Idempotency & Retry (Dispatch)

Simple Explanation

Run many agent calls safely: avoid duplicate work with idempotency keys, enforce timeouts and rate limits, and apply fail‑fast policies.

Technical Explanation

Use dispatch_agents() with DispatchResourceLimits for quotas and tracking. Idempotency cache stores results by key; repeated inputs return cached outputs. Rate limiter controls request pace; timeouts and failure‑rate guards enforce resilience.

When to use - Parallel orchestration with potential retries - Budget and reliability constraints

Key Primitives - dispatch_agents() — parallel execution with guardrails - DispatchResourceLimits — quotas and thresholds - Idempotency cache — avoids duplicates

Try it - Run the sample: samples/32-idempotency-retry/idempotency_retry_demo.py - See the README: samples/32-idempotency-retry/README.md

Why run this - See cache hits in action for duplicate inputs - Observe timeouts and failure‑rate guards under load - Tune rate limits and quotas to your API budgets

Gotchas - Bound retries; treat timeouts as failures when appropriate - Tune rate limits to API quotas - Use deterministic aggregation for partial results