ARC-Apache SURE Binary Memory Pack v5

ARC-Apache is the binary-first cryptographic memory substrate for the ARC ecosystem. It treats every durable knowledge object as a deterministic binary object first, then binds that object to hashes, Merkle roots, manifests, receipts, signatures, and optional encryption.

This package is designed as the bridge between:

• ARC-Core — authority, receipts, policy, event registration
• ARC Language Module — lexical / meaning spine mirrored into binary objects
• ARC-StreamMemory — screen, sensor, frame, terminal, and visual memory streams
• ARC-Neuron LLMBuilder — datasets, candidates, GGUF/model artifacts, benchmark lineage
• OmniBinary Runtime — binary intake and deterministic mirror discipline
• Arc-RAR — portable rollback / proof bundles
• SURE — seeded reconstruction math for large generated payloads
• Proto-Synth Grid Engine — visual cognition shell for graphs, receipts, streams, and memory maps
• ARC Lucifer Cleanroom Runtime / Cognition Core — event-sourced runtime and promotion lanes

Current capability

v5 provides a runnable reference implementation for the core proof loop:

python scripts/arc_apache.py init-store --store .arc_apache
python scripts/arc_apache.py keygen --store .arc_apache --name local-node
python scripts/arc_apache.py pack README.md --store .arc_apache --content-class document --codec utf8-text
python scripts/arc_apache.py verify .arc_apache/manifests/<manifest>.json --store .arc_apache
python scripts/arc_apache.py receipt .arc_apache/manifests/<manifest>.json --store .arc_apache --source manual
python scripts/arc_apache.py sign-receipt .arc_apache/receipts/<receipt>.json --store .arc_apache --key-name local-node
python scripts/arc_apache.py verify-receipt .arc_apache/receipts/<receipt>.json --store .arc_apache

It supports:

• deterministic binary envelopes
• chunked object storage
• SHA-256 payload hashes
• per-chunk SHA-256 hashes
• Merkle roots
• manifest hashes
• receipt hashes
• Ed25519 receipt signing when cryptography is installed
• AES-GCM encrypted object packing when cryptography is installed
• language-module binary mirroring
• stream/frame manifest generation
• SURE seed-recipe objects
• Arc-RAR bundle planning
• ARC-Core route stubs
• tests and smoke validation

Doctrine

All durable ARC information -> canonical binary object -> cryptographic proof -> receipt -> authorized projections.

Text, JSON, SQLite, UI views, markdown, screenshots, source files, model files, frames, benchmark rows, and language graphs are not the final truth surface. They are source forms or projections. The canonical durable form is the binary object plus its proof chain.

Install / run

No install is required for the core pack/verify/restore path. Python 3.10+ is recommended.

Optional cryptographic signing/encryption requires:

python -m pip install cryptography

Run tests:

python -m pytest tests

Run a no-pytest smoke check:

python scripts/arc_apache.py smoke --store .arc_apache_smoke

Public positioning

ARC-Apache is not marketed as a finished AGI. It is a real-world substrate for systems that need memory integrity, replayability, provenance, dataset lineage, model artifact lineage, and cryptographic proof for large binary/runtime objects.

The honest claim:

ARC-Apache gives ARC systems a binary-first, cryptographically verifiable memory layer that can support language, runtime, stream, model, and seeded-generation objects without treating loose text or opaque blobs as trusted truth.