indirect prompt injection

A parent document retriever for chunks too small to hold their own context

In these roadmaps