CoolFace
Apppublic

ffr0517/sepsis-spotter-ui

sourceHugging Facemitupdated 10mo agoView on Hugging Face
0likes
App README

Sepsis Spotter Project Repository Quick Links

Build API UI Manuscript

Spot Sepsis – UI Repository

Mobile-friendly research interface for interacting with the Spot Sepsis models (S1/S2). For research use only — not medical advice.

Current Implementation (Legacy-Compatible)

This version of the app connects to the previous Spot Sepsis API build (legacy endpoints). It provides a chat-based interface that collects clinical and laboratory information and runs the Spot Sepsis S1 and S2 models through either:

  • —Hugging Face Spaces (remote inference)
  • —Localhost (127.0.0.1) for self-hosted testing

Core Features

  • —Conversational intake assistant that builds a structured “Info Sheet.”
  • —Username/password research access gate.
  • —Sheet merge + resume for continuing sessions.
  • —“Run S1” / “Run S2” buttons gated by data completeness.
  • —Deterministic parser mode for use without an API key.

Planned Upgrades (Next Major Integration Phase)

This UI will be refactored once the new spot-sepsis-build, and spot-sepsis-api repositories reach their stable release.

AreaPlanned EnhancementTarget Behaviour
Input HandlingDropdown, CSV upload, enhanced LLM-assisted entryMultiple structured input paths
ValidationAutomatic parity testing with backend outputs≥ 99–100 % equivalence before deployment
TestingRed-team prompt and robustness testingIdentify edge/failure cases
Adaptive LearningOpt-in anonymised “live-model” modeAllows continual learning with clinic-controlled sharing
Version GovernanceBenchmark-driven model promotion (PR-AUC criteria)Only validated updates become active
User AnalyticsClinic dashboards and optional outcome loggingLocal record-keeping + contribution toggle
Privacy Controls“Share data with developers” toggleClear separation between private / shared data

Deployment

Automatic deployment is handled by .github/workflows/deploy-to-space.yml.

Workflow summary

  1. 1.Checks out this repository.
  2. 2.Clones the target Hugging Face Space using HF_SPACE token + SPACE_ID.
  3. 3.Syncs app.py, requirements.txt, README.md
  4. 4.Commits and pushes any changes.