Skip to content

Public Roadmap

Note

This roadmap is directional, not a promise. Dates are targets and may shift.

v1.0.x Hardening (Maintenance)

Focus: correctness and reliability fixes for the existing retry engine.

  • Use monotonic time for deadlines.
  • Do not retry cancellation/system-exiting exceptions.
  • Tighten HTTP status coercion in http_classifier.
  • Clarify classifier heuristics and safety guidance in docs.

v1.1 (Q1 2026) Policy, Breakers, and Practical Ergonomics

Focus: a coherent execution model plus integration ergonomics that avoid bespoke glue.

  • Introduce unified Policy + Retry components (sync + async), keeping RetryPolicy as convenient sugar.
  • Add a first-class CircuitBreaker with state transitions and events.
  • Result-based retries (result_classifier) for HTTP/SDK responses without forcing exceptions.
  • Classification context and strategy integration (classifiers can emit hints; strategies can consume them).
  • Retry-After support as a built-in strategy helper.
  • Cleaner outcome surface: stable stop reasons and typed terminal outcomes (attempts/last class).
  • Timeline capture (optional) for per-attempt debugging without custom hooks.
  • Observability improvements and an OpenTelemetry contrib hook.
  • Docs refresh with end-to-end recipes (HTTP calls, worker loops, graceful shutdown).

v1.2 (Q1 2026) Built-in Classifiers, Recipes, and Advanced Patterns

Focus: reduce bespoke wiring for common stacks, plus production guardrails.

  • Built-in classifiers for common libraries (aiohttp, grpc, boto3/botocore, redis, etc.) via extras.
  • Classifier authoring guidance and integration recipes.
  • Additional small strategy helpers driven by real-world patterns.
  • Retry budgets to prevent retry storms.
  • Testing utilities for deterministic retries.
  • Per-attempt timeouts in addition to overall deadlines.
  • Injectable sleeper / before-sleep hook to integrate with leases and external schedulers.

v1.3+ (Q2 2026+) Ecosystem Expansion

Focus: optional integrations and higher-level execution models.

  • Framework integrations (Django, Flask, Celery, FastAPI).
  • Contrib observability modules (Prometheus, Datadog, Sentry).
  • Non-blocking / externally scheduled retry execution (advanced).
  • Experimental hedging support (async-first).

Ongoing Documentation

  • Migration guides (Tenacity, Backoff).
  • Performance tuning and troubleshooting guides.