Heidi Health · Rate Limits

Heidi Health Rate Limits

Heidi Health does not publish explicit numeric rate-limit values in its public developer documentation. Practical limits are governed by the partner contract, per-API-key throttling at the registrar.api.heidihealth.com edge, recommended audio-chunking cadence, and asynchronous status flags on long-running operations. This file captures the operational ceilings and guidance documented in the developer portal so partners can budget capacity and design clients defensively.

Heidi Health Rate Limits is the machine-readable rate-limit profile for Heidi Health on the APIs.io network, conforming to the API Commons Rate Limits specification.

The profile also includes 8 backoff/retry policies defined.

Tagged areas include Healthcare, Health Tech, AI Medical Scribe, Ambient AI, and Clinical Documentation.

0 Limits
HealthcareHealth TechAI Medical ScribeAmbient AIClinical DocumentationClinical Decision SupportArtificial IntelligenceSpeech To TextTranscriptionEHR IntegrationElectronic Health RecordsTelehealthClinical CodingICD-10SNOMEDHIPAAGDPRSOC 2ISO 27001ISO 42001WearablesVoiceAudioAustraliaMelbourne

Policies

Each Heidi API key represents one EHR/tenant. Per-key rate enforcement occurs at the registrar.api.heidihealth.com edge. Numeric ceilings are not published; partners receive their key-specific limits as part of onboarding.
Every JWT issued by GET /jwt has an `expiration_time` (ISO 8601). Clients must refresh the token before expiry and present it as `Authorization: Bearer ` on every subsequent call. Tokens are bound to a specific EHR user (email + third_party_internal_id).
For live ambient capture via the chunked transcription endpoints, Heidi recommends splitting audio into 45-second to 1-minute chunks to balance latency and accuracy. Smaller chunks degrade transcript quality; larger chunks delay live results.
Audio uploaded via POST /sessions/{session_id}/upload-audio is stored and transcribed lazily on the next GET /sessions/{session_id}/transcript call. Clients must tolerate first-read latency proportional to audio length.
Consult-note generation (POST /sessions/{session_id}/consult-note), custom-template generation, and Ask Heidi (POST /sessions/{session_id}/ask-ai) return chunked `application/x-ndjson` responses (`{"data": "..."}` objects). Clients must keep the connection open and concatenate chunks; there is no documented maximum response length but generation duration is proportional to consult complexity and template size.
GET /sessions/linked-user is paginated via `page_size`. Responses include `count`, `total_count`, `has_next`, `has_previous`. Maximum `page_size` not documented; recommended ≤ 200.
Context document uploads accept .pdf, .jpg, .png, .docx, .doc. Documents are parsed asynchronously and progress through PARSING -> READY status. Maximum file size not publicly documented.
MP3 and OGG are the recommended formats for optimal transcription quality. Other audio formats are accepted but may yield lower transcript fidelity.

Sources