F POLARIS · FHIR
polaris platform · public technical docs

One FHIR contract
for German ambulatory
care.

Polaris sits between PVS systems, Aidbox, and the apps or agents above them. It normalizes KBV, Gematik, Basis-DE, and Cognovis profile families into one stable platform surface without asking partners to learn each PVS schema.

R4
One FHIR R4 contract above every adapter boundary.
2
Public guide families highlighted directly on this site.
CLI
Operator tooling for health, sync, and reconciliation.
public guide families
upstream IG stack
  • tier 1 hl7.fhir.r4.core · de.basisprofil.r4 HL7 / HL7 DE
  • tier 1 kbv.basis · kbv.ita.for · kbv.ita.erp KBV
  • tier 2 erezept-workflow · epa.medication · vsdm2 · kbv.ita.eau Gematik / KBV

One FHIR contract above the adapters.
The PVS stays the system of record.

// Apps and agents target one FHIR contract.
// Adapters translate between Aidbox and the external source system.

interface PvsAdapter {
  readonly name: string
  connect(): Promise<void>
  sync(opts: SyncOptions): Promise<SyncResult[]>
  writeback(res: FhirResource): Promise<void>
}

// Builders are PVS-agnostic.
// They know profile rules, not SQL table names.
buildPatient(data) -> KBV_PR_FOR_Patient
buildEncounter(data) -> Encounter profile
buildChargeItem(data) -> ChargeItem / Claim profile
system-of-record matrix
  • pvs Billing · Diagnoses · Source master data system of record
  • polaris FHIR contract · Validation · Writeback orchestration platform layer
  • apps & agents Scheduling · Automation · Partner workflows build on the contract

Polaris does not ask downstream teams to learn MSSQL schemas, proprietary export formats, or dental billing edge cases. It isolates that complexity behind adapters and publishes one FHIR-facing contract to the rest of the platform.

Shipped adapters, planned coverage, and operator tooling.

shipped

Humanmedizin ambulant

Compiled PVS adapter for ambulatory medical practices. Specific source-system support is discussed per rollout and depends on available customer-side access.

development

Dental ambulant

Dental PVS adapter in development, using the same compiled delivery model and FHIR builder layer, tuned for dental billing and treatment workflows.

planned

Weitere PVS-Adapter

Additional ambulatory PVS targets are planned. Polaris treats each connector as an adapter capability rather than a public vendor partnership claim.

tooling

Polaris FHIR SDK · adapter CLI

The public developer surface is the Polaris FHIR SDK, currently published through @polaris/fhir-de. Runtime operations use the adapter CLI for health, sync, and reconciliation.

Two patterns for building on Polaris.

Polaris is infrastructure. Apps and agents consume it through one SDK surface, no matter which PVS sits under the adapter. Teams building on Polaris pick the pattern that fits their product — both are first-class supported.

variant a
partner replicates

Your FHIR store, synced from Polaris.

Partner app hosts its own FHIR store in the cloud. Data syncs out of the PVS through Polaris adapters and into the partner store. Partner owns UX and hosting end to end.

Partner SaaS app · cloud
Own FHIR store · own frontend
↑ FHIR sync
Polaris Adapter NDJSON export
Extraction · normalization · writeback
↑ read
PVS system of record
  • +Full sovereignty over data and UX
  • +No dependency on Polaris install base
  • Partner owns PVS-drift maintenance
  • Data leaves the practice to partner cloud
variant b
partner connects in

Dock into the CDP, data stays local.

Polaris runs in the practice. Partner cloud reaches in via WireGuard VPN and the SDK. Data can be synced selectively — or tunneled stateless. The §203 problem is solved by construction.

Partner SaaS app · cloud
Frontend · business logic
↓ VPN · SDK · OAuth2
Polaris CDP local · in practice
FHIR · Identity · Agent API · Audit
↓ local network
PVS system of record
  • +Data stays in the practice by default
  • +Adapter maintenance handled by Polaris
  • +Structural §203 compliance
  • Depends on Polaris rollout in the practice
coding-agent-ready

Your LLM tooling works out of the box.

The Polaris SDK was built for humans and agents on equal footing. The same typed interface that your developers use, Claude Code, Codex, or an in-house MCP client can use too — with anonymization and audit handled at the platform boundary, not at the app.

schema-typed

JSON Schema on every tool

MCP-native tool discovery via tools/list. Agents know the contract before they call.

mode-aware

Anonymized or real

Every tool accepts mode. Default is anonymized; real mode requires Practitioner JWT.

auditable

Every call logged

Caller, tool, arguments, mode, response fingerprint. Audit by construction, not by convention.

narrow by design

Use-case tools, not raw FHIR

getEncounterForCodingSuggestion, not readEncounter. Less PII leaks, cleaner responses.

Public docs

Start with the platform. Then use the SDK.

Polaris FHIR should answer two technical questions immediately: how the platform is installed, how the SDK is used once the platform exists, and how agent access is meant to work. FHIR-specific links still go straight to fhir.cognovis.de.

platform install

Install Polaris as a platform, not as a bag of packages.

Aidbox, deployment modes, identity, install-data, and rollout guidance now live in a dedicated platform hub.

  • Single-tenant, multi-org platform model
  • Read-only and bidirectional deployment modes
  • Identity and access control at the platform layer
Platform hub →

sdk

Use the Polaris FHIR SDK from one coherent surface.

The SDK hub separates the current public package surface from the broader SDK work that is still landing.

  • Today the package surface is @polaris/fhir-de.
  • Canonical constants should be imported, never hard-coded.
  • Broader Polaris SDK surfaces are marked in work where needed.
SDK hub →

agents

Document agents and LLM access from the ADRs onward.

The agent hub now explains the MCP agent surface and the LLM gateway direction, while stating clearly what is accepted and what is not shipped yet.

  • Inbound MCP surface from ADR-012
  • Outbound LLM gateway from ADR-020
  • Anonymization, auth, and audit model explained publicly
Agent hub →

Canonical stability

Canonical URLs stay stable even when the presentation layer grows better docs around them.

Stable canonicals

Canonical URLs identify the guide artifacts themselves. Public landing pages and docs can evolve without redefining those identifiers.

Registry consistency

Package distribution remains centralized through npm.cognovis.de for predictable installation and version pinning.

No duplicate guide body

This site should orient, summarize, and link outward. It should not fork the canonical guide content into a second uncontrolled IG tree.

Technical contact

Need deployment guidance, SDK help, or a deeper platform conversation around Polaris?

Reach out for public guide questions, platform install reviews, adapter rollout support, or SDK-level integration discussions.

Contact the maintainers