ORIONPULSE

Six modules that take a question from raw table to measured decision.

Pulse Connect, Semantic Layer, Ask Orion, Forecast Studio, Boards and Decision Flows share one metric catalog and one permission model, so an answer in chat matches the board and the API.

Pulse Connect

Live connections push queries down to the warehouse. Business-app and file connectors land into a schema you own, on a schedule from 15 minutes to daily.

Warehouses
Snowflake, Databricks SQL, Google BigQuery, Amazon Redshift, Microsoft SQL Server, PostgreSQL
Streams
Apache Kafka, Amazon Kinesis (for sub-minute metric refresh)
Files
Parquet, CSV and Excel from Amazon S3, Google Cloud Storage, Azure Blob
Business apps
Salesforce, SAP S/4HANA, NetSuite, Shopify, Workday

metrics/net_revenue.yml

metric: net_revenue
label: Net revenue
owner: finance-analytics
sql: SUM(order_lines.gross_amount) - SUM(order_lines.discounts) - SUM(returns.amount)
grain: [day, week, month]
dimensions: [region, store_id, channel, category]
row_level: region IN user.allowed_regions
verified: true

Semantic Layer

Metrics, joins, dimensions and row-level rules are defined in version-controlled YAML. Sync it from GitHub or GitLab, review changes in pull requests, and mark metrics as verified. Ask Orion and every board read only from this catalog.

Ask Orion

An LLM analyst that turns plain-English questions into SQL against your semantic layer, runs it under the asking user's permissions, and explains the result with a chart. Analysts can edit the SQL, verify an answer, and pin it to a board.

Supported models

Each workspace picks a default and a fallback. Enterprise plans can bring their own key or a private endpoint.

  • AnthropicClaude Sonnet 5.5, Claude Haiku 5.5
  • OpenAIgpt-6.1-sol, gpt-6-luna
  • GoogleGemini 3.8 Flash
  • Private endpointsAmazon Bedrock, Azure
Illustrative product interface

Forecast Studio

Train demand, cash-flow and failure-risk forecasts without leaving the platform. Forecast Studio compares gradient-boosted trees, Prophet-style seasonal models and ARIMA on a holdout window, picks the best by MAPE or RMSE, and retrains on a schedule. Every forecast shows its features, backtest and drift alerts.

Illustrative product interface

Boards, built without code

Drag metric tiles, trend lines, tables, maps and forecast panels onto a grid. Filters, drill-downs and scheduled PDF or Slack snapshots are set from menus. Analysts who prefer code can open any tile in SQL or Python notebooks.

  • 14 tile types, cross-filtering and drill to row level
  • Refresh on warehouse sync or every 1 to 60 minutes
  • Embed any board with the Embed SDK and signed tokens

Decision Flows

Rules and model outputs open decision cards with an owner and evidence, notify Slack, Microsoft Teams, Jira or ServiceNow, and record the result so ROI is measured, not estimated.

Flow: stockout risk · Northeast

  1. Forecast Studio Stockout risk crossed 20% for 12 SKUs 07:02
  2. Decision Flows Decision card opened, assigned to Supply planning 07:02
  3. Planner Approved expedited transfer from DC-04 09:41
  4. Decision Flows Outcome logged: 11 of 12 SKUs stayed in stock Oct 25

Try every module on your own warehouse for 14 days.