Query external Parquet data

Lakehouse support registers Parquet sources and maps their columns to graph concepts. Source files remain external; ShrikeDB stores connection metadata, mappings, and query artifacts.

Use this workflow when external files remain the authoritative data. For application records that will be updated through ShrikeDB, use direct ingestion instead.

Before you begin

Lakehouse registration and execution require SHRIKEDB_LAKEHOUSE_ENABLED=1 on the server. You need access to a selected database, permission to manage its source definitions, and Parquet files accessible to the server. Start with local Parquet files and a mapping whose results you can check against the input.

  1. Register, review, and query explains the MCP requests and mapping lifecycle.

  2. Understand source selection explains result diagnostics.

  3. Tune lakehouse queries explains useful measurements.

  4. Configuration and source management covers feature settings, credentials, and refreshes.

Result types

Connection and mapping tools return workflow reports. Query responses include rows plus source-selection diagnostics. Some direct inspection paths can return diagnostic fixture rows; check pushdown_status before using a result as source data. The workflow chapter describes this distinction and verification steps.