// the find
samply/blaze
A FHIR® Server with internal, fast CQL Evaluation Engine
Blaze is a FHIR R4 server written in Clojure that embeds a CQL evaluation engine, so population-level CQL and $evaluate-measure queries run inside the server instead of against an exported copy of the data. It is aimed at health IT teams in research networks and biobanks that need interactive aggregate queries over millions of patients.
Running CQL inside the store avoids a separate extract-and-load pipeline before aggregate queries, which is the core architectural bet of the project. The distributed storage variant has a Cassandra and Kafka setup under .github/distributed-test, so horizontal scaling is exercised in CI rather than only described. The README says it is in production use at Medical Informatics Initiative sites and European biobanks, and the paper linked from it gives a citable reference point for the design.
Clojure on the JVM narrows the contributor pool. Most FHIR tooling is Java, TypeScript, or Python, so anyone who needs to debug or extend the server has to read Clojure. LOINC and SNOMED CT content is licensed, so you have to import it yourself before the terminology services return anything useful. The quick start is a single docker run, and deployment guidance lives only on the external docs site. The distributed variant adds Cassandra and Kafka to the deployment, which is a lot of moving parts for a team that only needs one node.