Roadmap

Flowli is alpha software. The engine, the patterns, the CLI and the HTTP service are implemented and tested. Now they need to prove themselves in real use. This page describes where the project is and what comes next. It states intent, not commitments: priorities shift with what we learn.

Current phase: testing and validation

The API is not frozen yet. Before 1.0, we want evidence that the model (a journaled coroutine execution stack, replayed from the top) is correct, sufficient, and pleasant to build on.

Finding bugs

  • Keep the test suite exhaustive: the in-memory backend for the rules, CairnDB on the filesystem backend for real contention, and mutation testing.

  • Hunt for the bugs that tests miss: long-running executions, large journals, many concurrent workers, and crashes at every point of the worker loop.

  • Report anything surprising. See Contributing.

Building applications on top of Flowli

The best test of an API is building real things with it. flowli-runner and flowli-codeflow, which run coding agents as workflows, are the first applications. They answer two questions:

  • Is the API useful? Do Context, the patterns and the operator commands express real workflows naturally?

  • Are primitives missing? Patterns that recur across applications are candidates for new patterns, or for new Context verbs.

The API may change as a result, with breaking changes recorded in the Changelog.

Next

  • The first release on PyPI.

  • Validation on the cloud storage backends of CairnDB, not only on the filesystem backend.

Later

  • Publishing flowli-runner and flowli-codeflow once their API settles.

  • A 1.0 release with a stable API, once the validation phase has settled the model.