Retrieval & Grounding

How a product finds source text and keeps an answer tied to it.

Five cards on this track are live. Concept cards are free, including the ones that are not written yet. Names without a card are not links. Practice for this track is the Retrieval & Grounding pack, and that pack is coming soon.

RAG Live Free card
Retrieve relevant text, then ask the model to answer from that text.
Chunking Live Free card
Splitting documents into pieces that retrieval can store and return.
Embeddings for retrieval Live Free card
Using embeddings to find stored chunks that are close to a query.
Hybrid search Live Free card
Combining word matching and embedding search so exact terms and paraphrases can both hit.
Query rewriting Coming soon Free card
Turning a follow-up question into a search query that stands on its own.
Reranking Live Free card
A second pass that reorders retrieved chunks before they go into the prompt.
Long context vs RAG Coming soon Free card
Choosing between putting the source in the window and retrieving pieces of it.
Document parsing Coming soon Free card
Turning a file, such as a PDF or a page, into text and structure you can chunk.
Vector databases Coming soon Free card
A store built to keep embeddings and return the nearest ones.
Citation & grounding check Coming soon Free card
Checking that a claim in the reply is tied to a source you actually retrieved.
Freshness & access control Coming soon Free card
Whether a source is current, and whether this user is allowed to see it.

Retrieval & Grounding pack Pricing