Changelog

All notable changes to the EduBotX ML API will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

---

[3.0.0]

Breaking Changes

• Removed deprecated V1 reusability endpoint (/api/predict/reusability with basic response)
• Removed deprecated endpoints: /predict, /recommend-treatment
• Renamed V2 reusability to main endpoint - /api/predict/reusability now returns the enhanced response with CPCB compliance
• Removed reusability_v2 model ID - Use reusability for all reusability predictions
• Updated WastewaterInput schema - Now includes optional effluent values (previously WastewaterInputV2)

Changed

• Unified reusability prediction - Single endpoint with full CPCB compliance, treatment recommendations, and effluent estimation
• Simplified API structure - Removed V1/V2 versioning in favor of single, feature-complete endpoints
• Updated model version - Reusability model now reports version 3.0
• Cleaner schema endpoint - Only reusability and treatment model IDs supported

Migration Guide

If upgrading from v2.x:

# Before (v2.x)
POST /api/predict/reusability      # Basic prediction
POST /api/predict/reusability/v2   # Enhanced with CPCB

# After (v3.0)
POST /api/predict/reusability      # Enhanced with CPCB (only option)

// Before (v2.x)
import { useReusabilityPredictionV2, WastewaterInputV2 } from "@/lib/ml-api";

// After (v3.0)
import { useReusabilityPrediction, WastewaterInput } from "@/lib/ml-api";

---

[2.1.0]

Added

• Treatment Endpoint Enhancements

  • use_case query parameter for treatment endpoint (Irrigation, Domestic Use, Industrial Use, Drinking Water, Aquaculture)
  • quality_score - Water quality score (0-100) based on input parameters
  • parameter_status - Good/Moderate/Poor classification for each parameter
  • chemical_suggestions - Chlorine and alum dosing recommendations, RO recommendation
  • risk_assessment - Risk level (Low/Medium/High) with reasons
  • cumulative_sequence - Full treatment sequence from Primary to recommended stage
  • predicted_stage vs recommended_stage differentiation (adjusted for use case)

• Documentation

  • Enhanced FastAPI OpenAPI documentation with proper tags and descriptions
  • Added CHANGELOG.md for version tracking
  • Updated API documentation with all new endpoints and features

Changed

• Cleaned up imports and removed unused dependencies
• Improved code organization with section headers
• Updated API version to 2.1.0

Fixed

• Removed deprecated validator import from pydantic

---

[2.0.0]

Added

• Reusability Prediction V2 (/api/predict/reusability/v2)

  • CPCB inland surface water compliance check
  • Treatment upgrade recommendations
  • Effluent value estimation when not provided
  • Smart verdict combining model prediction and CPCB status
  • Alternative reuse class suggestions

• Reference Endpoints

  • GET /api/class-info - Reusability class metadata
  • GET /api/cpcb-limits - CPCB discharge limits

• Batch Prediction (/api/predict/batch)

  • Process multiple samples in one request
  • Support for both reusability and treatment models

• PDF Report Generation (/api/report/generate)

  • Generate downloadable PDF treatment reports

• Model Schema Endpoint (/api/schema/{model_id})

  • Dynamic form generation support for Next.js

Changed

• Standardized API response format with status, data, message, timestamp
• Added global exception handlers for consistent error responses
• Migrated from on_event to lifespan context manager

---

[1.0.0]

Added

• Initial release

• Reusability Prediction (/api/predict/reusability)

  • Predict water reusability class based on wastewater parameters
  • Returns prediction, confidence, and class probabilities

• Treatment Recommendation (/api/predict/treatment)

  • Recommend treatment stage (Primary/Secondary/Tertiary/Advanced)
  • Returns treatment steps and descriptions

• Health Endpoints

  • GET / - API status
  • GET /health - Detailed health check with model status
  • GET /api/models - List available models

Models

• Reusability Predictor (v1) - Uses models_v1/ artifacts (joblib)
• Treatment Recommender (v1) - Uses models_v2/ artifacts (pickle)

---

Model Artifacts

models_v1/ (Reusability Prediction)

File	Description
model.joblib	Trained classifier model
imputer.joblib	Missing value imputer
scaler.joblib	Feature scaler
label_encoder.joblib	Class label encoder
features.joblib	Feature names list

models_v2/ (Treatment Recommendation)

File	Description
sih_model.pkl	Trained classifier model
sih_scaler.pkl	Feature scaler
sih_label_encoder.pkl	Stage label encoder

---

WebSocket Integration

Endpoints

Endpoint	Description
/ws/reusability	Real-time reusability predictions
/ws/treatment	Real-time treatment recommendations

SCADA Simulator

A Python script (ws_scada_simulator.py) is provided for testing WebSocket integration:

# Run the simulator
python ws_scada_simulator.py

Features:

• Connects to both WebSocket endpoints
• Generates realistic sensor data
• Supports sequential or parallel streaming
• Configurable frame count and interval

---

Last Updated: December 2024