Use Dagster ops and jobs to orchestrate a multi-step legacy pipeline with shared resources and run-scoped configuration

Steps

1. Define ops using the @op decorator; each op declares its input and output types and receives a context object giving access to resources and config
2. Declare required resources in the op using the required_resource_keys parameter; resources are initialized once per run and shared across all ops in that run
3. Define a job using @job and wire ops together inside a with-block using op return values to establish dependencies, or use the graph() abstraction for reusable subgraphs
4. Supply run configuration as a Python dict or YAML under the ops key, scoping config to individual ops by name; use ConfigSchema to validate config at submission time
5. Provide resource implementations at the job or repository level using resource_defs, enabling different implementations (mock vs real) for test and production environments
6. Launch the job from the Dagster UI, the CLI (dagster job execute), or via the GraphQL API, passing run config as a JSON payload