rohitg00--ai-engineering-from-scratch
wehub-resource-sync
91 行
3.3 KiB
JSON
91 行
3.3 KiB
JSON
{
|
|
"lesson": "29-production-runtimes",
|
|
"title": "Production Runtimes: Queue, Event, Cron",
|
|
"questions": [
|
|
{
|
|
"stage": "pre",
|
|
"question": "What are the six runtime shapes the lesson lists?",
|
|
"options": [
|
|
"Request-response, streaming, durable execution, queue-based, event-driven, scheduled",
|
|
"Read, write, exec, fork, exit, wait",
|
|
"REST, gRPC, GraphQL, WebSocket, SSE, MQTT",
|
|
"Map, filter, reduce, fold, scan, group"
|
|
],
|
|
"correct": 0,
|
|
"explanation": "These are the six runtime shapes; pick a shape before picking a framework."
|
|
},
|
|
{
|
|
"stage": "pre",
|
|
"question": "When is request-response a poor fit?",
|
|
"options": [
|
|
"On a 1-second classification",
|
|
"When the task takes longer than about 30 seconds and users hang up while workers pile up",
|
|
"On REST APIs",
|
|
"On internal services"
|
|
],
|
|
"correct": 1,
|
|
"explanation": "Synchronous HTTP is only viable for short tasks; long tasks need queue/durable/event shapes."
|
|
},
|
|
{
|
|
"stage": "check",
|
|
"question": "What is durable execution's core property?",
|
|
"options": [
|
|
"Faster inference",
|
|
"State is checkpointed after every step so the runtime can resume from the last successful step on failure",
|
|
"Lower token cost",
|
|
"Native voice support"
|
|
],
|
|
"correct": 1,
|
|
"explanation": "Checkpoint-and-resume is the differentiator; LangGraph is the lesson's reference."
|
|
},
|
|
{
|
|
"stage": "check",
|
|
"question": "Which Anthropic observation justifies queue-based runtimes for long-horizon agents?",
|
|
"options": [
|
|
"Tasks always complete in 5 seconds",
|
|
"Computer use announcement: dozens-to-hundreds of steps per task is normal",
|
|
"Tasks must run synchronously",
|
|
"Cron solves everything"
|
|
],
|
|
"correct": 1,
|
|
"explanation": "Anthropic flagged dozens-to-hundreds of steps per task as normal; that workload needs queues or durable runtimes."
|
|
},
|
|
{
|
|
"stage": "check",
|
|
"question": "What does the lesson mean by 'observability is load-bearing'?",
|
|
"options": [
|
|
"Logs are pretty",
|
|
"Without OTel GenAI spans and a Langfuse/Phoenix/Opik backend you cannot debug a multi-step agent that failed at step 40",
|
|
"Disk gets full",
|
|
"Providers require it for billing"
|
|
],
|
|
"correct": 1,
|
|
"explanation": "Observability is not optional for multi-step agents; it is the difference between debugging and replaying from scratch."
|
|
},
|
|
{
|
|
"stage": "post",
|
|
"question": "Why must queue workers have a dead-letter queue (DLQ)?",
|
|
"options": [
|
|
"Without DLQ, failed jobs vanish silently",
|
|
"DLQ is required by AWS",
|
|
"DLQ makes jobs faster",
|
|
"DLQ is encrypted"
|
|
],
|
|
"correct": 0,
|
|
"explanation": "Failed jobs without DLQ disappear; DLQ is the parking lot for failed jobs."
|
|
},
|
|
{
|
|
"stage": "post",
|
|
"question": "What does the lesson recommend pairing cron-shaped agents with?",
|
|
"options": [
|
|
"GPU autoscalers",
|
|
"Durable execution so a failing nightly run resumes next tick",
|
|
"Streaming SSE",
|
|
"WebRTC"
|
|
],
|
|
"correct": 1,
|
|
"explanation": "Cron + durable execution recovers cleanly from failed scheduled runs."
|
|
}
|
|
]
|
|
}
|