WASM Runtime Integration

[flyingrobots]

Migrated from Method backlog

This issue was created from a legacy filesystem backlog card. GitHub Issues are now the live work tracker; repository docs remain Method evidence.

Source backlog: docs/method/backlog/up-next/PLATFORM_wasm-runtime.md
Original lane: up-next
Original legend: PLATFORM

Original backlog card

WASM Runtime Integration

Milestone: First Light | Priority: P1 | Repo: Echo

Historical note: This roadmap item predates the observation-first ABI v2
and the intent-shaped ABI v3 control-plane rewrite. It is retained as
planning archaeology, not as the current WASM API contract. The live public
boundary is documented in SPEC-0009 WASM ABI.

Historical First Light plan for wiring warp-wasm exports to real engine
logic. It reflects an earlier step(...)-driven browser boundary that has
since been replaced.

T-4-1-1: Wire Engine lifecycle behind wasm-bindgen exports

User Story: As a web developer, I want the WASM module to expose a real Engine instance so that I can drive tick execution from JavaScript.

Requirements:

• R1: Instantiate an Engine (via EngineBuilder) in a module-scoped RefCell or OnceCell inside warp-wasm, gated behind an init() export.
• R2: dispatch_intent(bytes) must call engine.ingest_intent() and return the IngestDisposition encoded as CBOR via echo-wasm-abi.
• R3: step(budget) must call engine.tick() up to budget times, returning a CBOR-encoded StepResult containing tick count and commit status.
• R4: get_head() must return the latest Snapshot fields (tick, state_root, commit_id) as CBOR bytes.

Acceptance Criteria:

• AC1: init() export constructs an Engine with at least one demo rewrite rule (motion rule).
• AC2: Round-trip test: JS calls dispatch_intent then step(1) and receives a non-empty StepResult.
• AC3: get_head() returns a valid CBOR payload whose state_root changes after a step that applies a rewrite.
• AC4: Calling step before init returns an error CBOR (not a panic).

Definition of Done:

• Code reviewed and merged
• Tests pass (CI green)
• Documentation updated (if applicable)

Scope: Engine instantiation, intent ingestion, tick execution, head query. All through existing wasm-bindgen exports.
Out of Scope: Snapshot-at (time travel), render_snapshot, query execution. Multi-warp Engine configurations.

Test Plan:

• Goldens: CBOR golden vectors for get_head() after genesis init (zero-tick snapshot).
• Failures: step before init returns structured error. Malformed intent bytes rejected gracefully (no WASM trap).
• Edges: step(0) is a no-op returning current head. step(u32::MAX) runs until no pending rewrites remain.
• Fuzz/Stress: Proptest with random intent payloads; engine must not panic.

Blocked By: none
Blocking: T-4-1-2, T-4-2-1

Est. Hours: 6h
Expected Complexity: ~250 LoC

---

T-4-1-2: Snapshot and ViewOp drain exports

User Story: As a web developer, I want to drain ViewOps and request snapshots at specific ticks so that I can render simulation state in the browser.

Requirements:

• R1: drain_view_ops() must return all materialized ViewOp frames accumulated since the last drain, CBOR-encoded.
• R2: snapshot_at(tick) must replay from the provenance store and return the full Snapshot at the requested tick as CBOR bytes.
• R3: render_snapshot(bytes) must accept a Snapshot and produce ViewOps suitable for visualization.

Acceptance Criteria:

• AC1: After step(1) that fires a motion rule, drain_view_ops() returns at least one ViewOp frame.
• AC2: Calling drain_view_ops() twice in succession returns an empty array on the second call.
• AC3: snapshot_at(0) returns the genesis snapshot matching the initial get_head().
• AC4: snapshot_at(tick) for a future tick returns an error payload.

Definition of Done:

• Code reviewed and merged
• Tests pass (CI green)
• Documentation updated (if applicable)

Scope: ViewOp drain, snapshot-at replay, render_snapshot projection.
Out of Scope: Streaming ViewOps over WebSocket. Incremental snapshot diffs.

Test Plan:

• Goldens: ViewOp CBOR output for a single-motion-rule tick golden scenario.
• Failures: snapshot_at for non-existent tick returns structured CBOR error. render_snapshot with garbage bytes returns error.
• Edges: drain_view_ops on fresh engine (no ticks) returns empty array. snapshot_at(current_tick) matches get_head.
• Fuzz/Stress: Rapid alternation of step and drain_view_ops with random budgets.

Blocked By: T-4-1-1
Blocking: T-4-2-2

Est. Hours: 5h
Expected Complexity: ~200 LoC

---

T-4-1-3: JS/WASM memory bridge and error protocol

User Story: As a web developer, I want a clean TypeScript API wrapper around the raw WASM exports so that I do not deal with raw Uint8Array encoding/decoding.

Requirements:

• R1: Publish a @echo/wasm-bridge TypeScript package (or web/ directory) that wraps each wasm-bindgen export with typed async functions.
• R2: CBOR decode all return payloads into typed TypeScript objects.
• R3: Errors from the WASM side (structured CBOR with { error: string, code: number }) are surfaced as typed EchoWasmError exceptions.
• R4: Memory cleanup: large Uint8Array returns are freed on the WASM linear memory side after JS copies them.

Acceptance Criteria:

• AC1: TypeScript wrapper compiles with tsc --strict --noEmit.
• AC2: bridge.init() / bridge.step(n) / bridge.getHead() type signatures match the CBOR schema.
• AC3: An intentional error (e.g., step before init) throws EchoWasmError with a .code property.
• AC4: No WASM linear memory growth beyond O(1) per bridge call (alloc/free pairs verified via instrumentation test).

Definition of Done:

• Code reviewed and merged
• Tests pass (CI green)
• Documentation updated (if applicable)

Scope: TypeScript wrapper, CBOR decode, error mapping, basic memory lifecycle.
Out of Scope: React/Svelte bindings. Worker thread offloading. Streaming APIs.

Test Plan:

• Goldens: Snapshot: bridge.getHead() output matches known JSON fixture after genesis init.
• Failures: Every WASM error code has a corresponding EchoWasmError subclass or code. Unrecognized CBOR returns a generic decode error.
• Edges: Empty intent bytes. Zero-length ViewOp drain. Concurrent calls (JS is single-threaded but test re-entrant safety).
• Fuzz/Stress: 1000 rapid step/drain cycles; monitor WASM memory (must not monotonically grow beyond blob cache).

Blocked By: T-4-1-1
Blocking: T-4-2-1, T-4-4-3

Est. Hours: 5h
Expected Complexity: ~300 LoC (TypeScript)

---

[coderabbitai]

🔗 Related PRs

#329 - Stack Witness 0001 fixture-backed jedit walking skeleton [merged]
#330 - test(wasm): own package export smoke boundary [merged]
#353 - Fix WASM application intent authority boundary [merged]
#373 - chore: remove legacy ttd-browser surface [merged]
#380 - feat(core): expose retained evidence posture on readings [merged]

---

📝 Issue Planner

_{Check the box below or use the @coderabbitai plan command to generate an implementation plan and prompts that you can use with your favorite coding assistant.}

• Create Plan

---

🧪 Issue enrichment is currently in open beta.

You can configure auto-planning by selecting labels in the issue_enrichment configuration.

To disable automatic issue enrichment, add the following to your .coderabbit.yaml:

issue_enrichment:
  auto_enrich:
    enabled: false

💬 Have feedback or questions? Drop into our discord!

[flyingrobots]

Referenced by the Continuum Launchpad Browser Replay roadmap as the Echo WASM runway for GP3. Release bar: flyingrobots-labs/flyingrobots.dev#4. GP3: flyingrobots-labs/flyingrobots.dev#9. Project: Continuum Stack Convergence