Developer questions: multi-patient, multi-workflow agent memory for diagnostic centres #438
abhijeetsaherawat13
started this conversation in
General
Replies: 4 comments 1 reply
-
|
Hi Abhijeet, I lead DevRel at Memori Labs. You brought up excellent questions! What's a good email for you and we'll be happy to get you the answers to these questions. Best, |
Beta Was this translation helpful? Give feedback.
0 replies
-
|
Hello Jay, My email ID is - abhi.sahe13@gmail.com - looking forward to your answers Thanks |
Beta Was this translation helpful? Give feedback.
0 replies
-
|
Hey, Could you look into this please? |
Beta Was this translation helpful? Give feedback.
1 reply
-
|
$BODY |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
-
Hey! I'm building AI agents for diagnostic centres — multiple patients, multiple workflows (report gen, follow-up, billing), long-running sessions.
I came across memori's 81.95% on LoCoMo and want to reproduce something close to that in a real product setup before committing to the architecture.
A few quick questions:
Reproducing LoCoMo — is there a script or harness to run the LoCoMo eval against memori? I want to swap in a medical-context dataset (patient conversations, multi-session queries) and see if the numbers hold under that distribution.
LoCoMo vs. LongMemEval — LoCoMo tests multi-session recall across long conversations. Has memori been run on LongMemEval-S as well? For a medical use case, knowing how the system handles temporal reasoning and knowledge updates (e.g. a lab result superseding an older one) matters a lot.
What degrades the score in practice? Multi-user isolation, concurrent writes, mixed structured/unstructured memory — does any of this hurt retrieval quality vs. the benchmark setup?
User/agent scoping for multi-patient — what's the recommended ID schema when the same patient interacts across 3–4 different agent workflows?
Basically trying to go from benchmark → working pilot at one centre as fast as possible. Anyone done something similar?
Beta Was this translation helpful? Give feedback.
All reactions