Overview

Meteor runs metadata jobs described by recipes. A recipe wires together three kinds of plugins:

Recipe (YAML) → Source → Processors → Sinks

• A recipe is a YAML file that defines one job: what to extract, how to transform it, and where to send it.
• A source names the extractor plugin that pulls metadata out of a system, such as BigQuery or Kafka.
• Processors transform each record in flight. They are optional.
• Sinks receive the final records. A recipe can have many sinks.

Each extractor emits records. A record holds an entity (the resource and its metadata) and a list of edges (its relationships, such as ownership and lineage). The metadata model page describes both in detail.

To understand how Meteor executes a recipe — concurrency, batching, retries, and error handling — see Architecture. For the bigger picture of what this metadata enables, see Context Graph for AI.