Scale Dedicated deployments

Dedicated deployments give you reserved serving capacity — and you decide how much. replicas is the number of serving copies answering requests in parallel:

curl -X POST "$API_BASE/v1/deployments/dedicated/$DEPLOYMENT_ID/scale" \
  -H "Authorization: Bearer $TOKEN" \
  -H "X-Workspace-Id: $WORKSPACE_ID" \
  -H "Content-Type: application/json" \
  -d '{ "replicas": 3 }'

from predictai import PredictAI

client = PredictAI(token="pa_live_…", workspace_id="ws_…")

data = client.deployments.post_deployments_dedicated_by_deployment_id_scale(DEPLOYMENT_ID, json={"replicas": 3})

import { PredictAI } from "@predictai/sdk";

const client = new PredictAI({ token: "pa_live_…", workspaceId: "ws_…" });

const data = await client.deployments.postDeploymentsDedicatedByDeploymentIdScale(DEPLOYMENT_ID, { json: { replicas: 3 } });

{
  "success": true,
  "deployment_id": "3ac1f0b9-…",
  "replicas": 3,
  "message": "Deployment scaled to 3 replicas"
}

Request body

The call returns as soon as the new target is accepted; capacity converges in the background. Watch ready_deployments on Dedicated serving status to see it land, or subscribe to deployment.updated on the realtime channel.

When scaling matters

• Traffic spikes — add replicas ahead of a known peak so latency stays flat; drop back down afterwards.
• Throughput — one replica serves one request at a time per slot; more replicas mean more concurrent forecasts.
• Pausing — replicas: 0 parks the deployment without deleting anything. The record and its configuration stay; scale back up to resume serving.

Only Dedicated deployments scale this way — Shared deployments sit in the multi-tenant pool and the platform manages their capacity. Dedicated capacity is billed as dedicated hosting rather than per request; see Billing before running a high replica count.

Errors