Select an endpoint, edit the request body, and hit Run to call the live API.
Every endpoint returns a unified JSON envelope with prediction, confidence, risk level, and actionable insights.
Predicts diabetes probability using a Gradient Boosting model trained on the Pima Indians Diabetes Database (768 samples). Handles physiologically impossible zero values via median imputation.
| Field | Type | Required | Description |
|---|---|---|---|
| pregnancies | integer | Yes | Number of pregnancies (0–20) |
| glucose | number | Yes | Plasma glucose concentration (mg/dL) |
| blood_pressure | number | Yes | Diastolic blood pressure (mmHg) |
| skin_thickness | number | Yes | Triceps skinfold thickness (mm) |
| insulin | number | Yes | 2-hour serum insulin (μU/mL). Use 0 if unknown. |
| bmi | number | Yes | Body mass index (kg/m²) |
| diabetes_pedigree | number | Yes | Diabetes pedigree function (family history score) |
| age | integer | Yes | Age in years |
Random Forest model trained on 1,943 combined records from the Cleveland Heart Disease study and Heart Failure Prediction dataset. Cholesterol=0 is treated as missing and median-imputed.
| Field | Type | Required | Description |
|---|---|---|---|
| age | integer | Yes | Age in years |
| sex | integer | Yes | 0 = Female, 1 = Male |
| chest_pain_type | integer | Yes | 0=Asymptomatic, 1=Atypical Angina, 2=Non-Anginal, 3=Typical Angina |
| resting_bp | number | Yes | Resting blood pressure (mmHg) |
| cholesterol | number | Yes | Serum cholesterol (mg/dL). Use 0 if unknown. |
| fasting_bs | integer | Yes | Fasting blood sugar >120 mg/dL (1=True, 0=False) |
| max_hr | number | Yes | Maximum heart rate achieved (bpm) |
| exercise_angina | integer | Yes | Exercise-induced angina (1=Yes, 0=No) |
| oldpeak | number | Yes | ST depression induced by exercise |
Histogram Gradient Boosting with adjusted class weights (1:20) to handle severe class imbalance (4.8% positive rate in training data). Decision threshold is set to 0.30 for high-recall detection — high_risk_flag is true when probability ≥ 0.30.
| Field | Type | Required | Description |
|---|---|---|---|
| gender | string | Yes | Male, Female, Other |
| age | number | Yes | Age in years |
| hypertension | integer | Yes | 1=Yes, 0=No |
| heart_disease | integer | Yes | 1=Yes, 0=No |
| ever_married | string | Yes | Yes or No |
| work_type | string | Yes | Private, Self-employed, Govt_job, children, Never_worked |
| residence_type | string | Yes | Urban or Rural |
| avg_glucose_level | number | Yes | Average glucose level (mg/dL) |
| bmi | number | Yes | BMI (kg/m²) |
| smoking_status | string | Yes | formerly smoked, never smoked, smokes, Unknown |
SVM with RBF kernel trained on the Wisconsin Diagnostic dataset (569 samples, ~97% 10-fold CV accuracy). Input grouped into mean, standard error, and worst-case measurements of 10 nucleus characteristics.
Runs all three risk models from a single consolidated input. Returns individual scores for each condition plus an overall risk level and the highest-concern condition. Ideal for initial patient screening. Shared fields (age, sex, bmi, blood_pressure) are dispatched to each model automatically.
MobileNetV2 fine-tuned on ~7,200 brain MRI images, exported to INT8-quantized ONNX (~3.5MB). Classifies into: glioma, meningioma, pituitary tumor, or no tumor. Accepts base64-encoded JPEG or PNG.
| Field | Type | Required | Description |
|---|---|---|---|
| image_base64 | string | Yes | Base64-encoded JPEG or PNG brain MRI image |
Random Forest classifier trained on the Sleep Health and Lifestyle Dataset (374 samples). Predicts None, Insomnia, or Sleep Apnea probability. Also returns a composite wellness score (0–100) from sleep duration, quality, daily steps, and stress level.
| Field | Type | Description |
|---|---|---|
| occupation | string | Software Engineer, Doctor, Nurse, Teacher, Accountant, Lawyer, etc. |
| bmi_category | string | Normal, Overweight, Obese |
| systolic_bp / diastolic_bp | integer | Blood pressure components (split from the "126/83" format) |
Compares your daily activity metrics against pre-computed percentiles from 33 FitBit users (2016). Returns steps/calories/active-minutes percentiles and a fitness category. Reference is a small convenience sample — use as indicative benchmarks only.
Gradient Boosting Regressor trained on WHO data for 193 countries (2000–2015). Accepts 19 socioeconomic and health indicators and predicts life expectancy in years with a 90% confidence interval.
Returns pre-aggregated analytics from 55,500 patient records across 6 conditions and 5 age groups. Includes billing statistics, admission type breakdown, length of stay, medication patterns, and insurance distribution.
| Parameter | Required | Values |
|---|---|---|
| condition | Yes | Arthritis, Asthma, Cancer, Diabetes, Hypertension, Obesity |
| age_group | No | 0-17, 18-34, 35-49, 50-64, 65+ |
Three Gradient Boosting Regressors — mean estimate, 5th percentile (lower bound), 95th percentile (upper bound). Trained on 1,338 US insurance records. The smoker×BMI interaction term is included as an explicit feature.
| Field | Type | Description |
|---|---|---|
| age | integer | Age in years |
| sex | string | male or female |
| bmi | number | Body mass index |
| children | integer | Number of dependents |
| smoker | string | yes or no |
| region | string | southwest, southeast, northwest, northeast |
Returns the model registry JSON including algorithm name, training dataset, CV accuracy, number of samples, and artifact SHA256 hash for each model.
All models are pre-trained offline and loaded at startup. No training happens at inference time.
AUC = Area Under ROC Curve (classification). R² = coefficient of determination (regression). Metrics from 5-fold stratified cross-validation unless noted.