SOVEREIGN RUNTIME • 100% AIR-GAPPED
RELEASE: v2.2.2 (Production Freeze) • RUST 1.80+

Sovereign, in-process autonomous software engineering runtime. Transforms open-weights SLMs (3B to 7B) into verified compilers. Executes completely offline without cloud API tokens, external daemons, or telemetry.

INSPECT VERIFICATION KERNEL PLATFORM MATRIX

Sovereign Platform Support Matrix

Cross-Platform Execution Invariants
Linux (x86_64)
ACTIVE PRODUCTION CORE
  • Sandboxing bwrap (Kernel Namespaces)
  • Hardware AVX2 / AVX-512 & NVIDIA CUDA
  • Inference In-Process C-FFI (RAM link)
  • Lifecycle Continuous Production Release
macOS (Apple Silicon)
v2.2.2 FROZEN SNAPSHOT
  • Sandboxing Seatbelt (Deprecated) + RAM VFS
  • Hardware Metal GPU (libggml-metal)
  • Architecture arm64 (M1–M4 Native)
  • Lifecycle Frozen Community Build (Unsupported)
Windows (x86_64)
WSL2 SUPPORTED
  • Sandboxing WSL2 Linux Cleanrooms
  • Hardware Native CPU & CUDA Passthrough
  • Inference Zero-Copy VFS Mount
  • Lifecycle Continuous via WSL2 Integration

Deterministic Cleanroom Execution

Live Verification Daemon Spec
PID: 49201 // CLEANROOM ISOLATED
// Omni Engine CLI v2.2.2 — Autonomous Multi-File Compiler
$ omni run --target atlr_ledger --cleanroom --isolated
[1/4] Initializing In-Process C-FFI RAM model (Qwen2.5-Coder-7B)...
> C-FFI: linked llama.cpp directly in RAM. Zero network socket overhead.
> VFS: initialized RAM Virtual File System at /dev/shm/omni_vfs_gate
[2/4] Topological Directed Acyclic Graph (DAG) construction...
> DAG: planned dependencies: models.rs → ledger.rs → lib.rs
> Gate: Zero-Wipe Invariant (W = ∅) enforced. Regressions strictly prohibited.
[3/4] Spawning unprivileged cleanroom sandbox...
> Sandbox: unshare-all + --unshare-net (0 physical network access)
> Compiler: cargo check --lib --tests completed in cleanroom (Exit Code: 0)
[4/4] 50-thread concurrent ATLR stress test verification...
✓ EXIT CODE: 0 (Pristine). All 50 threads completed with total liquidity conservation.
✓ 3 modules committed to host disk: [models.rs, ledger.rs, lib.rs]

Mathematical Governance Pillars

Replacing ReAct Token Loops with Invariants
PILLAR 01

Direct In-Process C-FFI

Upstream inference is linked directly into engine memory. Eliminates HTTP/TCP socket latency, unlocking sub-millisecond causal KV-cache rollbacks and token steering.

PILLAR 02

Deterministic Cleanrooms

Code generation is staged in RAM VFS and verified inside Bubblewrap (Linux) or Apple Seatbelt (macOS). Unverified code never touches host storage.

PILLAR 03

Zero-Wipe Invariant ($W = \emptyset$)

Mathematical guarantee that verified functions and classes are permanently frozen in memory. Erasing working code during refactoring is physically impossible.

PILLAR 04

Topological DAG Planning

Multi-file projects are scheduled topologically according to strict type and symbol dependencies, preventing out-of-order synthesis hallucination loops.

Empirical ATLR Concurrency Benchmark

50-Thread Atomic Ledger High-Concurrency Stress Test
Runtime & Model Environment Makespan Exit Code Concurrency Logic Outcome
Omni Engine (Local 7B) Google Colab (T4 GPU) 37m 58s Exit 0 RwLock + Mutex 100% Passed (Clean)
Omni Engine (Local 4B) Commodity 8-Core CPU 1h 39m Exit 0 Canonical Min/Max Lock 100% Staged via Gate
Aider Local (4B Ollama) Commodity 8-Core CPU >5h 24m (Timeout) kill -9 Never Reached 0 files delivered
Aider Cloud (Gemini 2.5 Flash) Cloud Infrastructure 58 seconds Exit 101 sort_unstable FAILED (Panic test)