Skip to content

Guarantees and limitations

Atomic Flow is guided by a simple principle: never claim a stronger guarantee than the implementation can actually provide.

For every feature and implementation, the project aims to document:

  • Guarantee: what behavior is expected and how it is measured.
  • Assumptions: required properties of storage, external services, and user-supplied operations.
  • Failure modes: cases where execution or recovery can still fail.
  • Observability: how the caller can distinguish a successful outcome from an uncertain one.
  • Verification: the tests or evidence supporting the claim.

Atomic Flow does not aim to:

  • Replace native database transactions.
  • Guarantee global ACID transactions over independent services.
  • Reverse arbitrary external side effects.
  • Automatically repair every partial failure without application involvement.
  • Make an unverified recovery outcome look successful.

No stable Atomic Flow runtime guarantees are published yet. The current documentation is architectural direction. Supported guarantees will be stated precisely after behavior is implemented and tested.

For the implementation roadmap, see Development status.