heimdall

Processes personal finance data and a user-defined budget to generate indicators, control points, and warnings.

url
heimdall.dyseth.com

Heimdall is a system that analyzes personal financial information (income, expenses, recurring commitments, balances, and savings) along with a person’s budget and turns it into a continuously updated picture of financial health. The budget is the reference against which reality is measured. Its goal is not only to report where money went, but also to show how actual behavior compares with the plan, and to flag when the two are drifting apart.

Components

  • Ingestion and normalization. Data from bank exports, statements, and manual entries is parsed into a single consistent ledger. Transactions are deduplicated, categorized, and tagged by currency and account. The budget is loaded alongside it and mapped to the same categories and periods, so that planned and actual figures can be compared directly.
  • Budget model. The budget defines expected income, category limits, savings targets, and scheduled obligations such as rent, insurance, and tuition. It can be monthly, annual, or mixed, and it can be revised over time. Each version is retained, so past performance is always judged against the plan valid at the time.
  • Indicators. A set of metrics summarizes the state of the finances: savings rate, fixed-to-variable expense ratio, runway in months, category share of spending, debt-to-income ratio, and budget adherence (the gap between planned and actual by category). Each indicator is tracked over time, so trends are visible and not just snapshots.
  • Control points. Fixed checkpoints compare actuals against the budget at defined moments: mid-month, end of month, and at the close of each period. Each control point reports whether a category or goal is on track, at risk, or breached, and by how much. Projections to period end indicate where the current pace will land.
  • Warnings. Deviations are detected and ranked by severity. Examples include a category approaching or exceeding its limit, an unusual spike in spending, a recurring charge that has increased, a balance projected to fall below a safety threshold, a missing expected income, or a savings target that is no longer reachable. Every warning states what changed, why it matters, and which budget threshold it crossed.
  • Dynamic updating. The pipeline works incrementally. New transactions or budget changes update indicators, control points, and warnings as they arrive, without reprocessing the whole history.

Outcomes

  • Validation with synthetic ledgers and budgets that contain known deviations, as well as anonymized real-world personal finance data.
  • Metrics: detection delay, false-alarm rate, and precision and recall of the warnings, plus the accuracy of end-of-period projections against the budget and the stability of indicators under noisy or incomplete data.
  • Privacy by design: processing is local-first, and no financial data leaves the user’s control.

cd ..

Type the name of a menu item, such as scope, and press Enter to open it. Type help to list every command.

# cookie-settings

Choose which optional categories you allow. You can change this at any time from the cookie-settings link in the menu.