The Read Path

Everything between a query arriving and relevant memories sitting in the model's context: scoping to an owner, ranking candidates, and packing the survivors into a token budget.

This is the whole of a RAG pipeline, and roughly a fifth of a memory layer.

Why it matters in a memory layer

The read path is where memory looks most like retrieval, which is exactly why it is the part people over-invest in. It is genuinely important -- but a perfect read path over a corrupted store returns corruption faster.

Connections

Taught in: Anatomy of a Memory Layer · Memory Is Not RAG

Used in: Embedding Recall · Retrieval Is Not Enough · Should I Even Look?

Do not confuse with: The Write Path