bonfireDB vs Google Cloud Healthcare API

Cloud Healthcare API gives you a rock-solid, serverless FHIR store under a Google HIPAA BAA — but it's persistence-plus-eventing, not an application: you still build the app, the per-patient authorization, the realtime layer, and — once Google's FHIR-native semantic-search product sunsets (EOL 2027-05-15) — your own agent stack. bonfire is designed for the opposite starting point — typed clinical primitives, fresh-on-commit reads, and ABAC-enforced semantic search with FHIR emitted underneath.

TL;DR

Google Cloud Healthcare API is a genuinely mature, GA managed service — multi-modal FHIR/HL7v2/DICOM, profile validation on write, de-identification, BigQuery/Pub-Sub integration, a native (read-side) Consent engine, and named enterprise users like Mayo and Hackensack Meridian — and for interoperability, research data harmonization, and analytics it is a strong, low-ops foundation. But it is a data plane, not an app: a small team still builds the UI, domain logic, per-patient/per-tenant authorization, and the entire realtime fan-out, and FHIR search is eventually consistent with no strong-read option. bonfireDB is early-stage and largely vision, so its column reflects design intent, not a shipped, certified product. Where the two genuinely diverge is app-builder ergonomics: fresh-on-commit operational reads, a reactive client cache, write-enforced ABAC with auto-written audit, build-your-own-MCP over clean projections, and built-in ABAC-enforced semantic search — areas where Google's offering is eventually consistent, read-only-consent, log-based-audit, and (for semantic search) shipping today but sunsetting in 2027.

At a glance

Google Cloud Healthcare API is a mature, fully-managed FHIR/HL7v2/DICOM data plane for enterprise interoperability and analytics; bonfireDB is an early-stage app-native clinical backend that generates FHIR underneath so a small team can ship an outpatient or AI healthcare product fast.

Google Cloud Healthcare APIbonfireDB
TypeManaged FHIR/HL7v2/DICOM data plane (not an app)Agent-native clinical backend; FHIR generated underneath
Language / stackClosed-source Google-managed; REST/gRPC + client libsTypeScript + Postgres + pgvector
License / costProprietary pay-as-you-go; free tier, no license feeApache-2.0 core (free) + managed tier
HostingGoogle-managed only; no self-host or VPCYour AWS (OSS) or managed
App-native typed APINo — raw FHIR resources, no typed app primitivesYes — typed clinical primitives
Fresh-on-commit app readsNo — search indexed async; only identifier synchronousYes — committed operational read models
Reactive realtime cacheNo — Pub/Sub change events, not client pushYes — useClinicalQuery
Built-in semantic searchShips today — Agent Search for healthcare is FHIR-native, but sunsets 2027-05-15Built in — pgvector, ABAC-enforced
Agent / MCP layerNone built in — roll your own over FHIRBuild-your-own-MCP over clean projections
Clinical authorization (write-enforced)IAM (store-level) + read-only Consent; writes not enforcedRead + write enforced ABAC, auto-audit
Automatic audit / provenanceCloud Audit Logs (platform), not auto FHIR AuditEventsAutomatic
FHIR conformance / exportMature DSTU2/STU3/R4; validation, history, Bulk $exportFHIR R4 generated underneath; scoped conformance (honest: not a full enterprise FHIR server)
Best forEnterprise interop, data harmonization, analytics pipelinesBuilders of AI-native health apps (scribes, copilots, agents)

Where Google Cloud Healthcare API genuinely wins

  • Truly managed, serverless FHIR store with zero engine ops — no servers to patch or scale, an 11-nines durability claim, and a Google Cloud HIPAA BAA out of the box, which removes major compliance and infra burden from a small team.
  • Multi-modal by design: FHIR + HL7v2 + DICOM in one service, plus built-in de-identification (dataset/store/resource level) and FHIR Bulk Data $export — strong for interoperability and research/analytics pipelines.
  • First-class GCP integration: Pub/Sub change notifications, streaming and batch export to BigQuery, Dataflow, and Vertex AI make it easy to bolt analytics and ML onto clinical data.
  • Standards-conformant FHIR (DSTU2/STU3/R4) with CapabilityStatement, history/versioning (vread), transactions/bundles via executeBundle, conditional create/update, and StructureDefinition/IG profile validation on write.
  • A native FHIR Consent engine (patient consents, admin policies, cascading policies via X-Consent-Scope) plus SMART-on-FHIR scope and patient-context support — more access-control machinery than most managed FHIR stores ship, if you stay within its constraints.
  • Proven at enterprise scale with named, high-credibility production users (Mayo Clinic, Hackensack Meridian Health, SulAmerica), backed by ISO 27001/27017/27018 and PCI DSS at the platform level.

Where bonfireDB is built different

  • Fresh-on-commit app reads: Cloud Healthcare API indexes FHIR search asynchronously, so writes don't immediately appear in search results — only search-by-identifier is synchronous, and there is no caller-selectable strong-consistency option. bonfire is designed around committed operational read models so create-then-list flows are immediately consistent. (Verified: Google docs state resources are indexed asynchronously with identifier as the only synchronous index.)
  • Write-enforced clinical authorization: GCH IAM stops at the store level and its FHIR Consent enforcement is read-only — POST/PUT/PATCH/DELETE are explicitly not consent-enforced, with caps of 200 active consents per patient and 200 admin policies per store. bonfire's design enforces ABAC on both reads and writes. (Verified: Google docs say write methods are not supported by consent enforcement.)
  • Automatic in-store audit/provenance: GCH gives you Cloud Audit Logs (GCP platform logs), not auto-written FHIR AuditEvent/Provenance resources on every read and write. bonfire is designed to emit audit automatically as a first-class clinical record. (Verified: GCH surfaces operations via Cloud Audit Logs; FHIR AuditEvent/Provenance are app-authored, not auto-generated.)
  • Built-in, durable, ABAC-enforced semantic search: credit where due — Google's Agent Search for healthcare ships real FHIR-native semantic search today, which most FHIR stores don't have. The catch is durability and access control: it is sunsetting (EOL May 15, 2027, after which you'd rebuild on something else) and even today applies ACLs app-side rather than enforcing per-patient consent at query time. bonfire is designed to build semantic search in (pgvector) with ABAC applied per hit. (Verified: Google docs carry the explicit "won't be available after May 15, 2027" sunset notice on the currently-shipping FHIR-native search.)
  • App-state and agent layer as first-class: GCH is FHIR persistence only, so app/draft/workflow state needs a second database and any agent/MCP layer is yours to build. bonfire is designed with typed clinical primitives plus build-your-own-MCP over clean projections so app state has a home and agents work against typed data rather than raw FHIR.
  • Reactive realtime to the client: GCH realtime is Pub/Sub fire-and-forget change notifications (not FHIR Subscriptions and not client push) — you build the entire fan-out to the browser yourself. bonfire is designed with a reactive query cache (useClinicalQuery) so the UI updates without you hand-rolling the realtime pipeline.

Which should you choose?

Choose Google Cloud Healthcare API if…

Choose Google Cloud Healthcare API if you need a battle-tested, fully-managed, multi-modal (FHIR + HL7v2 + DICOM) store under a Google HIPAA BAA — especially for interoperability, data harmonization, de-identification, and BigQuery/Vertex analytics at enterprise scale — and you have the engineering to build the application, authorization, realtime, and agent layers on top.

Choose bonfireDB if…

Choose bonfire if you're a small team shipping an outpatient or AI healthcare app and want app-native typed primitives, fresh-on-commit reads, write-enforced ABAC with automatic audit, and built-in semantic search — accepting that bonfire is early-stage with scoped (not full enterprise) FHIR conformance, and that you'll still emit standards-conformant FHIR R4 underneath.

bonfireDB is early-stage; this compares its design and positioning against Google Cloud Healthcare API as it ships today (2026). Verify current Google Cloud Healthcare API capabilities before deciding.

You build the app. Bonfire is the clinical data layer underneath.

Open source. Runs in your AWS. The clinical layer is handled.

FAQ

Frequently asked questions

What's a good Google Cloud Healthcare API alternative?

Google Cloud Healthcare API is a mature managed FHIR/HL7v2/DICOM data plane, but it's persistence, not an application. bonfireDB is designed as the app backend that generates FHIR underneath (not a FHIR server): typed clinical primitives, fresh-on-commit reads, and write-enforced ABAC, with FHIR R4 generated underneath. It's early access, so it fits if you want app-builder ergonomics over a full enterprise interop store.

Is Google's FHIR-native semantic search being sunset?

Google's docs carry an explicit notice that the currently-shipping FHIR-native search (Agent Search for healthcare) won't be available after May 15, 2027 (verify current status with Google). bonfireDB is designed to build semantic search in directly via pgvector, with ABAC applied per hit at query time.

Is Google Cloud Healthcare API FHIR search strongly consistent?

No. Per Google's docs, FHIR resources are indexed asynchronously, so writes don't immediately appear in search results; only search-by-identifier is synchronous, and there's no caller-selectable strong-read option. bonfireDB is designed around committed operational read models so create-then-list flows are immediately consistent.

Does Google Cloud Healthcare API enforce per-patient authorization on writes?

Per Google's docs, IAM stops at the store level and its FHIR Consent enforcement is read-only — write methods aren't supported by consent enforcement. bonfireDB's design enforces ABAC on both reads and writes, with audit written automatically. Verify Google's current consent capabilities before deciding.

Can I self-host Google Cloud Healthcare API in my own VPC?

No — as of 2026 it's Google-managed only, with no self-host or VPC option (verify current). bonfireDB's Apache-2.0 core is designed to run in your own AWS, with an optional managed tier. It's early access today.