How manoj-devvvvv/drug-food-ai-upgraded works
This repository builds and serves a drug–food interaction recommender. Its user workflow is a drug/food request to the FastAPI service, recommendation and risk result; the serving engine combines model-based prediction with curated rules and supporting resolution, applicability-domain, calibration, and explanation capabilities. The broader research workflow covers cleaning and labeling data, generating pair features, selecting and training models, then evaluating them. Evidence is limited to the supplied source excerpts and SOURCE INDEX; several preparation-stage connections are represented as components without guessed links.
Main entry points: recommend.py Recommendation engine · common.py Shared config and utilities · brick8_models.py Model training · features.py Molecular pair features
Components 20
Serving and recommendations
- FastAPI service
web/app.py - Recommendation engine
src/recommend.py - Name resolution
src/resolve.py - Curated interaction rules
src/rules.py - Applicability domain
src/domain.py - Probability calibration
src/calibration.py - Prediction explanation
src/explain.py
Data and features
- Drug cleaning
src/brick2_clean_drugs.py - Food cleaning
src/brick3_clean_foods.py - Pair labeling
src/brick4_labels.py - Molecular pair features
src/features.py
Model development
- Feature selection
src/brick7_selection.py - Model training
src/brick8_models.py - Deployed model selection
src/brick17_single_model.py
Validation and evidence
- Data splits
src/splits_extra.py - Model evaluation
src/brick9_evaluate.py - Domain calibration analysis
src/brick14_domain_calibration.py - Generalization analysis
src/brick15_generalization.py - Shared config and utilities
src/common.py
Other components
- Drug/food user
Connections 30
- Drug/food user → FastAPI service: submits request
web/app.py - FastAPI service → Recommendation engine: requests recommendation
web/app.py - FastAPI service → Curated interaction rules: reports rule version
web/app.py - Recommendation engine → Name resolution: resolves names
src/recommend.py - Recommendation engine → Curated interaction rules: checks rules
src/recommend.py - Recommendation engine → Applicability domain: checks applicability
src/recommend.py - Recommendation engine → Probability calibration: calibrates probabilities
src/recommend.py - Recommendation engine → Prediction explanation: generates explanation
src/recommend.py - Recommendation engine → Molecular pair features: builds pair features
src/recommend.py - Recommendation engine → Model training: loads model
src/recommend.py - Drug cleaning → Shared config and utilities: uses utilities
src/brick2_clean_drugs.py - Food cleaning → Shared config and utilities: uses utilities
src/brick3_clean_foods.py - Pair labeling → Shared config and utilities: uses utilities
src/brick4_labels.py - Feature selection → Molecular pair features: builds features
src/brick7_selection.py - Feature selection → Shared config and utilities: uses utilities
src/brick7_selection.py - Model training → Molecular pair features: builds features
src/brick8_models.py - Model training → Shared config and utilities: uses utilities
src/brick8_models.py - Deployed model selection → Model training: reuses model methods
src/brick17_single_model.py - Deployed model selection → Shared config and utilities: uses utilities
src/brick17_single_model.py - Model evaluation → Molecular pair features: builds features
src/brick9_evaluate.py - Model evaluation → Model training: uses estimators
src/brick9_evaluate.py - Model evaluation → Shared config and utilities: uses utilities
src/brick9_evaluate.py - Domain calibration analysis → Model training: uses model artifacts
src/brick14_domain_calibration.py - Domain calibration analysis → Applicability domain: fits domain thresholds
src/brick14_domain_calibration.py - Domain calibration analysis → Data splits: uses split helpers
src/brick14_domain_calibration.py - Domain calibration analysis → Shared config and utilities: uses utilities
src/brick14_domain_calibration.py - Generalization analysis → Deployed model selection: reuses model setup
src/brick15_generalization.py - Generalization analysis → Molecular pair features: builds features
src/brick15_generalization.py - Generalization analysis → Data splits: uses split helpers
src/brick15_generalization.py - Generalization analysis → Shared config and utilities: uses utilities
src/brick15_generalization.py