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

# BigQuery public time series

> Search public tables, preview an explicit series and forecast observed values.

## Find a table

In Quantura Search, enter **bigquery weather**, **bigquery google trends**, or an exact
identifier such as **bigquery noaa\_gsod.gsod2025**. You can also choose BigQuery in
Research → Economic data. Catalog discovery reads metadata; it does not run SQL.
Search returns a bounded set of tables, with schema-verified forecast availability.
Use an exact `dataset.table` identifier when a broad catalog search omits your table.

## Select and preview

Choose a date/time column, a numeric value, and exact filters that identify one
series. For NOAA GSOD, the supported `year`/`mo`/`da` combination can supply a UTC
date. For example, select `temp`, `stn = 725030`, and `wban = 14732` in `gsod2025`.
Preview the values, units, missing-value codes and coverage before forecasting.
The source's interval must match your intended series; choosing an interval alone
does not resample it. Select an aggregation explicitly to produce completed UTC
hourly, daily, Monday-weekly or monthly buckets. Missing intervals are not filled.

A table without supported date and numeric columns cannot be forecast. Duplicate
timestamps require additional exact filters or an explicit aggregation.

## Cost and access

History queries require sign-in. Credentials and billing stay on the backend.
Only `bigquery-public-data` tables are accepted; arbitrary SQL and private projects
are not accepted. Column identifiers are checked against the schema, and filter
values use query parameters.

Each uncached query is dry-run first, then limited to **100 MiB billed bytes**.
A row limit does not reduce scanned bytes. The service also reserves a daily
project budget of 10 GiB and allows 10 uncached queries per user per UTC day.
Repeated identical queries share a one-hour process cache. Catalog/schema metadata
is cached for 24 hours. A latest query uses the last completed UTC hour; explicit
historical cutoffs remain exact. Location comes from the dataset, including EU.

Public histories can be revised. A date cutoff excludes later observations but
cannot reconstruct the data available at the original publication time.

## API example

pay-as-you-go or Pro (including an active 14-day trial), enterprise and administrator accounts can use personal API keys.
Active 14-day trials include API keys and MCP access. Public-data query limits still apply.

```bash theme={null}
curl -H "Authorization: Bearer $QUANTURA_API_KEY" \
  'https://quantura.studio/api/v1/market-search?source=bigquery&q=noaa_gsod.gsod2025'

curl -X POST -H "Authorization: Bearer $QUANTURA_API_KEY" \
  -H 'Content-Type: application/json' \
  -d '{"source":{"type":"economic_series","provider":"bigquery","dataset_id":"noaa_gsod","table_id":"gsod2025","time_field":"__year_month_day","value_field":"temp","dimensions":{"stn":"725030","wban":"14732"},"frequency":"1D","aggregation":"none","limit":500}}' \
  https://quantura.studio/api/v1/market-data/history
```

Use the same `source` object with [ensemble forecasting](/docs/ensemble-api).
The worker receives the normalized immutable input snapshot, not Google credentials.

## Backend configuration

Enable BigQuery API on the backend billing project. The runtime service account
needs BigQuery Job User there and read access to the selected public tables.
Quantura reuses its existing server service-account credentials or Application
Default Credentials. `BIGQUERY_BILLING_PROJECT` can override the billing project;
no Google credential is exposed to the browser.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.