FHIR-native APIs, mature SDKs, clear architecture docs. Integrate in days, not quarters — without ever moving raw data.
from lumkomdx import Federation
fed = Federation(token=TOKEN)
cohort = fed.cohort(
diagnosis="E11",
age_range=(40, 65),
last_hba1c__gt=7.5,
)
print(cohort.count(), cohort.sites)Identical semantics across SDKs and HTTP. The federation layer handles routing, consent, and audit transparently.
from lumkomdx import Federation
fed = Federation(token=TOKEN)
# Build a federated cohort — runs at every site
cohort = fed.cohort(
diagnosis="E11", # Type 2 diabetes
age_range=(40, 65),
last_hba1c__gt=7.5,
)
print(f"{cohort.count():,} patients across {cohort.sites} sites")
# › 12,847 patients across 14 sitesimport { Federation } from "@lumkomdx/sdk";
const fed = new Federation({ token: TOKEN });
// Build a federated cohort — runs at every site
const cohort = await fed.cohort({
diagnosis: "E11", // Type 2 diabetes
ageRange: [40, 65],
lastHbA1cGt: 7.5,
});
console.log(`${cohort.count} patients · ${cohort.sites} sites`);
// › 12,847 patients · 14 sitescurl -X POST https://api.lumkomdx.com/v1/cohort \
-H "Authorization: Bearer $LUMKO_TOKEN" \
-H "Content-Type: application/json" \
-d '{
"diagnosis": "E11",
"age_range": [40, 65],
"last_hba1c_gt": 7.5
}'Every clinical resource the spec defines, exposed as a federated endpoint. Your query fans out to participating sites; aggregated results return.
| Method | Endpoint | Purpose |
|---|---|---|
| GET | /v1/Patient | Pseudonymous resource list |
| GET | /v1/Observation | Lab + vitals (site-local) |
| POST | /v1/cohort | Federated cohort query |
| POST | /v1/insight | Studio reasoning + cite |
| POST | /v1/consent/verify | Pre-flight check |
| GET | /v1/audit/{id} | Signed audit trail |
Production SDKs with identical semantics, type-safe handles, and built-in retry. Pick the one your team already runs.
pip install lumkomdxnpm install @lumkomdx/sdkgo get github.com/lumkomdx/sdk-goinstall.packages('lumkomdx')Adapters for every major EMR, deployable pipelines, and a federation node that runs inside your VPC. Your data never leaves the building.
Ship-ready adapters for every major EMR — plus a generic FHIR adapter for anything else, including bespoke schemas.
Run normalisation and enrichment at the source. The federation node reads them; raw data never leaves the host.
0 2 * * *Nightly cohort refreshon: patient.admitTrigger on EMR eventkafka://hl7-v2Continuous ingestDeploy the federation node to your own VPC with Terraform or Helm. We ship the modules; you keep the keys, the audit log, and the traffic.
# Helm — Kubernetes cluster inside your VPC
helm repo add lumkomdx https://charts.lumkomdx.com
helm install fed lumkomdx/federation-node \
--set token=$LUMKO_TOKEN \
--set vpc.id=$VPC_ID
# Terraform — same shape, infra-as-code
terraform apply -var="region=af-south-1"Operator-controlled. The federation node runs on your infrastructure with your secrets. LumkoMDX never has a copy of your raw clinical data — and is structurally unable to access it.
Guides for getting up the curve, a complete API reference, and architecture deep-dives — written like the code: precise and short.
https://api.lumkomdx.com/v1
Long-form deep dives on federation, identity boundaries, the consent model, and threat surface — written for senior engineers.
Most institutions are running production federated queries within two weeks. The federation node is a single deployment; your existing FHIR endpoints handle the rest.
A single federation node — a containerised service that receives queries, executes them locally against your data, and returns aggregated results. No raw data ever leaves.
Yes. The federation layer ships with adapters for OMOP, i2b2, and bespoke schemas. Custom adapters typically take a week.
Every query is signed, logged, and consent-checked at the patient level — at execution time, not sign-up. The audit log is yours; we never have copies.
Sandbox is best-effort. Production is 99.9% with regional failover, monitored at the federation node and the API gateway. SLO dashboards are part of the operator console.
The API is versioned in the URL (/v1). SDKs follow semver. Breaking changes get six months of overlap. Architecture guides flag deprecations as they happen.
Get a working federated environment, sample synthetic cohorts, and a token. No NDA required for the sandbox.
# Step 1 — install
pip install lumkomdx
# Step 2 — authenticate
export LUMKO_TOKEN=sandbox.****
# Step 3 — your first federated query
python -c "from lumkomdx import Federation; \
print(Federation().sites)"
# ['hospital_jhb', 'clinic_lagos', 'lab_nairobi', ...]