Deployments

Deploy any model from the BRICQS catalog to a real GPU runtime. Each deployment gets a live HTTPS endpoint, auto-scaling replicas, and real-time BRICQS metrics with no instrumentation required.

Environments

Every deployment belongs to one of three environments. A project can run all three simultaneously:

bashbricqs deploy staging --model meta-llama/Llama-3-8B-Instruct --env preview
bricqs deploy prod    --model meta-llama/Llama-3-8B-Instruct --env production

Promote to Production

Promote a running preview deployment to production from the dashboard — no downtime. BRICQS provisions a new production deployment, waits for it to reach running state, then retires the old production app.

Auto-scaling

Set min and max replicas per deployment. With min_replicas: 0, idle deployments scale to zero within minutes of last traffic and cost nothing until the next request arrives (cold start: ~30–60s).

bricqs deploy my-api --model meta-llama/Llama-3-8B-Instruct --min 1 --max 4

Metrics

Real-time BRICQS metrics are available for every running deployment from the dashboard Metrics tab — no app-level instrumentation required:

• CpuPercentage — container CPU utilization
• MemoryPercentage — container memory utilization
• GpuUtilizationPercentage — GPU utilization (GPU models only)
• Requests — total requests per time window
• ResponseTime — average response latency in ms
• Replicas — current replica count

Deploy via API

bashcurl -X POST https://api.bricqsai.com/deployments/deploy \
  -H "Authorization: Bearer <token>" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "my-api",
    "model_id": "meta-llama/Llama-3-8B-Instruct",
    "environment": "production",
    "min_replicas": 0,
    "max_replicas": 1
  }'

Rollback

Every deployment keeps a revision history. Roll back to any previous revision from the dashboard Revisions tab, or trigger a rollback via API. Rolling back re-provisions the prior container image without touching your current production endpoint until the rollback completes.