Control Flow Patterns

Simple Explanation

Common orchestration patterns to structure complex workflows: sequence, branch, map/reduce, retry with backoff, and circuit breakers. Use them to keep workflows predictable and resilient.

Technical Explanation

Compose tasks using explicit control flow: DAG or step functions. Implement branching by predicates, fan-out/fan-in for parallel work, and resilience via retries, backoff, and circuit breakers. Track idempotency keys to avoid duplicate work.

When to use - Multi-step workflows with conditional paths - Parallelizable subtasks with aggregation - External dependencies that can fail intermittently

Key Primitives - DAG planner or step executor - Branch predicates and guards - Fan-out/fan-in with concurrency limits - Retry/backoff; circuit breaker; idempotency cache

Try it - Run the sample: samples/15-control-flow/demo_patterns.py - See the README: samples/15-control-flow/README.md

Gotchas - Keep side-effects behind idempotent interfaces - Make aggregation robust to partial failures - Add timeouts and budget limits per branch