AUXY tracks token consumption by calculating the real API cost across models and features, applying a global retail multiplier to burn credits from a transactional ledger while enforcing pre-run gates and auto-reloads for pay-as-you-go accounts.

sequenceDiagram
  autonumber
  actor User
  participant Gate as Pre-Run Spend Gate
  participant Exec as Execution Engine
  participant Usage as Usage Recorder
  participant Ledger as Credit Ledger & Balance
  participant Stripe as Auto-Reload (Stripe)

  User->>Gate: Trigger Run (Probe / Optimizer)
  alt Balance <= 0 & Cannot Auto-Reload
    Gate-->>User: 402 Blocked (Prompt Recharge)
  else Balance > 0 or Auto-Reload Active
    Gate->>Exec: Execute Model Workload
    Exec->>Usage: Record Token Counts & Latency
    Usage->>Ledger: Calculate Real Cost x Multiplier
    Ledger->>Ledger: Deduct Entry & Update Balance
    alt Balance < Threshold & Auto-Reload On
      Ledger->>Stripe: Charge Off-Session PaymentIntent
      Stripe->>Ledger: Post-Payment Credit Entry
    end
    Exec-->>User: Return Run Output
  end