A Privacy-First Telehealth Intake Workflow

Intake Is Where Privacy Starts

The first text a patient sends is often the most sensitive. A privacy-first telehealth workflow keeps that rawest data local for as long as possible. The SPRAPP suite supports this with on-device TinyLM. SPRAPP is a tool, not a clinician; nothing here is medical advice.

Local Intake With TinyLM

The initial structuring of a patient's intake message runs on TinyLM. Because the eeny and meeny models run on-device, the raw message can be turned into a structured draft without leaving the patient's or clinician's device. For privacy regimes, keeping the rawest data local is the cleanest starting posture.

Screening Inbound Messages

Telehealth portals accept free text from patients, which is untrusted content. Sprappy Filter scores inbound messages across its 25 categories before any model processes them, catching injection attempts and malformed payloads.

De-Identified Second Opinions With Panel

When a clinician wants a second perspective on a complex case, a de-identified summary — not raw patient data — goes to SPRAPP Panel. The panel converges and reports agreement and disagreement. Divergence is a prompt to slow down, never a diagnosis. The clinician decides.

Compact, Long-Lived Records With smoltext

Telehealth records carry many short strings — appointment IDs, status codes, triage tags — over long retention windows. smoltext compresses these short strings well where gzip struggles on small payloads, keeping the record store compact.

The Human Stays in Charge

TinyLM keeps intake local, Filter guards the boundary, Panel offers de-identified perspectives, and smoltext keeps records lean. Every clinical judgment is human.

Cautious Adoption

Pilot TinyLM intake structuring in a non-production setting first, confirm the privacy posture with your compliance team, then consider de-identified Panel reviews.