Jayrun¶
Jayrun is an artifact-centric execution framework for computational graphs that need iteration, shared resources, hardware placement, supervision, and reliable lifecycle management.
Artifacts make data flow explicit. Operators transform artifacts or terminate a flow with an external side effect, while Jayrun coordinates concurrent execution, automatic clearing and retention, reusable resources, placement capacity, failure containment, and shutdown.
Start here¶
Follow the documentation in this order:
Introduction — decide whether Jayrun fits your workload.
Getting Started — build and execute one complete graph.
Scope and Lifetime Model — understand declaration, runtime, context, and execution ownership.
Data Model — distinguish declarations, graph-local definitions, IDs, and runtime values.
Data — understand the common runtime value container.
Artifacts and Data Flow — declare flowing data and manage artifact lifetime.
Configuration — provide graph-scoped computational values.
Resources — declare and share runtime-managed data and capabilities.
Operators and Executions — define reusable transformations and terminal side-effect operators, then understand invocation behavior.
Graph Construction — combine flows, derive layout, express routing patterns, bind resources, inspect declarations, and understand compilation.
Graph Validation — validate artifact contracts and interpret text, programmatic, and plotted diagnostics.
Operational Interfaces — use execution, context, runtime, and capacity capabilities safely from components.
Placement and Capacity — understand device leases, contention, and admission.
Engine and Context Lifecycle — submit, wait for, control, and inspect context runs safely.
Execution Settings — distinguish engine-wide policy from per-context policy and understand override rules.
Failure and Reliability Model — understand failure containment, escalation, rollback, and cleanup guarantees.
Observability and Inspection — interpret context runs, reports, diagnostics, and external telemetry boundaries.
After the conceptual path, use the task-oriented tutorials:
Use the API Reference for object-level lookup and Troubleshooting for operational diagnostics.
Installation¶
python -m pip install jayrun
Jayrun requires Python 3.11 or later.