How cjdearaujo/kharviml works
KharviML is a fish price and catch-volume prediction product for Turkish coastal fisheries. Users submit fish, area, time, and environmental measurements in the frontend; the API shapes them into model features and returns price or catch predictions for display. The repository also includes raw-data preprocessing and drift-reporting capabilities. Training and orchestration are documented by the presence of DVC configuration and stored MLflow artifacts, but their implementation sources were not provided, so those flows are not detailed here.
Main entry points: main.py Prediction API · mlartifacts Stored models · app.js Prediction form · preprocess.py Data preprocessing
Components 12
Prediction experience
- Prediction form
frontend/app.js
Prediction serving
- Prediction API
src/api/main.py - Feature builder
src/api/main.py - Price model
src/api/main.py - Catch model
src/api/main.py - Stored models
mlartifacts
Data preparation
- Raw fish data
data/raw/fish_catch_data_turkey.csv - Data preprocessing
src/data/preprocess.py - Processed dataset
Model monitoring
- Drift reporting
monitoring/drift_report.py - Drift reports
Other components
- Prediction user
Connections 11
- Prediction user → Prediction form: enters inputs
frontend/app.js - Prediction form → Prediction API: requests predictions
frontend/app.js - Prediction API → Feature builder: builds features
src/api/main.py - Prediction API → Price model: predicts price
src/api/main.py - Prediction API → Catch model: predicts catch
src/api/main.py - Stored models → Price model: loads price model
- Stored models → Catch model: loads catch model
- Raw fish data → Data preprocessing: reads data
- Data preprocessing → Processed dataset: writes dataset
src/data/preprocess.py - Processed dataset → Drift reporting: reads dataset
monitoring/drift_report.py - Drift reporting → Drift reports: saves reports
monitoring/drift_report.py