sim_platform.market.connectors

The interface for connecting a data vendor without bypassing look-ahead protection.

A connector reads from a vendor (Bloomberg, Refinitiv, FactSet, a database) and yields records that carry their own publication time. export_connector writes them in the formats the platform already loads, so every experiment reads vendor data through the same point-in-time stores, tools and disguise as files: a connector never hands data to an Agent directly. Market Lab ships no paid-data connector; write one with this Protocol.

class sim_platform.market.connectors.DataConnector(*args, **kwargs)[source]

Bases: Protocol

Yields point-in-time records for one symbol; each method may yield nothing.

fundamentals(symbol)[source]

Rows with symbol, metric, period_end, value, available_date, optional unit/form.

available_date must be when the figure became public, never the period end.

Parameters:

symbol (str)

Return type:

Iterable[Dict[str, Any]]

news(symbol)[source]

Rows with published_at (ISO timestamp), headline and optional id/period.

Parameters:

symbol (str)

Return type:

Iterable[Dict[str, Any]]

documents(symbol)[source]

Rows with published_at, title, text and optional id/form.

Parameters:

symbol (str)

Return type:

Iterable[Dict[str, Any]]

sim_platform.market.connectors.export_connector(connector, symbol, folder, session_start, *, delay_seconds=0, provenance=None)[source]

Write a connector’s records as fundamentals CSV, news JSON and documents JSON.

Returns the paths written (None for a kind with no records); pass them as fundamentals_path, the experiment’s news file and documents_path.

Parameters:
  • connector (DataConnector)

  • symbol (str)

  • folder (str | PathLike[str])

  • session_start (str)

  • delay_seconds (int)

  • provenance (str | None)

Return type:

Dict[str, str | None]