Skip to content
cccskills
FreePublished on GitHub

abi-superpower-wdbx-compute

WDBX compute superpower. Cycle-free accelerator contracts, deterministic CPU SIMD fallback, scoped Metal execution, output-checked CoreML inference, and honest CUDA/Vulkan boundaries.

The English summary is being prepared. Showing the original description.

donaldfilimon/abi19 stars(GitHub 19 + cccskills 0)

Run score Not tested

See how to install

Files included (1)

  • SKILL.md6.6 KB

SKILL.md (original)

Review the instructions the agent will receive before you install.

WDBX moved out of this repository on 2026-08-22. It now lives in the sibling repo ../wdbx together with abi-compute, abi-foundation, abi-core, and abi-telemetry; abi consumes them by relative path. Source paths below therefore read ../wdbx/crates/.... Run WDBX-only tests from that repo (./tools/cargo.sh test --locked --manifest-path ../wdbx/Cargo.toml --workspace < /dev/null, run from the abi checkout), and abi's gate (./tools/check.sh) from here.

Under the Abbey System Constitution (docs/superpowers/specs/2026-08-22-abbey-system-constitution.md) WDBX is the provenance-aware episodic substrate, not a vector store. Most of the evidence half of its specification is unimplemented; the measured gap list is in docs/superpowers/specs/2026-08-22-wdbx-conformance-gap-analysis.md. Do not describe an episodic capability as Current on the strength of the vector-store features that do exist.

ABI Superpower: WDBX Compute

Exposes the cycle-free abi-compute accelerator contract and WDBX selector. Unsupported or unverified paths degrade to deterministic CPU SIMD. Supported macOS builds can execute Metal dot/cosine/norm/batch-cosine; capability becomes runtime-verified only after successful CPU-oracle parity. CoreML evidence is an output-checked tiny-model inference under a .cpuAndNeuralEngine request, not placement or residency proof. CUDA/Vulkan runtime remains unverified.

Actions

info

Show compute backend matrix (availability + dispatch status):

abi wdbx compute info

Output includes:

  • CPU: scalar, AVX2, AVX-512, NEON through Rust std::simd
  • GPU: Metal/CUDA/Vulkan five-state evidence; compilation is not execution
  • NPU: ANE hardware presence plus output-checked CoreML inference, not residency
  • TPU: Report-only endpoint metadata (ABI_REMOTE_COMPUTE_ENDPOINT) plus a separately tested reference TCP transport

select

Backend selection is internal to the Rust compute API. There is no public select subcommand; abi wdbx compute info reports the requested/effective selection and its fallback reason.

backend

There is no public per-backend subcommand. Use abi wdbx compute info for the compat table, the per-backend evidence ladder, and WDBX selection; abi backends reports the native-kernel state and points there for the table (since 2026-09-16).

remote

ABI_REMOTE_COMPUTE_ENDPOINT opts abi wdbx compute info into one authenticated, bounded DOT probe. The reference transport attempts the configured endpoint and falls back to deterministic local CPU when the endpoint is unavailable. This is execution evidence for the small reference request, not a production remote-TPU service or a general WDBX offload path.

Backend Matrix (Honest)

BackendAvailableDispatchesReality
CPU SIMDAlways✅ AlwaysRust std::simd with host-matched width
MetalSupported macOS buildsScopedDot/cosine/norm/batch-cosine; runtime-verified only after CPU-oracle agreement, otherwise CPU fallback
CUDAAdapter may compile❌ UnverifiedCompilation/tool detection does not prove runtime initialization or execution
VulkanAdapter may compile❌ UnverifiedCompilation/tool detection does not prove loader/device execution
CoreML / ANEOptional inference helper❌ Residency unverifiedExecutes an output-checked tiny model under .cpuAndNeuralEngine; does not prove inference placement or ANE residency
Remote acceleratorReference DOT onlyScoped proberemote_compute.rs has an authenticated, timeout-bounded transport; the explicit environment endpoint is probed by compute info with deterministic CPU fallback

Implementation

ComponentSourceRole
Contracts / CPU SIMD../wdbx/crates/abi-compute/src/Object-safe accelerator, five-state evidence, deterministic SIMD/top-k
WDBX batch search../wdbx/crates/abi-wdbx/src/v2/index.rsAccelerator injection with result validation and CPU parity/fallback
GPU adapterscrates/abi-gpu/src/{adapters,metal_kernels}.rsMetal execution/oracle; CUDA/Vulkan unverified; CoreML output-checked without residency claim
NPU Detection../wdbx/crates/abi-compute/src/backend.rsane_hardware_present() — hardware metadata only
Remote accelerator../wdbx/crates/abi-wdbx/src/remote_compute.rsAuthenticated bounded reference DOT transport; compute info probes an explicitly configured endpoint and otherwise does no network I/O
Selection parity tests../wdbx/crates/abi-compute/src/cpu.rs, ../wdbx/crates/abi-wdbx/src/v2/index.rsAccelerator requests preserve CPU-reference results

CLI Access

abi wdbx compute info

Build and runtime boundary

abi-compute, abi-wdbx, and abi-gpu are normal Rust workspace crates. The WDBX selector retains deterministic CPU fallback; no compiled/available adapter is promoted to execution without later evidence states. The real CLI surface is abi wdbx compute info.

Claim Boundary

Per docs/spec/wdbx-north-star.mdx §3.3 and docs/contracts/external-claims-audit.mdx:

  • ✅ Cycle-free object-safe accelerator contract and five-state evidence
  • ✅ CPU SIMD parity across accelerator requests and fallback
  • ✅ Scoped Metal numerical execution when initialized and oracle-verified
  • ✅ ANE hardware detection and output-checked CoreML tiny-model inference
  • ✅ Reference loopback TCP DOT transport with local fallback
  • ⚠️ Metal execution remains supported-host/runtime scoped; availability alone is not execution evidence
  • ❌ CUDA/Vulkan runtime execution is not verified
  • ❌ CoreML inference under a compute-unit request is not placement or ANE residency
  • ⚠️ ABI_REMOTE_COMPUTE_ENDPOINT drives only a bounded reference DOT probe with deterministic CPU fallback; it is not general WDBX offload or production remote acceleration
  • ❌ No blanket accelerator speedup claim without reproducible benchmark evidence

Reviews

No reviews yet. Tell others what you thought.

Comments

No login needed. Share how it worked for you, ask a question, or note where it did not run.

No comments yet. Be the first to write one.

Comments are published right away. Do not include personal information such as email addresses or phone numbers. Up to 2 URLs.

0 / 1,000 characters

This form is protected by reCAPTCHA, and the Google Privacy Policy and Terms of Service apply.

More skills from this repository

Maintain the Abbey / Aviva / Abi Discord system prompts in the Rust abbey-bot (`src/ask.rs`) and keep the hosted Grok bots (Grok Bot Abbey, Aviva Grok) in step with them. Use when the user says "improve grok bots", "abbey prompt", "persona prompt", "ask.rs", "system prompt", "tip floor", asks to change how Abbey/Aviva/Abi speak on Discord, or after an abbey-bot PR merges and the hosted bot's tip floor needs bumping. Runs as /abbey-system-prompt.

The English summary is being prepared. Showing the original description.

donaldfilimon/abi19Updated Oct 27, 2026

Validate ABI docs and external collateral against repo evidence and docs/contracts/external-claims-audit.mdx. Use when auditing claims, scanning docs for unsupported QPS/sharding/FHE/AES language, checking one claim before publish, or re-confirming the external-claims audit. Agent procedure only — not a CLI or slash command. Pair with abi-doc-claims-sync when editing docs.

The English summary is being prepared. Showing the original description.

donaldfilimon/abi19Updated Oct 27, 2026

Synchronize ABI Markdown and docs claims with executable behavior. Use before editing README, AGENTS.md, CLAUDE.md, GEMINI.md, docs/**/*.mdx, walkthroughs, changelogs, or any old ABI Markdown that mentions WDBX, MCP, GPU, clustering, compression, FHE, agents, or benchmarks.

The English summary is being prepared. Showing the original description.

donaldfilimon/abi19Updated Oct 27, 2026

Plan and execute ABI repo work from the TODO board, roadmap/spec docs, stable Rust 1.99.1 constraints, and validation gates. Use when the user asks to find ABI todos/roadmaps, organize a large implementation from tasks/todo.md, or decide the next safe work slice. Distinct from the goals skill — that skill owns the persistent ledger tasks/goals.md; this skill plans from the active board and north-star specs.

The English summary is being prepared. Showing the original description.

donaldfilimon/abi19Updated Oct 27, 2026

Validate ABI MCP JSON-RPC stdio and its custom loopback HTTP compatibility endpoint, including auth, limits, notifications, and shutdown.

The English summary is being prepared. Showing the original description.

donaldfilimon/abi19Updated Oct 27, 2026

Review ABI's 16 compiled-in plugin fixtures, manifest validation, manual bundled registry, and real CLI/MCP run dispatch without claiming dynamic loading.

The English summary is being prepared. Showing the original description.

donaldfilimon/abi19Updated Oct 27, 2026

See all skills by donaldfilimon

Report a problem with this skill