Find the first domino
A side effect, mistaken for a new illness, treated with another medicine.
FirstDomino finds that chain across visits and asks for the one missing fact.
01Why a graph
Every answer is one graph query.
- Q1
1walk
Medicine → class → pattern → symptom → second medicine. One traversal.
- Q2
1question
The missing fact most findings wait on, and who can answer it.
- Q3
0re-checks
Change an unrelated medicine and nothing reopens.
02Patient lab
Don’t take our word for it. Build your own.
Nothing is precomputed. Without FalkorDB, the lab has no answer at all.
Open the patient lab- 1Place medicines on a timelineand say who reported each one
- 2Write it to FalkorDBa new synthetic patient
- 3Watch the graph walk itsame queries, same rules, in milliseconds
03Proof
Checked, not claimed.
Run on the graph engine. Agreement with our documented rules, not clinical validation.
12/12
synthetic cases agree with the documented rules
98/98
checks agree, case by case
5/5
deliberately wrong expectations caught
0
facts lost after restarting FalkorDB and the API
04Memory
Nothing is overwritten.
Story 3: a late note changes a date. Watch what the graph keeps.
v1 · letter
swelling began day 28
v2 · late GP note
swelling began day 14
revision 1
Reviewed
revision 2
Needs a fresh look
clinician review
Kept on revision 1
Attached to what they actually saw.
05The stories
Three synthetic stories.
For a pharmacist's medication review. The clinician decides.
Reading the medication story
06Graph Hacks 2026
Two tracks, one graph.
- Agents That Act on Connected Data
The review agent walks the patient graph with allowlisted tools and records the next question.
- Agent Memory and Coordination
Facts, findings, agent steps and decisions are versioned in FalkorDB and survive restarts.