Rate limits & credits

predictAI doesn't enforce fixed requests-per-second limits. Instead, usage is governed by two mechanisms that map to what actually consumes resources:

Credits meter actions

Compute-heavy actions cost credits, priced per action and charged when the work runs:

• training runs (per run, scaled by intensity),
• inference calls,
• discovery runs and tournaments,
• large ingestion volumes (streaming events beyond the workspace's included allowance),
• BYOM uploads and validation.

If your balance can't cover an action you'll get a 402 — the action simply doesn't run. Check your balance and burn rate any time:

curl "$API_BASE/v1/billing/balance" \
  -H "Authorization: Bearer $TOKEN" \
  -H "X-Workspace-Id: $WORKSPACE_ID"

from predictai import PredictAI

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

data = client.billing.get_billing_balance()

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

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

const data = await client.billing.getBillingBalance();

Estimate before you spend with POST /v1/billing/estimate — every chargeable surface can be priced in advance.

Quotas cap objects

Your plan caps how many of each object a workspace can hold: models, pipelines, deployments, goals, and more. Hitting a cap returns a 403 with a message naming the limit. One call shows your live usage counts:

curl "$API_BASE/v1/billing/usage" \
  -H "Authorization: Bearer $TOKEN" \
  -H "X-Workspace-Id: $WORKSPACE_ID"

from predictai import PredictAI

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

data = client.billing.get_billing_usage()

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

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

const data = await client.billing.getBillingUsage();

Being a good citizen

• Prefer Realtime subscriptions over tight polling loops — training, deployment, and goal events are pushed to you.
• Batch historical data through Ingestion imports rather than thousands of individual signal/push calls.
• Cache stable reads (catalogs, plan info); they change rarely.

Full pricing, plans, and quota detail: Billing.