Apppublic

numanakhter124/ai-symptoms-triage

sourceHugging Faceupdated 6mo agoView on Hugging Face

1likes

App README

AI Symptoms Triage Environment

An Openenv RL environment where an AI agent triages patients by reading symptoms and deciding severity, care pathway, and urgency — like a nurse doing intake triage.

Motivation

Medical triage errors cause preventable deaths. This environment trains agents to correctly classify patient urgency using symptom patterns.

Observation Space

Field	Type	Description
patient_id	str	Unique patient identifier
symptoms	List[str]	Reported symptoms
age_group	str	child / adult / elderly
duration	str	hours / days / weeks
existing_conditions	List[str]	Pre-existing conditions

Action Space

Field	Values
severity	mild / moderate / severe
pathway	homecare / seedoctor / emergencyroom / callambulance
timeframe	within24h / within4h / immediate

Tasks

—Easy: Obvious textbook symptoms (e.g. chest pain, runny nose)
—Medium: Multiple symptoms requiring reasoning (e.g. fever + headache)
—Hard: Red flag combinations where context changes everything (e.g. elderly + confusion + chest pain)

Reward

—Severity: 50%, Pathway: 30%, Timeframe: 20%
—Penalty: -0.3 if severe case called mild

Baseline Scores

—Easy: 1.0
—Medium: 1.0
—Hard: 1.0
—Overall: 1.0

Setup

bash

uv sync
uv run server

Inference

bash

export HF_TOKEN=your_token
python inference.py