Cortex vs Non-Cortex Baselines

NON-CORTEX BASELINE (WITHOUT CORTEX)CORTEX (WITH GRAPH + SMART RETRIEVAL)

Full-context baseline re-sends complete source documents every query. Cortex avoids this via retrieval + graph context, reducing tokens and cost.

Non-Cortex Configuration

Baseline method

LLM model for token/cost comparison

Source / Memory Scope

Scope type

Talk to specific item

0 knowledge available in current tenant

Cortex Configuration

Top-N chunks

Default N is 5 for now. Tune this to compare quality/cost trade-offs.

Further reading: long-context failure modes and retrieval quality