US-C.3not implemented Train a classification model on my own categorised data
As a user who has manually categorised hundreds of transactions over time I want use my categorised transaction history as training data for a custom finetuned classification model so that the AI suggestions become increasingly accurate for my personal spending patterns.

Acceptance Criteria

  • A training dataset is generated from previously classified transactions.
  • The finetuned model is stored locally.
  • The model can be retrained as more data accumulates.
  • Performance metrics (accuracy, precision, recall per category) are reported.
No DAG diagram: Not yet implemented — training pipeline; produces a model artifact rather than data flowing through the transaction DAG layers.
Demo video coming soon

DAG Diagram