Data models

Task

interface Task {
  id: string;
  status: "queued" | "running" | "complete" | "failed";
  repo: string;
  issueNumber: number;
  agent: string;
  prompt: string;
  machineId?: string;
  prUrl?: string;
  createdAt: Date;
  completedAt?: Date;
  error?: string;
}

Field intent:

• id: globally unique task identifier.
• status: current lifecycle phase.
• machineId: Fly runtime handle for operational introspection.
• prUrl: set when task results in a pull request.
• error: terminal failure details for UI and debugging.

User config

interface UserConfig {
  flyToken: string;
  anthropicKey: string;
  ampKey?: string;
  openaiKey?: string;
  defaultAgent: string;
}

Notes

• status tracks the end-to-end execution lifecycle.
• machineId links a task to Fly runtime state.
• prUrl is set when agent execution results in a pull request.

Modeling recommendations

• Keep status enum small and explicit.
• Use nullable optional fields for runtime-dependent metadata.
• Store timestamps in UTC and render locale at presentation time.

Extension ideas

• Add durationMs for performance analytics.
• Add attempt for retries and replay workflows.
• Add actor metadata to correlate request ownership and audit trails.