ideogram-observability

# Ideogram Observability ## Overview Monitor Ideogram AI image generation for latency, cost, error rates, and content safety rejections. Key metrics: generation duration (5-25s depending on model), credit burn rate, safety filter rejection rate, and API availability. Ideogram's API is synchronous, so all observability is request-level instrumentation. ## Key Metrics | Metric | Type | Labels | Alert Threshold | |--------|------|--------|-----------------| | `ideogram_generation_duration_ms` | Histogram | model, style, speed | P95 > 25s | | `ideogram_generations_total` | Counter | model, status | Error rate > 5% | | `ideogram_credits_estimated` | Counter | model | >$10/hour | | `ideogram_safety_rejections` | Counter | reason | >10% rejection rate | | `ideogram_image_downloads` | Counter | status | Download failures > 1% | ## Instructions ### Step 1: Instrumented Generation Wrapper ```typescript import { performance } from "perf_hooks"; interface GenerationMetrics { duration: number; model: string; style: string; status: "success" | "error" | "safety_rejected" | "rate_limited"; seed?: number; resolution?: string; } const metricsLog: GenerationMetrics[] = []; async function instrumentedGenerate( prompt: string, options: { model?: string; style_type?: string; aspect_ratio?: string } = {} ) { const model = options.model ?? "V_2"; const style = options.style_type ?? "AUTO"; const start = performance.now(); try { const response = await fetch("http...