New enhanced_reasoning parameter on agent chat requests. When enabled, the agent adds an evaluation step after tool execution to assess whether the user's intent was fulfilled, needs more info, or encountered an error. This produces more reliable, self-correcting responses.
Agent interactions now support real-time streaming via Server-Sent Events (SSE)! Stream agent responses incrementally and get visibility into tool execution as it happens. Available on the API (/agent/stream-chat) and all SDKs.
Construe now includes a terminology server API for managing custom code systems. Use it to store and query proprietary medical terminologies alongside standard vocabularies like SNOMED CT, LOINC, and ICD-10.
The lang2fhir document endpoint now automatically detects file types, eliminating the need for manual fileType parameter specification.
Extracted medical codes now include citations showing where codes appear in source documents, improving traceability and auditability of AI-generated coding decisions.
fhir2summary now supports International Patient Summary mode!
When you use fhir2summary with IPS mode it will take your FHIR bundle and generate International Patient Summary narratives per ISO 27269 and HL7 FHIR IPS IG standards. Auto-creates sections for allergies, medications, problems, immunizations, procedures, and vital signs. No template is required, you just set mode to ips. Requires Bundle with exactly one Patient resource with identifiers.
Use cases: context engineering, care transitions, referrals, cross-border exchange, standardized summaries.
We've released a Claude Code plugin that lets you use Claude as a forward deployed agent for building healthcare data workflows.
Describe what you need in plain language (e.g. FHIR mapping pipelines, data transformations, EHR integrations) and Claude orchestrates Phenoml APIs to build it for you. The plugin connects Claude directly to our workflows and FHIR Provider APIs!
Use it to:
Check out the tutorial video to see it in action!
We're excited to announce the beta release of fhir2summary - a new feature of lang2FHIR that enables you to compress and summarize FHIR resources!
At release fhir2summary supports two modes: narrative and flatten. Flatten mode converts FHIR resources to a stringified form inspired by the work of Sam Schiffman. Narrative mode enables you to create custom templates for summaries you want to generate from FHIR resources. You can create summary templates for use cases ranging from context engineering, FHIR data compression, to chart summarization. Provide fhir2summary an example summary and representative resources and our AI will generate a reusable template for you to use on subsequent resources.
Coming soon:
We're excited to announce the beta release of workflows - orchestrate healthcare data operations using natural language!
Workflows uses lang2FHIR to translate your natural language instructions into structured FHIR operations and generates a DAG of steps – FHIR searches, resource creation, and decision nodes. Pass in unstructured data like clinical notes or legacy system outputs, and workflows handles extraction, transformation, and orchestration across your FHIR systems.
You can execute workflows across multiple FHIR providers (Epic, Cerner, Athena Health, Elation, Medplum, Canvas Medical, Google Healthcare API, HAPI) or a single one by specifying the fhir_provider_id in your workflow requests! Control whether to use deterministic templates (default) or dynamic generation. One seamless API for healthcare data orchestration.
Coming soon:
