Critical bug – Failure in the application of the automatic categorization rules engine

I am reporting a bug related to the automatic categorization rules mechanism again, as the problem persists and continues to directly impact the reliability of automation.

Currently, the system recognizes the existence of a certain categorization rule—even blocking the creation of a new rule with the same criteria due to uniqueness violation—which shows that the rule is correctly persisted in the database. However, the categorization engine is not applying this rule to newly imported transactions.

In this specific case, there is an active rule configured with the criterion "Description contains PAYGO," associated with the "Food and beverages" category. The rule was created correctly, remains visible in the management panel, and the system confirms its existence by preventing the creation of another identical rule. However, recent transactions whose description contains "PAYGO" continue to be classified as "Health and Wellness," explicitly contradicting the criteria defined by the user.

This behavior highlights a disconnect between the rule storage layer and the processing pipeline responsible for categorizing transactions. The interface indicates that the rule exists and is active, but the evaluation mechanism does not apply it when processing new entries.

Technically, the scenario suggests a failure in the integration between the rule engine and the transaction ingestion/import flow. Among the plausible technical hypotheses are:

• The rules are being stored correctly, but are not being consulted when the categorization engine is executed.

• There is a precedence conflict between user-defined rules and the default categorization algorithm, with possible subsequent overwriting.

• The categorization engine may be running before the custom rules are loaded or synchronized.

• There may be a failure in the normalization of the description string (case sensitivity, trim, encoding, invisible characters, or textual variations).

• The import process is not triggering a reassessment of existing rules.

• There may be an asynchronous recategorization job that is overwriting the classification applied by the rule.

The expected behavior is deterministic: active user-defined rules should take precedence over any default automatic categorization and should be applied consistently to all transactions that meet the established criteria.

Additionally, in case of conflict or inability to apply, the system should provide explicit feedback to the user, rather than appearing to function normally while silently ignoring the configured rule.

The problem is recurring, occurring with several transactions and not just with the case mentioned. This requires constant manual recategorization, generates rework, and significantly compromises confidence in the automation mechanism.

I therefore request a detailed verification of the integration between:

• the rule storage layer;

• the rule evaluation engine;

• transaction ingestion/import flow;

• automatic processing execution order;

• precedence logic between user rules and standard categorization;

• possible asynchronous recategorization processes.

Additionally, it would be highly recommended to provide a mechanism for manual reprocessing of rules for already imported transactions, allowing the current rules to be reapplied in batches if the engine is not applying them correctly at the time of ingestion.

The automatic rules functionality is essential for efficient financial organization. In its current state, it is not operating in a reliable or deterministic manner, which compromises its practical usefulness.

<img data-width="100%" data-align="left" alt="" data-featurebase-content-key="68b875d3f0d1b7dbae80ebab/post/698e4a843c0138d7eaaf7e5e/019c53d2-c0c8-7c77-a743-a82405d4f825/b64u-cGF5Z28ucG5n.png" data-featurebase-content-filename="paygo.png" src="https://68b875d3f0d1b7db...