# vkuzina/cassed architecture

> Architecture diagram of the GitHub repository vkuzina/cassed, generated by GitDiagram from its file tree, README and a sample of source files (last updated 2026-10-02). It is an AI-made overview: check details against the source.

- Interactive diagram: https://gitdiagram.com/vkuzina/cassed
- Repository: https://github.com/vkuzina/cassed

## Overview

CASSED detects multilabel sensitive-data categories in structured table columns. The main workflow converts standard CSV values and labels into Flair sentence data, then trains a BERT-backed classifier and evaluates it; README also documents using pretrained models. Source excerpts establish preparation and training wiring, but test_model.py and several documented runtime details were not sampled, so those links rely on README evidence.

Main entry points: [`train_classifier.py`](https://github.com/vkuzina/cassed/blob/main/train_classifier.py) (Training workflow), [`test_model.py`](https://github.com/vkuzina/cassed/blob/main/test_model.py) (Model testing), [`config/run_config.py`](https://github.com/vkuzina/cassed/blob/main/config/run_config.py) (Run configuration), [`prepare_data.py`](https://github.com/vkuzina/cassed/blob/main/prepare_data.py) (CSV converter)

## Diagram (Mermaid)

```mermaid
flowchart TD

subgraph group_data["Data preparation"]
  node_standard[("Standard datasets")]
  node_config["Run configuration<br/>[run_config.py]"]
  node_prepare["CSV converter<br/>[prepare_data.py]"]
  node_prepared[("Prepared datasets")]
end

subgraph group_model["Model training"]
  node_train["Training workflow"]
  node_corpus["Flair corpus"]
  node_embedding["Transformer embeddings"]
  node_classifier["Multilabel classifier"]
  node_sentence["Sentence splitting"]
  node_trainer["Flair trainer"]
end

subgraph group_evaluation["Evaluation and inference"]
  node_saved[("Trained model")]
  node_test["Model testing<br/>[test_model.py]"]
  node_evaluation["Classifier evaluation"]
  node_results["Evaluation results<br/>[run_config.py]"]
  node_pretrained[("Pretrained models")]
end

node_user(("Dataset user"))

node_user -->|"provides data"| node_standard
node_user -->|"sets paths"| node_config
node_config -->|"configures paths"| node_prepare
node_standard -->|"reads CSVs"| node_prepare
node_prepare -->|"writes sentences"| node_prepared
node_config -->|"configures training"| node_train
node_prepared -->|"loads splits"| node_corpus
node_train -->|"creates corpus"| node_corpus
node_train -->|"initializes"| node_embedding
node_train -->|"constructs"| node_classifier
node_classifier -->|"splits sentences"| node_sentence
node_train -->|"initializes and runs"| node_trainer
node_trainer -->|"trains"| node_classifier
node_trainer -->|"saves model"| node_saved
node_user -->|"runs testing"| node_test
node_test -->|"loads model"| node_saved
node_test -->|"selects model"| node_pretrained
node_test -->|"evaluates predictions"| node_evaluation
node_train -->|"evaluates test split"| node_evaluation
node_evaluation -->|"reports metrics"| node_results

click node_standard "https://github.com/vkuzina/cassed/tree/main/datasets"
click node_config "https://github.com/vkuzina/cassed/blob/main/config/run_config.py"
click node_prepare "https://github.com/vkuzina/cassed/blob/main/prepare_data.py"
click node_prepared "https://github.com/vkuzina/cassed/tree/main/datasets/dessi_prepared"
click node_train "https://github.com/vkuzina/cassed/blob/main/train_classifier.py"
click node_corpus "https://github.com/vkuzina/cassed/blob/main/train_classifier.py"
click node_embedding "https://github.com/vkuzina/cassed/blob/main/train_classifier.py"
click node_classifier "https://github.com/vkuzina/cassed/blob/main/extended_flair/extended_text_classifier.py"
click node_sentence "https://github.com/vkuzina/cassed/blob/main/extended_flair/extended_sentence.py"
click node_trainer "https://github.com/vkuzina/cassed/blob/main/train_classifier.py"
click node_test "https://github.com/vkuzina/cassed/blob/main/test_model.py"
click node_evaluation "https://github.com/vkuzina/cassed/blob/main/extended_flair/extended_text_classifier.py"
click node_results "https://github.com/vkuzina/cassed/blob/main/config/run_config.py"

classDef toneNeutral fill:#f8fafc,stroke:#334155,stroke-width:1.5px,color:#0f172a
classDef toneBlue fill:#dbeafe,stroke:#2563eb,stroke-width:1.5px,color:#172554
classDef toneAmber fill:#fef3c7,stroke:#d97706,stroke-width:1.5px,color:#78350f
classDef toneMint fill:#dcfce7,stroke:#16a34a,stroke-width:1.5px,color:#14532d
classDef toneRose fill:#ffe4e6,stroke:#e11d48,stroke-width:1.5px,color:#881337
classDef toneIndigo fill:#e0e7ff,stroke:#4f46e5,stroke-width:1.5px,color:#312e81
classDef toneTeal fill:#ccfbf1,stroke:#0f766e,stroke-width:1.5px,color:#134e4a
class node_standard,node_config,node_prepare,node_prepared,node_user toneBlue
class node_train,node_corpus,node_embedding,node_classifier,node_sentence,node_trainer toneAmber
class node_saved,node_test,node_evaluation,node_results,node_pretrained toneMint
```

## Components

### Data preparation

- [Standard datasets](https://github.com/vkuzina/cassed/tree/main/datasets): `datasets`
- [Run configuration](https://github.com/vkuzina/cassed/blob/main/config/run_config.py): `config/run_config.py`
- [CSV converter](https://github.com/vkuzina/cassed/blob/main/prepare_data.py): `prepare_data.py`
- [Prepared datasets](https://github.com/vkuzina/cassed/tree/main/datasets/dessi_prepared): `datasets/dessi_prepared`

### Model training

- [Training workflow](https://github.com/vkuzina/cassed/blob/main/train_classifier.py): `train_classifier.py`
- [Flair corpus](https://github.com/vkuzina/cassed/blob/main/train_classifier.py): `train_classifier.py`
- [Transformer embeddings](https://github.com/vkuzina/cassed/blob/main/train_classifier.py): `train_classifier.py`
- [Multilabel classifier](https://github.com/vkuzina/cassed/blob/main/extended_flair/extended_text_classifier.py): `extended_flair/extended_text_classifier.py`
- [Sentence splitting](https://github.com/vkuzina/cassed/blob/main/extended_flair/extended_sentence.py): `extended_flair/extended_sentence.py`
- [Flair trainer](https://github.com/vkuzina/cassed/blob/main/train_classifier.py): `train_classifier.py`

### Evaluation and inference

- **Trained model**
- [Model testing](https://github.com/vkuzina/cassed/blob/main/test_model.py): `test_model.py`
- [Classifier evaluation](https://github.com/vkuzina/cassed/blob/main/extended_flair/extended_text_classifier.py): `extended_flair/extended_text_classifier.py`
- [Evaluation results](https://github.com/vkuzina/cassed/blob/main/config/run_config.py): `config/run_config.py`
- **Pretrained models**

### Other components

- **Dataset user**

## Connections

- Dataset user → Standard datasets: provides data (evidence: `README.md`)
- Dataset user → Run configuration: sets paths (evidence: `README.md`)
- Run configuration → CSV converter: configures paths (evidence: `prepare_data.py`)
- Standard datasets → CSV converter: reads CSVs (evidence: `prepare_data.py`)
- CSV converter → Prepared datasets: writes sentences (evidence: `prepare_data.py`)
- Run configuration → Training workflow: configures training (evidence: `train_classifier.py`)
- Prepared datasets → Flair corpus: loads splits (evidence: `train_classifier.py`)
- Training workflow → Flair corpus: creates corpus (evidence: `train_classifier.py`)
- Training workflow → Transformer embeddings: initializes (evidence: `train_classifier.py`)
- Training workflow → Multilabel classifier: constructs (evidence: `train_classifier.py`)
- Multilabel classifier → Sentence splitting: splits sentences (evidence: `extended_flair/extended_text_classifier.py`)
- Training workflow → Flair trainer: initializes and runs (evidence: `train_classifier.py`)
- Flair trainer → Multilabel classifier: trains (evidence: `train_classifier.py`)
- Flair trainer → Trained model: saves model (evidence: `README.md`)
- Dataset user → Model testing: runs testing (evidence: `README.md`)
- Model testing → Trained model: loads model (evidence: `README.md`)
- Model testing → Pretrained models: selects model (evidence: `README.md`)
- Model testing → Classifier evaluation: evaluates predictions (evidence: `README.md`)
- Training workflow → Classifier evaluation: evaluates test split (evidence: `train_classifier.py`)
- Classifier evaluation → Evaluation results: reports metrics (evidence: `extended_flair/extended_text_classifier.py`)
