Skip to main content
Initial Run is the first modeling journey. It is available only in Staging.

​
What Initial Run does

  1. Reads selected ClickHouse source schemas and tables.
  2. Profiles columns and source structure.
  3. Creates dbt source definitions.
  4. Generates staging models.
  5. Generates warehouse dimensions and facts.
  6. Generates marts for downstream consumption.
  7. Validates dbt in an isolated workspace.
  8. Creates a proposal only after validation passes.
Initial Run does not publish directly to production. It creates a proposal that can be reviewed, applied to Staging, run, and promoted.

​
Before running

  • At least one source is loaded or externally available.
  • Tables are selected in Operations.
  • GitHub is configured.
  • Assistant/OpenAI is configured.
  • The active environment is Staging.

​
Cost preview

Initial Run generates models with the customer’s OpenAI key, so Platinur shows a scope and cost estimate before anything is spent. When the operator starts Initial Run, the portal first counts the tables and columns in the selected source schemas. Sources in selected-tables mode count only the selected tables; sources in all-tables mode count the whole schema. The confirmation dialog then shows:
  • How many tables across how many sources will be profiled.
  • The configured OpenAI model.
  • The estimated total token usage, with an estimated dollar cost when a cost rate (ai.estimated_cost_per_million_tokens) is configured.
  • The token budget at which the run stops automatically.

​
What the estimate is based on

The estimate is derived from the live table and column counts plus a fixed planning overhead, multiplied by a repair-loop factor. It is intentionally pessimistic: a rough order of magnitude, not a quote. The estimate is advisory — if the scope cannot be read, the portal falls back to a generic confirmation instead of blocking the run.

​
Confirm and cancel

  • Confirm starts Initial Run. From this point tokens are spent on the configured OpenAI key.
  • Cancel closes the dialog. Nothing runs and nothing is spent.

​
Token budget

The run stops automatically once total token usage exceeds the configured budget (Run Token Budget in Configuration, default 1,500,000). A stopped run can be resumed after raising the budget; files already generated are restored from the sandbox, so completed work is not repeated.