v0.0.7

What's Changed

Added

• EngineConfig — Frozen dataclass for engine-managed mounts. Accepts url (simple) or engine_factory (advanced, e.g. custom pool/connect_args via create_async_engine_factory). Supports schema, create_tables, and custom model overrides. Engine is created at mount time and disposed on unmount/close.
• SessionConfig — Frozen dataclass for app-managed mounts. Wraps an existing session factory; Grover does not dispose the engine. Dialect inferred from the factory's bind or set explicitly.
• create_async_engine_factory() — Helper that captures create_async_engine args and returns a zero-arg callable for deferred engine creation.
• Engine lifecycle on Mount — Mount.engine field tracks Grover-owned engines. unmount() and close() dispose them automatically.

Changed

• add_mount() API redesigned — engine=, session_factory=, dialect=, file_model=, file_version_model=, file_chunk_model=, db_schema= parameters replaced by engine_config= and session_config=. filesystem and mount are now keyword-only. path remains positional.
• DatabaseFileSystem constructor simplified — Config parameters (dialect, schema, model classes) removed. Only provider kwargs remain. New _configure() method applies settings from EngineConfig/SessionConfig at mount time.
• LocalFileSystem constructor simplified — Model and schema parameters removed. Accepts only workspace_dir, data_dir, and provider kwargs.
• UserScopedFileSystem constructor simplified — Config parameters removed. Accepts only share_model and provider kwargs.
• deepagents integration — from_database() and from_database_async() now accept EngineConfig instead of raw engine/session_factory.

Fixed

• Sync Grover in Jupyter — Pre-created AsyncEngine bound to the wrong event loop no longer causes failures. EngineConfig defers engine creation to mount time, ensuring it runs on the correct loop.

Full Changelog: v0.0.6...v0.0.7