ogar-loco: hoist the vocabulary-agnostic call ABI out of ogar-blockly

[AdaWorldAPI]

What

The skeleton for the reusable low-code surface, per the operator direction: "Elixir should become just a rails-shaped semantic over classid index, 256:256. Not much different than blockly, just different vocabulary. It should be a reusable surface for any other purposes." Named next consumers: lance-graph's elixir-shaped compiled templates, then Power-Automate-style flows.

The surface splits in two — this PR builds the shared half and leaves the customizations to the wiring side:

• crates/ogar-loco (new, zero-dep) — everything that is the same no matter what the bytes mean:
  • FnIndex / Call / LaneShape / FunctionBody / call_in_slab + layout constants + budgets, moved verbatim from ogar-blockly (bytes, semantics, tests unchanged).
  • DOMAIN_FLOOR generalizes DEVICE_FAMILY_FLOOR: below the floor is the shared computational core, byte-stable across every vocabulary; at/above is the classid-selected vocabulary's own range.
  • vocabulary::shared_core — the core's stack_arity / body_refs / branches / min_shape tables defined once (transcribed from the proven blockly-rs tables: the two-quantity split + expression arities). Uncovered shared-core bytes (WAIT, STOP, RETURN, …) refuse everywhere.
  • trait Vocabulary + vocabulary::conformance::check — the seam a sibling codebook implements, and the mechanical no-drift gate every vocabulary crate must run. Both failure modes verified able to fire: a drifted ADD arity and a truncating min_shape are caught by name.
  • node / pool / program hoisted from blockly-abi, generalized where palette-typed (references_are_resolvable / branches_of are vocabulary-parameterized).
• ogar-blockly becomes the Blockly/Scratch vocabulary crate over the core: re-exports the old surface unchanged (the palette census test now doubles as the re-export completeness proof; blockly-rs compiles with zero changes), keeps BlockConcept / SoaSplit / BLOCKS_DOMAIN, and adds BlocklyVocabulary — domain hooks empty-and-refusing until the device families mint, because that range is reserved, not allocated.

What is deliberately NOT here (the wiring side)

Recorded in docs/BLOCK-EDITOR-PLAN.md § W6:

1. The template vocabulary crate — gated on lance-graph's rung-2 144-verb unification (O7: the two shipped 144-vocabularies diverge; an ordinal-pinned codebook forces that ruling) and its concept-domain mint.
2. The flow vocabulary crate (Power Automate) + the structured PARALLEL/JOIN / try-catch-shaped control mints its runAfter semantics need.
3. The blockly-rs flip off its local node/flow/program/pool copies — a scheduled obligation, not an aspiration; until it lands, ogar-loco's copies are the canonical ones for any second vocabulary.
4. Every concept mint (M1–M3 unchanged, plus the template/flow domains).

Gates

• cargo fmt --check, cargo clippy --all-targets --all-features -- -D warnings, cargo test --all-features (36 ogar-loco + 7 ogar-blockly), RUSTDOCFLAGS="-D warnings" cargo doc --no-deps, and the density example — all green, scoped -p ogar-loco -p ogar-blockly.
• Back-compat is proven from inside: ogar-blockly's census test compiles every named palette constant through the re-export path, exactly as downstream consumers do.

🤖 Generated with Claude Code

https://claude.ai/code/session_011DN5odWu4eisdevH9bPWyz

---

Generated by Claude Code

[cursor]

Bugbot couldn't run - usage limit reached

Bugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit.

A user or team admin can review and increase usage limits in the Cursor dashboard.

(requestId: serverGenReqId_b4cd93b9-4244-4789-b568-89038f511d9e)