Check document against regex rules asynchronously.

Comprehensive evaluation of the model. Args: model: The trained model (already loaded) tokenizer: Th

Analyze prediction errors.

Compute comprehensive metrics for token classification.

Create confusion matrix from predictions and labels.

Get model predictions on the test dataset.

Create and save confusion matrix visualization.

Print evaluation summary.

Save evaluation results to file.

Chunk the large document between max and min context length of BERT-based model

Tokenizes the input document and predicts credential tokens.

Convenience function to call from training loop. Args: model: The model being trained tokenizer: The tokenizer t

Match and remove NoseyParker results

Prompt for the LLM

Scan for secrets using finetuned xlm-RoBERTa-base model

Run LLM check and confirmation on the retrieved xlm-RoBERTa-base model

Load all regex patterns from a JSONL file.

Removes all noseyparker regex matched patterns from the document.

Main function to scan the document for secrets

Set seed for all random number generators to ensure reproducibility. Args: seed (int): The seed value to use

Load regex patterns from a JSONL file.

Initialize the evaluator. Args: save_dir (str): Directory to save evaluation results and plots

DeepPass2 API call

Match a single regex pattern synchronously (runs in a thread pool).

Extracts tokens from text and applies tokenization.