Introduction

Power BI’s legacy real‑time “streaming” and “push” semantic models are in long‑term support only. As of October 31, 2024, Microsoft no longer allows creation of new real‑time semantic models (Push, Streaming, PubNub, Streaming tiles). Existing models continue to run but are scheduled for retirement on October 31, 2027. This cookbook shows how Nimble Way customers can: (a) keep existing streaming/push dashboards working safely, (b) push rows via REST with correct schemas and throttling, and (c) plan a forward path to Microsoft Fabric Real‑Time Intelligence or Azure Stream Analytics while continuing to use Nimble for live web data.

What still works (December 15, 2025)

• Existing Streaming or Push semantic models continue to accept rows via the Power BI REST “PostRows” endpoint and can back real‑time dashboard tiles and pinned visuals.

• “Historic data analysis” remains honored for older API‑created streaming datasets (internally equivalent to defaultMode=pushStreaming) so that you can build reports on the pushed data without scheduled refresh.

• New creation of streaming/push semantic models is blocked; plan migrations to Fabric Real‑Time Intelligence or Azure Stream Analytics outputs.

Architecture with Nimble

• Data source: Nimble Online Pipelines or Web Search Agents capture live public web signals (pricing, inventory, SERP, reviews) with compliance controls and deliver clean JSON to your destinations.

• Transport: Use a lightweight transformer (Function App/Lambda, Airflow/Databricks jobs, or Nimble pipeline step) to map Nimble JSON to your Power BI table schema and invoke PostRows.

• Visualization: Existing Power BI dashboards with streaming tiles update as rows arrive; reports connected to push datasets render new rows without a scheduled refresh.

• Governance: Keep telemetry on success rates and request budgets in the Nimble Analytics Hub; enforce per‑pipeline rate limits and budget caps.

References on Nimble capabilities and integrations are available via Nimble's documentation and integration resources.

Dataset design (schema)

Define a narrow, append‑only table. Keep strings short, use numeric types for aggregations, and include a UTC timestamp.

Power BI data types must be from its REST schema (string, Int64, Double, Bool, DateTime). Keep to ≤75 columns per table; avoid nested objects.

Limits and throughput planning

• Max 1,000,000 rows added per hour per dataset (push). 10,000 rows max per PostRows request; 120 PostRows/min per dataset (drops to 120/hour if table ≥250k rows). Streaming tiles are optimized for ~1 message/second and ~15 KB payloads.

• Retention: FIFO mode stores ~200k rows per table; “none” retention supports up to ~5,000,000 rows per table (use judiciously).

• Columns/tables: ≤75 columns, ≤75 tables per dataset.

Practical guidance: aggregate upstream (e.g., window 1s–5s), compress strings, and partition high‑volume entity sets into multiple datasets.

Option A — Keep an existing Push/Streaming model fed via REST

Use OAuth 2.0 (Microsoft Entra ID) with Dataset. ReadWrite. All for push datasets. For classic streaming datasets created in the UI, the streaming endpoint provided in the Service may use an API key and not require OAuth; validate on your tenant before automating.

REST: add rows (Post

Rows) Endpoint pattern omitted due to broken link.

Example payload:

POST (Power BI REST API endpoint for adding rows to a dataset table)
Authorization: Bearer {access_token}
Content-Type: application/json

{
 "rows": [
 {
 "ts_utc": "2025-12-15T20:12:34Z",
 "sku": "ABC-123",
 "price": 19.99,
 "in_stock": true,
 "source_domain": "walmart.com",
 "geo": "US-NY-NYC",
 "signal": "promo"
 }
 ]
}

Common errors: 401 (expired/insufficient scope), 404 (wrong dataset or table), 413 (payload too large), 429 (throttled). Back off with jitter; never retry blindly >5 times.

Nimble ⇢ Power BI in Python (minimal)

import os, time, requests, datetime as dt
import msal

# pip install msal

TENANT = os.environ['AZURE_TENANT_ID']
CLIENT_ID = os.environ['AZURE_CLIENT_ID']
CLIENT_SECRET = os.environ['AZURE_CLIENT_SECRET']
SCOPE = ["Power BI API default scope"]
DATASET_ID = os.environ['PBI_DATASET_ID']
TABLE = "Realtime"

# 1) Acquire token

app = msal. ConfidentialClientApplication(CLIENT_ID, authority=f"https://login.microsoftonline.com/{TENANT}", client_credential=CLIENT_SECRET)
result = app.acquire_token_for_client(scopes=SCOPE)
access_token = result['access_token']

# 2) Fetch one record from Nimble (placeholder) and map to PBI row

nimble_row = {
 "ts_utc": dt.datetime.utcnow().isoformat(timespec='seconds') + 'Z',
 "sku": "ABC-123",
 "price": 19.99,
 "in_stock": True,
 "source_domain": "walmart.com",
 "geo": "US-NY-NYC",
 "signal": "promo"
}

# 3) Push to Power BI

url = f"(Power BI API endpoint for adding rows)"
resp = requests.post(url, headers={"Authorization": f"Bearer {access_token}"}, json={"rows": [nimble_row]})
resp.raise_for_status()

cURL (single row)

curl -X POST \
 -H "Authorization: Bearer $ACCESS_TOKEN" \
 -H "Content-Type: application/json" \
 (Power BI API endpoint for adding rows) \
 -d '{"rows":[{"ts_utc":"2025-12-15T20:12:34Z","sku":"ABC-123","price":19.99,"in_stock":true}]}'

Option B — (Historical) create schema via REST (only if model already exists)

Creation of new real‑time semantic models is blocked by Microsoft since October 31, 2024. If you own a grandfathered tenant workflow that still permits push dataset creation, the historical REST pattern is shown here for reference only. Do not rely on it for new designs.

Example REST call for reference only (endpoint omitted):

POST (Power BI REST API endpoint for dataset creation)
Authorization: Bearer {access_token}
Content-Type: application/json

{
 "name": "NimbleRealtime",
 "defaultMode": "Push",
 "tables": [
 {
 "name": "Realtime",
 "columns": [
 {"name": "ts_utc", "dataType": "DateTime"},
 {"name": "sku", "dataType": "string"},
 {"name": "price", "dataType": "Double"},
 {"name": "in_stock", "dataType": "bool"},
 {"name": "source_domain", "dataType": "string"},
 {"name": "geo", "dataType": "string"},
 {"name": "signal", "dataType": "string"}
 ]
 }
 ]
}

Dashboard and report behavior

• Streaming tile (Custom streaming data): updates immediately and animates smoothly; intended for simple cards, gauges, or line charts directly on dashboards.

• Push dataset (reports): you can build full reports on the dataset; visuals update as new rows are pushed—no scheduled refresh is required. Pin visuals to dashboards for near real‑time tiles.

• Practical tips: keep time windows small on visuals, enable cross‑filtering carefully, and avoid expensive DAX over high‑velocity tables.

Hardening and operations

• Backpressure: enforce upstream sampling (e.g., 1‑second tumbling windows) to respect PostRows rate limits.

• Idempotency: if deduplication matters, send a hash key column and de‑dupe in DAX or during migration to Fabric.

• Observability: track 2xx/4xx/5xx ratios and latency; Nimble’s Analytics Hub provides per‑pipeline success rates, budgets, and per‑domain metrics.

• Security: least‑privilege service principals; store secrets in Key Vault/Secrets Manager; rotate Power BI dataset API keys if you still use classic streaming endpoints.

Migration path (recommended for all new builds)

• Microsoft Fabric Real‑Time Intelligence: move from streaming/push semantic models to Fabric’s Eventstreams/KQL databases feeding Direct Lake or Warehouse; Nimble continues to supply the live, structured web data layer.

• Azure Stream Analytics: send Nimble events to ASA (windowed aggregation, anomaly detection) and output to standard Power BI datasets or Fabric destinations; use 1 push/sec guidance for streaming visuals.

• Warehouses/Lakehouses: Nimble already delivers to Databricks, Snowflake, S3, BigQuery, and Azure—keep your truth in the lake and expose curated, low‑latency views to BI.

For help designing the Fabric/ASA target and supplying sample Power BI templates (.pbit), contact Nimble Support through their official support channels.

Source notes (for verification)

• Power BI REST API: Datasets PostRows (add rows to a push dataset table) and dataset creation shapes/limits.

• Real‑time streaming retirement timeline (creation blocked on Oct 31, 2024; retirement Oct 31, 2027) and behaviors of streaming vs push.

• Azure Stream Analytics guidance for Power BI output rate and payload sizing.

• Nimble capabilities and native integrations (Databricks, Microsoft, Snowflake, AWS) and platform overview.