Why footprintjs?

The problem with opaque pipelines

Backend systems — loan processors, order workflows, ML pipelines — are increasingly complex chains of async operations. When they fail or behave unexpectedly, answering “why did it take that branch?” requires digging through logs, adding trace points, and often re-running with debug flags.

footprintjs makes pipelines self-documenting by construction. The narrative isn’t written after the fact — it’s a side effect of execution.

What you get for free

Decision evidence

When decide() routes to a branch, the narrative includes the exact values, operators, and thresholds that triggered it — without writing a single log statement.

AI-readable structure

chart.toSpec() returns the full graph structure as JSON. LLMs can traverse it, ask “what happens if X fails?”, and get accurate answers because the graph is the source of truth.

OpenAPI & MCP built-in

chart.toOpenAPI() and chart.toMCPTool() describe the chart as an API endpoint or AI tool — no separate schema maintenance.

Time-travel debugging

getSubtreeSnapshot() returns the full memory state at any subflow depth. Replay any branch from any point.

vs. plain async functions

// Before: opaque, no structure, manual logging
async function processLoan(input: LoanInput) {
  const score = await getCreditScore(input.id);
  if (score > 700) {
    await approveLoan(input.id);          // WHY was this taken?
  } else {
    await rejectLoan(input.id);
  }
}

// After: structured, self-documenting
const chart = flowChart<LoanState>('Score', async (scope) => {
  scope.creditScore = await getCreditScore(scope.$getArgs().id);
}, 'score')
  .addDeciderFunction('Route', (scope) =>
    decide(scope, [
      { when: { creditScore: { gt: 700 } }, then: 'approved', label: 'Good credit' }
    ], 'rejected')
  , 'route')
  // ...branches
  .build();