Model Hangar
Use the right engine Models are replaceable providers. Choose by task, privacy, quality, latency, and cost—and record what was actually tested.
REFERENCE PATTERN
How to organize it
GPT → complex reasoning Local model → approved private work Image model → still artwork Video model → motion generation Fallback → documented alternative
This is a public educational pattern, not a live connection to the computer behind EchoNet. Adapt it to your own environment.
PRACTICAL START
Starter prompt
Choose models for [TASK]. Compare privacy, capability, cost, latency, context, fallback, and verification. Never assume availability without testing.
Inspect first, keep the smallest useful scope, preserve rollback, and verify with direct evidence.