AI production checklist

Pre-launch checklist for Livepeer AI applications: gateway selection, authentication, error handling, cost estimation, and monitoring.

---

Gateway selection

Environment	Gateway URL	When to use
Development / testing	https://dream-gateway.livepeer.cloud	Free, no API key required, unpredictable latency
Production	https://livepeer.studio/api/beta/generate	Studio-managed, authenticated, SLA-backed

Switch before launch. The public gateway has no rate limits, no SLA, and no support. Do not ship to users on the public gateway.

Authentication

• API key stored in environment variables, never in source code
• Backend API key used for server-side requests; CORS-enabled key for browser-side
• Key rotation schedule set (90-day recommended)
• .env files excluded from version control via .gitignore
• Dependency scanning (e.g., Dependabot, Snyk) configured to detect accidental key exposure

Error handling

• 401 handler: log and surface a configuration error — do not retry; the key is wrong
• 422 handler: log the full response body to identify the failing field; fix the request shape
• 503 handler: retry with exponential backoff; cold model load is expected behaviour, not a failure
• 429 handler: back off and retry; check rate limits in Studio dashboard
• Global timeout set on all requests (recommended: 300 seconds for cold model scenarios)
• SDK retry configuration enabled with exponential backoff (see SDK Integration Guide)

Model selection

Choose warm models for latency-sensitive paths. Cold model load times range from 30 seconds to several minutes. Warm models respond immediately.

If you need…	Use warm model
Text to image	SG161222/RealVisXL_V4.0_Lightning
Image to image	timbrooks/instruct-pix2pix
Audio transcription	openai/whisper-large-v3
LLM completion	meta-llama/Meta-Llama-3.1-8B-Instruct

Custom model_id values cold-start on every request until an orchestrator warms them. Test custom models under expected load before launch.

Cost estimation

AI inference pricing is orchestrator-set. The current indicative rate for text-to-image is approximately $0.019 per megapixel of output.

Illustrative example for text-to-image at 1024x1024:

• 1024 x 1024 = 1,048,576 pixels = 1.05 megapixels
• At $0.019/MP: approximately$0.020 per image
• At 1,000 images/day: approximately $20/day

Actual rates vary by orchestrator and pipeline. Check your Studio billing dashboard after the first week of production traffic to calibrate estimates against real costs.

Monitoring

• Log HTTP status codes for all gateway requests
• Track p50, p95, p99 latency per pipeline
• Alert on sustained 503 rate above 5% (indicates warm model unavailability)
• Alert on sustained 5xx rate above 1% (indicates inference failures)
• Monitor Studio billing dashboard weekly during ramp-up

Livepeer does not provide an SDK-level metrics collector. Instrument at the HTTP client layer.

Launch readiness summary

• Production gateway URL configured (livepeer.studio/api/beta/generate)
• Backend API key in secrets manager, not in code
• CORS-enabled key used for browser-side calls
• Retry policy implemented with exponential backoff
• All error types (401, 422, 503, 429) handled
• Warm models used for latency-sensitive paths
• Cost projection completed and budget alert set in Studio
• Status page subscribed: https://status.livepeer.studio
• Incident response contact: Livepeer Discord #builders or [e-mail]

Related pages

AI Authentication

API key types, CORS keys, and rotation strategy.

AI Troubleshooting

Diagnose errors by status code.

SDK Integration Guide

Retry configuration and error handling in TypeScript, Python, and Go.