Slides: Stop Using RAG as Memory — Daniel Chalef, Zep
Source Video
Stop Using RAG as Memory — Daniel Chalef, Zep
Relationship To World's Fair 2026
These slides are extracted from a public AI Engineer YouTube video connected to World's Fair 2026. Speaker-matched clips are supporting context unless later confirmed as exact session recordings; official livestream recordings are day-level/event-level source material.
Related Scheduled Sessions
- No individual scheduled session mapping has been assigned yet; treat this as an event livestream deck.
Extracted Slides
OCR text:
aWws
ee)
@®Graphite W Windsurf 4) MongobB
Mdaily £3augmentcode WorkOS
OCR text:
cme
One Size Fits None
Why Memory Must Reflect Your Business Domain
p- 7
a a a Microsoft @yr{?
OCR text:
entityfields,
EntityType,
}from"egetzep/zep-cloud/wrapper/ontology"
export constfinancialGoalSchema:EntityType={
description:"A specific financial objective the user wants to achieve."
fields:{
goal_type: entityFields.text(
"Type of financial goal (emergency_fund,house_down_payment,retirement,vacation, debt)
AIE
target_amount:entityFields.float（"Target dollar amount for the goal"),
priority:entil
lds.text（"Goalpriority level （high,nedium,low)"),
export const expenseCategorySchena:EntityType ={
fields:(
category_name: entityFields.text(
Name of expense category (housing,food, transportation，entertainment,utilities,shop
monthly_spend:entityFields.float(
"Average monthly spending amount in this category
optimization_potential:entityFields.text(
"Potential for reducing spending in this category （high,medium,low,none)
export const debtAccountSchema:EntityType={
description:"A debt obligation that impacts the user's financial health.
fields:{
debt type:entityFields.text(
OCR text:
bear
Daniel Chaief
Zep Al
danielégetzep.coa
ABSTRACT
We imruduce Zep. a novel memory layer service for Al agents that outperforms the current sate:
of-the-art system, MemGPT. sn the Deep Memory Retrieval (DMR) benchmark. Addivonally, Zep
2% excels in more comprehcauve and challenging cvaluatwoas than DMR that better reflect real- works
coterpoe use cases, While curtng retneval-augmentied gonctanon (RAG) frameworks for large
language model (11M -hased agents arc limited to statec document retneval, cntcrpnse applicapoas
demand dynamic Loowledge integraioa from diverse wurces including ongoung cuavenations and
buunews data Zep addresses this fundamental lmitabon through its core component Graplnty—-a
bs, lemporally-aware knowledge graph engine that dynanucally syntheuses both unstructured conver:
Senora) data and structured buvness dala while masntaining hivioncal relavonships. In the OMR
benchmark. = hoch the MemGPT team eusdinhed as ther primary evaluation meine. Zep demon-
strates wiperwe performance (94.4% +0 93.4%). Beyond DMR. Zep's capabalues are farther vali.
dated through the more challenging LongMcmi:val benchmark, which better reflects enterpnse use
sasen through compiles temporal reasonang Lav Ja Gus evalusbon. Zep actueves substantial rewults
lex with accuracy improvements of up to 15.5% while umultancoud) reducing response latency 5)
90% compared to baseline implementations. These resulls are parbcularly pronounced in caterpnse-
snbcal Lasks such at cfoss-sesucn anformation sy athesss and long-term contest maintenance, demon-
strating Zep’s effectiveness for deployment sn real-workd applications
te iy a ty
D ‘i a
; https://ze.. “te .. er
f 2
A ‘ fy F
= 5 => Sd
lanl
a ea
-
Slide-Derived Subjects To Review
Subject extraction uses video title, related session titles/descriptions, transcript context, and OCR text when available. OCR is best-effort and should be reviewed against the embedded slide images.