> ## Documentation Index
> Fetch the complete documentation index at: https://offthepace.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Run SQL against the gold marts in your browser

> A full SQL editor backed by DuckDB-Wasm. Write raw queries against every registered gold mart no credentials, no server, results in milliseconds.

Query Lab is the escape hatch from the curated charts: a full SQL editor wired straight to the same DuckDB-Wasm engine that powers every other page. The registered gold-mart Parquet files load into your browser, and your query runs locally in WebAssembly no credentials, no backend, results in milliseconds. If a chart doesn't ask the question you have, Query Lab lets you ask it yourself.

## What it shows

Query Lab is a two-pane workspace rather than a chart:

* **Editor pane**: a SQL editor where you write DuckDB SQL against any registered table.
* **Results grid**: the rows your query returns, rendered as a sortable table.
* **Registered tables**: every gold mart published by the dbt pipeline is registered by name (`fct_lap_residuals`, `fct_driver_skill_features`, `fct_cliff_prediction_features`, `dim_compounds_season`, and the dimension tables), queryable exactly as they appear in the warehouse.

<Note>
  Query Lab is the one shipped feature that is **not** a FeaturePage module it is a bespoke editor page (`/query`) with no chart methodology drawer. It is documented here by hand rather than harvested from a `methodology.tsx`.
</Note>

## How to use it

1. **Open Query Lab** from the sidebar (or at `/query`).
2. **Wait for the engine indicator** in the top bar to turn green DuckDB-Wasm has finished loading the Parquet files.
3. **Write a query** against any registered table and run it; results render in the grid below.
4. **Iterate locally** every run executes in your browser, so there is no rate limit and nothing leaves the page.

## Reading the signal correctly

Query Lab queries the **published gold marts**, the same tables the charts read so a number you compute here will match the corresponding chart when you apply the same filters (clean-lap gates, anomaly exclusions, minimum samples). If a hand query disagrees with a chart, the difference is almost always an inclusion filter the chart applies that your query does not.

<Warning>
  The marts are modelling outputs, not raw timing data columns like `driver_skill_residual_s` already have car, tyre, fuel, and conditions removed. Read the relevant model's reference page before interpreting a column, so you know what has already been controlled for.
</Warning>

## Data source

Query Lab registers the full set of published gold marts via DuckDB-Wasm in the browser; there is no single backing model. The complete column-level catalog for every table lives in the **Reference** tab.

| Table                           | Role                                                         |
| ------------------------------- | ------------------------------------------------------------ |
| `fct_lap_residuals`             | Per-lap seven-term decomposition (the core mart)             |
| `fct_driver_skill_features`     | Driver skill residual aggregates                             |
| `fct_cliff_prediction_features` | Per-lap degradation / cliff feature matrix                   |
| `dim_*`                         | Driver, constructor, circuit, event, and compound dimensions |

The data window covers **2018–2024**.

For the full table and column catalog, browse the [model reference](/reference/models/fct/fct_lap_residuals) and the other entries in the Reference tab.