Written and maintained by CASRAI Editorial Board
Last updated
The Nabla Core API is a headless ambient clinical documentation API that enables electronic health record (EHR) vendors, telehealth platforms, and digital health engineering teams to embed real-time medical speech recognition, multi-speaker diarization, structured clinical note generation, and normalized medical coding directly into their own applications. Alongside this developer API, Nabla Copilot provides turnkey integration with Epic and enterprise EHRs via HL7 SMART on FHIR, Epic Haiku/Hyperspace context launch, and browser DOM extensions.
What Is the Nabla Core API?
While clinicians can use Nabla as a standalone web app or Chrome extension, health systems and software vendors building custom clinical workspaces often require white-label, API-level control over audio capture and note rendering. The Nabla Core API (`docs.nabla.com`) exposes Nabla’s ambient AI pipeline as programmable WebSocket and REST endpoints without forcing clinicians to leave their native EHR or telehealth user interface.
The Nabla Core API architecture separates client-side audio streaming from server-side authentication and clinical LLM synthesis:
- Server-to-Server JWT Token Exchange: Your backend server authenticates with the Nabla Core API using a long-lived server API key to mint a short-lived, encounter-scoped JSON Web Token (JWT) (`access_token` and `refresh_token`). This short-lived token is passed to the clinician’s browser or mobile app (`Authorization: Bearer <JWT>`), ensuring secret keys are never exposed in client-side code.
- Real-Time WebSocket Audio Streaming & Diarization: The client application opens a persistent secure WebSocket connection (`wss://core.api.nabla.com/v1/core/user/transcribe_ws`) and streams raw PCM (`pcm_s16le`, 16 kHz) or Opus-encoded audio frames, along with configuration parameters (encounter language such as English, French, or Spanish, and single mixed microphone stream vs. dual separated audio channels for telehealth). The WebSocket returns real-time incremental and finalized transcript objects labeled by speaker role (`PATIENT`, `DOCTOR`, or `OTHER`).
- REST Clinical Note Generation: Once the encounter concludes (or on demand mid-encounter), the client or server invokes the note-generation REST endpoint (`/v1/core/server/generate_note` or `/v1/core/user/generate_note`), passing the structured transcript segments, patient context (age, sex, active problems), and a desired template schema (e.g., SOAP note, 14-system Review of Systems, specialty template, or custom section prompts).
- Normalized Clinical Entity Extraction: Using the `/v1/core/server/generate_normalized_data` endpoint, the API extracts discrete clinical entities from the generated note and maps them to standard medical terminologies—including ICD-10-CM diagnosis codes, RxNorm medication concepts, SNOMED CT findings, and LOINC vital signs and laboratory observations.
How Does Nabla Copilot Integrate with Epic and EHR Systems?
For health systems deploying Nabla across ambulatory, emergency, or inpatient care, Nabla Epic integration and broader EHR connectivity operate through three tiered integration patterns depending on the health system’s IT governance and Epic App Market (Showroom) configuration:
| Integration Tier | Target EHR Environment | Technical Mechanism & Launch Context | Data Read & Writeback Capabilities |
|---|---|---|---|
| 1. Native Epic SMART on FHIR & Haiku / Hyperspace Integration | Epic Hyperspace (desktop), Epic Haiku (iOS/Android), Epic Canto (iPad). | Launched directly inside the active patient chart via HL7 SMART on FHIR OAuth 2.0 EHR launch or Epic ambient voice integration APIs. | Automatically reads patient demographics, encounter context, problem list, and active medications; writes approved note sections back to Epic SmartData Elements (SDEs) or note templates. |
| 2. Direct API / Bi-Directional EHR Connectors | athenaOne (athenahealth Marketplace), Greenway Health, NextGen, Canvas Medical, Elation Health. | Authenticated REST / FHIR R4 partner APIs linked to the provider’s schedule and encounter ID. | Pulls daily appointment schedule and patient chart summary; pushes discrete HPI, ROS, Physical Exam, and Assessment & Plan fields into the open encounter note. |
| 3. Web Browser Extension & Desktop Companion | Any web-based or Citrix-hosted EHR (eClinicalWorks, MEDITECH Expanse, Oracle Health Millennium, Cerner). | Chrome/Edge extension side-panel or macOS/Windows desktop app requiring zero hospital interface engine build. | One-click field mapping (DOM injection) or structured clipboard copy-paste into EHR note sections. |
In an Epic AI scribe deployment, clinicians typically initiate ambient recording on their mobile device (Epic Haiku or the Nabla mobile app synced via SMART on FHIR context) when entering the exam room, and open Epic Hyperspace at their workstation to find the drafted HPI, Physical Exam, Assessment & Plan, and patient instructions staged for review and signature.
What HL7 or FHIR Interfaces Does Nabla Copilot Support?
Health informatics architects integrating Nabla into enterprise clinical data pipelines rely on standardized HL7 FHIR R4 resources and terminology bindings to ensure AI clinical documentation maps cleanly into downstream analytics and billing workflows:
- HL7 SMART on FHIR Launch Specification (OAuth 2.0 / OpenID Connect): Supports standard EHR launch (`launch`, `patient/*.read`, `encounter/*.read`, `user/*.read`, `DocumentReference.write`, `Observation.write` scopes), automatically binding the ambient recording session to the active FHIR `Patient` and `Encounter` resource IDs so notes can never be accidentally filed under the wrong patient chart.
- Inbound Context Resources (FHIR R4 Read): Queries `Patient` (age, administrative sex, preferred language), `Encounter` (visit type, specialty department), `Condition` (active problem list for chronic disease context), and `MedicationRequest` / `MedicationStatement` (current regimen to improve ASR recognition of complex drug names).
- Outbound Clinical Documentation & Discrete Entities (FHIR R4 Write / Core API JSON):
- Clinical Note Narrative: Exported as a FHIR
DocumentReference(orComposition) resource coded with LOINC document ontology codes (e.g., LOINC11506-3Progress note or34117-2History and physical note) and section-level LOINC codes (`10164-2` History of Present Illness, `10187-3` Review of Systems, `29545-1` Physical findings, `51847-2` Assessment and Plan). - Structured Vitals & Findings: Numeric vital signs spoken aloud during the physical exam (blood pressure, heart rate, temperature, SpO2, weight) are extracted by the Nabla Core API into normalized JSON objects that map directly to LOINC-coded FHIR
Observationresources. - Coded Problems, Allergies, and Orders: Candidate diagnoses map to ICD-10-CM and SNOMED CT (`Condition`), newly mentioned allergies map to `AllergyIntolerance`, and discussed laboratory or imaging orders stage as draft `ServiceRequest` / `MedicationRequest` candidates for physician sign-off.
- Clinical Note Narrative: Exported as a FHIR
Data Privacy, Zero Audio Retention, and Security Architecture
When comparing Nabla against other Abridge AI competitors (such as DAX Copilot, Ambience Healthcare, Suki, and DeepScribe), enterprise security teams evaluate how patient audio and transcripts are governed:
- Stateless / Zero Persistent Audio Storage: By default, Nabla processes audio streams in memory and does not persistently store patient voice recordings after the clinical note is generated. Customers using the Nabla Core API can operate in stateless mode where neither audio nor transcripts are retained on Nabla servers post-response.
- HIPAA BAA, SOC 2 Type II, and ISO 27001 Certification: Nabla executes HIPAA Business Associate Agreements (BAAs) with covered entities and technology partners, maintains SOC 2 Type II and ISO/IEC 27001 certifications, and complies with EU GDPR health data processing rules.
- No Foundation Model Training on Customer PHI without Consent: Patient audio and clinical transcripts processed under an enterprise BAA or Nabla Core API agreement are strictly isolated and are not used to train shared commercial foundation models.
Frequently Asked Questions
How does Nabla Copilot integrate with Epic and EHR systems?
Nabla integrates with Epic via HL7 SMART on FHIR launch inside Epic Hyperspace and mobile workflows alongside Epic Haiku/Canto, automatically pulling patient and encounter context and pushing clinician-approved note sections back into Epic note templates or SmartData Elements. For other EHRs (such as athenaOne, NextGen, Greenway, eClinicalWorks, and Cerner), Nabla connects via direct REST/FHIR partner APIs, the headless Nabla Core API, or the Nabla Chrome extension.
What HL7 or FHIR interfaces does Nabla Copilot support?
Nabla supports HL7 SMART on FHIR (OAuth 2.0 / OpenID Connect) for contextual EHR launch, HL7 FHIR R4 read interfaces (`Patient`, `Encounter`, `Condition`, `MedicationRequest`) for pre-encounter context, and outbound FHIR `DocumentReference` / `Observation` mapping alongside the Nabla Core API’s WebSocket (`/transcribe_ws`) and REST (`/generate_note`, `/generate_normalized_data`) endpoints that output structured clinical notes coded with ICD-10-CM, SNOMED CT, RxNorm, and LOINC.
