hemx/PLAN.md

# Active frontier — peak idiomatic examples

**Parent ID:** `examples-idiom/001`

- **User value:** A newcomer can move from first app to production reference and advanced integration while seeing one coherent hemx model: plain `.heml`, generated resources, typed Rust handlers/effects, real links/forms/URLs, server-owned truth, native fallback, and explicit leaf adapters.
- **State:** Ready — the canonical starter and Workout exemplar are already strong; bounded teaching leaks remain in the HTML gallery, client-local example, SaaS reference, and advanced Kanban boundary.
- **Non-goals:** no new framework primitives, client router, VDOM, selector authoring API, reactive expression language, global client store, CSS framework/design system, generalized asset pipeline, visual redesign, or feature expansion. Do not churn `examples/workout` or `examples/techdemo` without a concrete failing contract.
- **Build:**
  - **`examples-idiom/002` — Done: the SaaS reference tells the production truth.** Settings is a real `/settings` link enhanced by the existing page-swap runtime, direct `/settings` renders the fallback page, and `/events` now sends an initial canonical batch followed by ongoing server-owned status updates; `?once` remains only as a bounded production-reference probe. The removed handler no longer simulates navigation with response effects. Raw selector/form construction remains confined to test adapters because generated handles are the asserted boundary, not an app authoring API. req: canonical_authoring/001 req: nav/001 req: nav/002 req: nav/004 req: push/003 req: examples/003 req: examples/004 req: examples/014
  - **`examples-idiom/003` — Make the HTML pattern gallery mechanically copyable.** In `examples/html_examples/src/main.rs`, generated `gallery` resources, and nearby templates/tests, express the dependent-select flow through typed generated values/partials instead of duplicated string-to-option mapping, and remove request fields discarded only to satisfy example plumbing where the generated handler contract permits it. Preserve each visible htmx slug, native form semantics, server-owned state, inserted-content behavior, and no-reload smoke. req: htmx_equivalents/001 req: htmx_equivalents/003 req: htmx_equivalents/005 req: canonical_authoring/001 req: form/004
  - **`examples-idiom/004` — Make client-local code visibly a leaf adapter.** In `examples/client_local/src/lib.rs` and its `.heml` surface, project the generated client event/state into a tiny typed counter-domain input/output instead of rendering raw event kind and encoded state as the example's product value. Keep the ordinary `#[hemx::handler(client)]` shape, generated event/state boundary, native event semantics, and no durable client state graph. req: client_local/001 req: client_local/003 req: client_local/004 req: canonical_authoring/017
  - **`examples-idiom/005` — Separate the advanced Kanban adapter from ordinary hemx app code.** Move the cohesive sync/presence/session/storage transport responsibility from `examples/kanban/src/main.rs` behind one clearly named local integration module with a small route/state contract; keep board templates and ordinary handlers nearby and unchanged where possible. Update `examples/kanban/README.md` to label the fixture as the advanced local/offline/sync north-star, route beginners to the starter/Workout/gallery first, and name legacy `/sync-demo`/`sync.js` as a compatibility probe rather than recommended authoring. Preserve replay, export, deletion, reconnection, auth, and multiplayer browser proof. req: state/001 req: state/002 req: local/001 req: local/002 req: milestone/001 req: sync/001 req: sync/008
  - **`examples-idiom/006` — Publish and enforce the example ladder.** In `README.md`, example READMEs, and the nearest existing xtask/example checks, identify `app new`/`examples/v0` as the first canonical app, `html_examples` as the pattern gallery, Workout as the product exemplar, SaaS as the production integration reference, `client_local` as the narrow leaf-adapter proof, Techdemo as exhaustive verification, and Kanban as advanced north-star integration. Add the smallest repository-owned guard that fails when beginner/reference authoring regresses to raw IDs, selectors, raw wire/effect constructors, `$OUT_DIR` includes, or app-authored DOM mutation; keep legitimate advanced/test adapters scoped rather than banning tokens globally. req: examples/001 req: examples/003 req: examples/004 req: examples/005 req: examples/007 req: examples/008 req: examples/009 req: canonical_authoring/001 req: progressive_disclosure/001 req: progressive_disclosure/002 req: progressive_disclosure/003
- **Blocked by:** none. The separate v1 legal release gate remains blocked on the owner license decision but does not block example work.
- **Proof:** each slice must make its user path observable, preserve the named native/recovery path, and pass its focused package/browser proof. Parent closure requires `cargo run -p hemx-xtask -- html-examples-smoke`, `cargo run -p hemx-xtask -- workout test`, focused SaaS/client-local/Kanban package tests, `cargo run -p hemx-xtask -- test`, `cargo check --workspace`, `cargo fmt --check`, `redgate list`, `redgate refs`, and a clean diff. Frustration signals to reject: a beginner must author selectors/raw IDs/wire ops; navigation loses URL/history; JavaScript becomes durable truth; an example claims live/recovery behavior with a one-shot stub; or advanced sync machinery appears to be the default app model.

## Blocked release decision retained

- [ ] **State:** Needs decision — choose the repository distribution license and approved third-party SPDX set, then add root license file(s), workspace package metadata, `deny.toml`, and rerun the release matrix.
- **Blocked by:** owner/legal authority. Current third-party set includes Apache-2.0, Apache-2.0 WITH LLVM-exception, BSD-3-Clause, BSL-1.0, MIT, Unicode-3.0, and Unlicense; all 20 workspace packages currently lack license metadata.
- **Proof:** `cargo deny check advisories sources licenses` and the full `docs/v1-readiness.md` matrix pass, then readiness changes from NO-GO to GO without publication or deployment.