Skip to content

What is Baldur?

A self-healing reliability layer for Python applications — circuit breaker, retry, fallback, and dead-letter queue behind one decorator.

Every application depends on things that fail: a payment gateway times out, a database gets slow, a third-party API starts returning errors. Baldur wraps the calls you care about in a composed resilience pipeline, so those failures are contained, retried, and recovered from automatically — instead of becoming customer-visible errors and after-hours pages.

import baldur


@baldur.protected("charge-customer")
def charge(order_id: str) -> dict:
    return payment_gateway.charge(order_id)

That decorator is the whole integration. With zero configuration it runs on an in-memory backend — no Redis, no Docker, no environment variables.

What it does when a call fails

  • Stops hammering a dependency that is down. A circuit breaker trips after repeated failures and fails fast until recovery probes confirm the dependency is healthy again.
  • Retries what deserves retrying. Transient errors are retried with backoff, so a failure that would succeed on the next attempt never reaches your users.
  • Falls back to a safe answer. When the call cannot succeed, your fallback runs instead of an exception propagating.
  • Sets work aside instead of dropping it (PRO). Failed work lands in a durable dead-letter queue and can be replayed — from code or from the web console — once the dependency recovers.

The concept behind all four, and why hand-rolling them in every project goes wrong, is covered in What is self-healing?.

Where it sits in your stack

Baldur is a library inside your process — pip install, decorate, done. It is not a proxy, not a sidecar, and not a hosted service; the framework runs entirely inside your own application, and nothing about your deployment changes.

  • Framework-agnostic core with adapters for Django, FastAPI, Flask, and Celery.
  • Zero-config by default: in-memory state for a single process; point it at Redis when you scale to multiple workers.
  • Operable from the browser: a built-in web console shows what is failing and lets you recover — no separate deployment.
  • Observable: Prometheus metrics and OpenTelemetry traces out of the box.

Already running a service mesh? Baldur complements it rather than competing with it — see Baldur and your service mesh.

OSS and PRO

The core is free and Apache-2.0 licensed (pip install baldur-framework), and covers the resilience patterns themselves: circuit breaker, retry, fallback, idempotency, health checks, metrics, and the web console. Baldur PRO is a separate package you add on top for critical workloads: a durable dead-letter queue with replay, audit trail, emergency mode, governance gates, unified notifications, and more. PRO adds capability — it never replaces or relicenses anything in the core.

See the tier model for how to decide which tier you need, the capability matrix for the feature-by-feature map, and Pricing for plans.

Where to go next

  • Getting Started — a protected endpoint in your framework, starting from zero infrastructure.
  • What is self-healing? — the concept behind the library.
  • Web Console — operate and recover from the browser.
  • Pricing — the OSS core is free forever; PRO plans start at $149/month.