Writing
Practical lessons, architectural decisions, and engineering notes from building distributed systems and software that has to hold up in production.
Featured series
A practical series on designing external operations to remain explicit, observable, and recoverable when dependencies fail, responses arrive late, or the outcome is uncertain.
A practical guide to HTTPX retry policies in Python covering connection failures, 429/5xx responses, exponential backoff, Retry-After, timeouts, async cancellation and idempotency.
How to configure OpenAI timeouts in Python without hiding retries, exceeding operation budgets, or mishandling long reasoning and streaming responses.
RelPrim 0.9.0 adds provider-aware retry delays, maximum wait enforcement and fallback integration for external operations rejected by rate limits.
RelPrim 0.8.0 adds idempotency policies, concurrent execution joining and successful result replay for external operations that must not run twice.
Engineering Notes
Get new articles, architecture notes, and open-source updates by email.
