---

title: "a11oy — Command Center" emoji: "🛡️" thumbnail: "https://a-11-oy.com/og-card.png" colorFrom: indigo colorTo: gray sdk: docker app_port: 7860 pinned: true license: apache-2.0 short_description: "a11oy source — product origin a-11-oy.com, proof a11oy.net" tags:

• governance
• agentic-ai
• doctrine-v11
• a11oy
• slsa-l1
• apache-2.0 ecosystem-stage: "operational" models: [SZLHOLDINGS/SZL-Khipu-1.5B, SZLHOLDINGS/SZL-Forge-1.5B-ReceiptAgent] datasets: [SZLHOLDINGS/a11oy-verifiable-corpus, SZLHOLDINGS/szl-lake]

---

a11oy

This repository is the source for the product origin https://a-11-oy.com. Proof lives at https://a11oy.net.

a11oy.com is not a surface of this project.

a11oy is a governed-AI Command Center: deny-by-default policy, trust ceiling 0.97, and a signed receipt for every decision. This tree is that system's source. It is not the proof registry.

Invariant

receipts.in ≡ receipts.out

A governed action is admitted only as a signed receipt and leaves only as that same signed receipt. The two sides are identical. A mismatch is a failed verification, not a display error.

Live surfaces

Measured 2026-09-04: the apex a-11-oy.com serves the Space runtime (server=szl, x-szl-space: a11oy). Hugging Face runtime.domains lists a-11-oy.com as READY. HTTP 200 is reachability, not a production certificate. www.a-11-oy.com TLS is UNAVAILABLE.

a11oy-factory is a bind of this source. It is not a second flagship.

Archive: Warhacker v1.0.0 is ARCHIVED.

---

The proof backbone

The trust math behind a11oy is pinned in Lean 4 and checked by a proof machine:

• 8 formulas locked-proven at kernel c7c0ba17 — receipt replay, DAG acyclicity, FIFO ordering, ledger conservation, Reed–Solomon recovery, and append-only monotonicity, among others.
• Λ unconditional uniqueness = Conjecture 1 — machine-checked false (we found a counterexample). Conditional uniqueness is proven axiom-free (Theorem U). We say both out loud.
• SLSA L1 honest · L2 build-attested · L3 roadmap. No FedRAMP or ATO claimed.

Full proof library: szl-holdings/lutar-lean

---

Verify it yourself

# Verify the build attestation
gh attestation verify oci://ghcr.io/szl-holdings/a11oy:latest --repo szl-holdings/a11oy

# Check live doctrine posture on the product apex (Space-backed as of 2026-09-04)
curl -s https://a-11-oy.com/api/a11oy/v1/honest | jq .doctrine_lock.lambda
# → "Conjecture 1" when the runtime answers that contract

---

Runtime API surfaces

These paths exist on the Space and, as of 2026-09-04, on the apex.

Persistent receipt storage (HF Space)

The protected deployment workflow attaches the existing SZLHOLDINGS/szl-evidence Storage Bucket read-write at /data, preserving any other attached volumes and failing closed if another volume already claims that mount. The Series-A database is namespaced at:

A11OY_SERIES_A_DB=/data/a11oy/series-a/control-plane.sqlite3

Production also sets A11OY_REQUIRE_PERSISTENT_STORAGE=1, A11OY_SERIES_A_REQUIRE_MOUNT=/data, and the network-filesystem-safe SQLite rollback journal. If the bucket is detached or the database path escapes the mount, Series-A registration fails closed instead of falling back to /tmp. The unified Khipu and energy ledgers use separate /data/a11oy/* paths.

Required HF Space secrets for full signing integrity:

• SZL_COSIGN_PRIVATE_PEM — canonical ECDSA P-256 private PEM shared by all receipt surfaces. The deployment sets A11OY_REQUIRE_PERSISTENT_SIGNING=1, so an absent or malformed key disables signing instead of minting a replacement identity.

Check current signing and storage status at GET /api/a11oy/v1/signing-status and GET /api/a11oy/v1/series-a/status on the product origin.

---

Honest status

---

Truth gates

Round-10 payload wiring: three stdlib-first gates plus the 12-step acceptance demo, run by the truth-gates workflow on every PR and push.

python3 tools/docs_lexicon_gate.py       # docs language gate: banned phrases, empty truth states
python3 tools/commercial_release_gate.py # commercial truth gate: COMMERCIAL_LEDGER + claims-ledger
python3 tools/demo_harness.py --conformance evidence/conformance/eu-ai-act-article-12.v1.yaml

Expected first-run exit codes: docs_lexicon_gate.py 1 and commercial_release_gate.py 1 — those failures are intended. The lexicon output is the Week 1 documentation fix list; the release-gate output is the 24 raise-blocking COMMERCIAL_LEDGER rows still UNKNOWN (the real work, per the funding discipline). demo_harness.py must stay 0 (12/12 steps) on every PR. The two observe-mode gates run continue-on-error until their finding lists reach zero; flip procedure: docs/RUNBOOK_WEEK1.md.

Naming note (no file was overwritten): tools/lexicon_gate.py and tools/release_gate.py already exist here as the round-5 gates, so the round-10 gates are wired under the names above; both generations run. The v1 Article 12 logging conformance profile lives at evidence/conformance/eu-ai-act-article-12.v1.yaml because the CANON path is occupied by an older, structurally different artifact. Fresh scaffolds are created non-destructively with python3 tools/szl_master_bootstrap.py --run.

---

Shared modules (must not drift)

a11oy_agent_loop.py, a11oy_mcp_client.py, and operator_shell_v4.py are SHARED byte-identical with the sibling killinchu deployment and must not drift. An in-repo ratchet pins their SHA-256 in .shared_module_hashes.json; the Shared-module hash lock workflow fails if any of them changes without the lock being regenerated. When a change is intentional, regenerate the lock in the same PR and mirror the edit to killinchu (cross-repo enforcement is a follow-up):

python3 .github/shared-module-hash-check.py --update

---

Governed Delta Workspace

Runtime write status is configuration-bound. GDW reports REAL only when its secret-managed credential registry, canonical governance gates, verified persistent storage, exact schema, and a fresh generation-bound supervised outbox pass are ready. Otherwise it reports UNAVAILABLE and writes fail closed.

The public deployment remains UNAVAILABLE until this corrective source is protected-merged, exact-source relocked, and the production credential and persistence contracts are observed live. Source tests are not deployment evidence.

GDW Frontier Push Pack is a MODELED instrumentation and verification extension for the Governed Delta Workspace. It provides load testing, operator validation, hybrid scheduling research hooks, KDA-vs-MLA memory benchmarking, and Lean-oriented proof export. It does not claim frontier benchmark superiority, proprietary activation access, or production-scale guarantees beyond the measured harness outputs.

The authenticated runtime, Postman collection, load tools, offline dashboard, memory benchmark, proof-input bridge, and fail-closed readiness conditions are documented in docs/gdw-frontier.md. A checked theorem is reported separately from an exported theorem input, and every throughput result is scoped to its captured run.

Learn more

• WILLAY API reference
• Governed run-loop recipe
• Proof library — lutar-lean
• Associated research-program concept DOI — 10.5281/zenodo.19944926
• Existing formal-artifact record — 10.5281/zenodo.20434276
• A11oy software releases — the v1.1.0 software-version DOI stays PENDING_ZENODO_READBACK until Zenodo resolves the immutable release
• Canonical product surface · runtime Space · public proof registry a11oy.net

---

---

◇ Part of the SZL Holdings estate — governed AI you can prove

One sovereign substrate, many organs — every decision carries a signed, checkable receipt.

◇ Holographic Estate — the showcase · 🛡️ a11oy · 🧬 IMMUNE · 🦅 killinchu · 🫀 anatomy · 🌌 cosmos · 🛰️ SDA · 🌊 yarqa · 🤗 all Spaces

_{Doctrine v11 · Λ = Conjecture 1, never green · honest by design · public data only.}