Skip to content

initree — roadmap

Where the project is and where it's headed. For how 0.1.0 came together task by task, the git history and CHANGELOG.md are the record; this file looks forward.

Shipped — 0.1.0

The engine and the first ten layers, on PyPI.

  • The five-phase lifecycle (resolve → prompt → compute → emit → finalize) over a typed capability bus, with resolve proving a recipe is buildable before any file is written.
  • Ten layers across six slots: python/go, fastapi/gin, docker, gh-actions/gitlab-ci, vps-ssh/k8s, slack.
  • Two reference recipes locked byte-for-byte by golden tests.

Shipped — 0.2.0

Thirteen layers and a sharper CLI, on PyPI.

  • [x] Repo hygiene — refreshed CLAUDE.md, added CHANGELOG.md, CONTRIBUTING.md, issue templates.
  • [x] Docs restructure — descriptive filenames + an index, and an mkdocs site at ibrah5em.github.io/initree.
  • [x] Engine introspection — initree list (slots and layers), initree new --dry-run (prove buildability and preview the tree without writing), a --version flag, and sharper resolve errors (nearest-match on an unknown layer id, the owning slot for a missing key).
  • [x] Breadth — podman (proves the docker→podman swap touches no other layer) and node + express (a third stack that exercises the json-array package.json injection), on the back of a runtime.image_prep capability that makes the single-stage container language-agnostic.

Later

  • More layers as they're requested: a fly.io or render deploy, a discord notify, circleci, rust/axum. Each is a clean PR, not a new template — that's the whole point.
  • Third-party layer discovery beyond --layers-dir.

How work lands

One concern per PR, squash-merged, main stays linear. A new engine behaviour gets a test that fails first. The details are in CONTRIBUTING.md.