// the find
awslabs/fhir-works-on-aws-deployment
A serverless implementation of the FHIR standard that enables users to focus more on their business needs/uniqueness rather than the FHIR specification
A serverless FHIR R4/STU3 server running on AWS Lambda, DynamoDB, S3, and Elasticsearch, deployed via CDK/Serverless Framework. Aimed at healthcare organizations or ISVs who want a compliant FHIR API without standing up their own infrastructure. Note: this repo has been migrated to aws-solutions/fhir-works-on-aws and is effectively archived.
- The component split (routing, authz, persistence, search as separate packages) means you can swap individual pieces without rewriting everything—practical for orgs with custom auth requirements.
- Includes a Java HAPI validator Lambda for strict IG validation, which is something most DIY FHIR servers skip entirely and then regret during ONC certification.
- Integration test suite covers bulk export, multi-tenancy, subscriptions, and SMART auth—more coverage than typical AWS Solutions launches.
- SMART on FHIR support as an alternative to Cognito RBAC is genuinely useful for interoperability scenarios where the client app manages its own IDP.
- Repo is abandoned—last push April 2023, README says it migrated to aws-solutions org. You'd be adopting a dead codebase with security notices about deprecated auth packages already in the README.
- Elasticsearch dependency is a cost and ops burden; AWS renamed it to OpenSearch and the migration path for existing deployments is manual and data-lossy per the troubleshooting section.
- Transaction bundle limit of 25 entries is hardcoded by default and the batch limit (750) is derived from a rough Lambda payload heuristic—neither is a real FHIR compliance guarantee and will bite anyone ingesting real-world CCD payloads.
- History search (type, system, and instance level) is explicitly marked 'not implemented', which is a mandatory FHIR capability for many regulatory use cases like USCDI compliance.