FisiHome — take-home physiotherapy SaaS

Problem

Physiotherapy and pilates professionals prescribe take-home exercises through conversation, so he/she can’t track if the patient practiced at home — correctly or not. Gym-training focused apps do not model clinical rehab workflows (per-body-region treatment tracks, consultation vs. home days, professional review) and might not address Brazilian health-data protection under LGPD.

Approach

FisiHome is being built as a two-persona Flutter app — professional and patient — on top of an async Python/FastAPI API with SQLAlchemy and Alembic. Professionals register patients by invite only, organise assignments into treatment tracks by clinical classification and pull exercises from a shared library or upload patient-specific videos; patients see each day’s assignments, watch the exercise video, record themselves and receive written feedback. This allows a more consistent and agile clinical progression, since movement repetition on a daily basis allows a progressive strenghtening, subject to constant rectification.

Key architectural decisions:

• Pooled multi-tenancy on PostgreSQL Row-Level Security, with a path to bridge/silo isolation depending on hired tier by clinics.
• Videos never touch the cluster — clients upload straight to S3 object storage with presigned URLs and play back through signed CDN URLs, so the compute tier only carries API traffic.
• Self-hosted auth in the API: JWT access/refresh, Argon2id hashing and TOTP MFA for professionals, matching the invite-only onboarding rules.
• Development environment on a 2-node k3s cluster (Raspberry Pi 5) exposed via Cloudflare Tunnel, provisioned with Terraform and instrumented with OpenTelemetry into Grafana/Prometheus/Loki/Tempo, designed to be moved to Cloud once patient volume justifies it.

Impact

• Foundation milestone in progress: API skeleton, domain model, RLS tenancy, auth and invite flows, plus the Kubernetes and Terraform baselines.
• Roadmap sequenced into: prescribe → video pipeline → close-the-loop review → LGPD export/delete, so each milestone ships something a real clinic can use.
• Cloud-native media from day one keeps the eventual managed-cloud migration limited to the API and database.