Skip to main content

CORTEX C5-REAL execution kernel (BABYLON-60): BFT ledgers, onco-transducer, exergy pipelines

Project description

babylon60

Tamper-evident cryptographic ledger for autonomous AI agents
Local-first · Hash-chained · Single-writer SQLite WAL · Optional Rust core

PyPI version Python versions SQLite WAL Rust optional License

Architecture · Security Model · Experimental Features


What It Solves

Autonomous agents lack a reliable way to persist decisions with causal traceability. babylon60 provides a local-first memory substrate:

Validation → Single-Writer Queue → Hash-Chain Ledger → Git Sentinel

Every entry is linked to the previous via SHA3-256 hash. Every write carries a causal taint (who / when / why). Duplicates are rejected via UUID v5 idempotency keys. The chain is verified on read — not assumed correct.

[!IMPORTANT] tamper-evident ≠ tamper-proof. Hash-chains detect modifications after the fact. They do not prevent an attacker with filesystem access from replacing the entire database. See SECURITY_MODEL.md for the full threat model.


Maturity Matrix

Component Status Coverage Use
BFT Ledger (SQLite WAL + hash-chain) Beta ~72% Development
Cryptographic chain (BLAKE3 / SHA3-256) Beta ~68% Development
Python SDK (babylon60.*) Beta ~65% Development
Rust core (strike_rs / PyO3) Alpha ~41% Experimental
MCP integration Alpha ~30% Experimental
Vector memory (sqlite-vec) Alpha ~35% Experimental
BFT Swarm / quorum Prototype ~20% Research only
Lean 4 formal proofs Prototype ~15% Research only
LoRA daemon / HDC Design 0% Do not use

Installation

pip install babylon60
# With cryptographic signing (Ed25519, Argon2)
pip install "babylon60[crypto]"

# With voice transcription (macOS Apple Silicon)
pip install "babylon60[voice-full]"

Quickstart

from babylon60.api.client import CortexClient

client = CortexClient()
print(client.status())
# {"ledger": "ok", "chain_valid": True, "entries": 0, "version": "1.0.2"}

Start the local REST + WebSocket API:

uvicorn babylon60.api.server:app --reload
# → http://localhost:8000

Verify chain integrity programmatically:

from babylon60.bft.ledger_actor import verify_chain

result = verify_chain(db_path="master_ledger.db")
# {"valid": True, "entries": 4821, "broken_at": None}

Architecture

flowchart TD
    A[Agent / SDK Call] --> B[Validation Layer\nschema · UUID v5 idempotency]
    B --> C[asyncio.Queue\nsingle writer]
    C --> D[BFTLedgerActor]
    D --> E[(SQLite WAL\nmaster_ledger.db)]
    D --> F[SHA3-256 Hash-Chain\nprev_hash linkage]
    F --> G[Git Sentinel\nauto-commit + causal taint]
    E --> H[Read / Verify\nchain integrity check]

    style A fill:#2B3BE5,color:#fff,stroke:none
    style D fill:#1a1a2e,color:#fff,stroke:#2B3BE5
    style E fill:#0A0A0A,color:#4CAF50,stroke:#4CAF50
    style G fill:#0A0A0A,color:#CE422B,stroke:#CE422B
Ledger Entry Contract (click to expand)

Every write must satisfy this schema. Violations are rejected at the validation layer — not silently ignored.

{
    "id":           "uuid-v5",       # Idempotency key — rejects duplicates
    "prev_hash":    "sha3_256-hex",  # Chain link — breaks chain on mismatch
    "payload":      {...},           # Structured content
    "causal_taint": "agent:reason",  # Creation trace — mandatory
    "lamport_t":    int,             # Logical clock — total ordering
    "agent_id":     "str",           # Writer identity
}
Consensus Topology (M12 Levels)
Level Mechanism Status
L1 — AP/CRDT Local caches, telemetry ✅ Stable
L2 — CP single-writer master_ledger.db WAL ✅ Stable
L3 — External witness Git Sentinel (local repo) ✅ Stable
L4 — BFT quorum N≥3f+1 distributed swarm 🧪 Prototype
L5 — Blockchain anchor OTS / BTC OP_RETURN 🔬 Research

L1–L3 is the stable core. L4 and L5 are opt-in research extensions — not required for local agent memory.


Guarantees

PropertyMechanismWhat It Proves
IntegritySHA3-256 hash-chainEntry not modified after write
ProvenanceCausal taint + timestampWho wrote it and when
OrderingLamport clock + WALSerialized, reproducible history
IdempotencyUUID v5 keyNo duplicate writes
IsolationPer-tenant DB scopeWorkspace separation

Local Development

python3 -m venv .venv && source .venv/bin/activate
pip install -e ".[dev]"
pytest -x --tb=short

Build Rust core (optional — Python fallback available):

cd strike_rs && cargo build --release

Experimental

BFT Swarm, Vector Memory, Lean 4 proofs, LoRA daemon, Blockchain anchoring, and HDC are explicitly experimental. See docs/EXPERIMENTAL.md for status and known limitations. Do not use in production.


Titular Civil: CORTEX Core Dev · AKA: Borja Moskv (borjamoskv)
All Rights Reserved — Proprietary & Trade Secret
Art. 6.1, 6.2, 14 LPI (España) · Convenio de Berna · Ley 1/2019 de Secretos Empresariales

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

babylon60-1.0.2.tar.gz (173.1 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

babylon60-1.0.2-py3-none-any.whl (185.7 kB view details)

Uploaded Python 3

File details

Details for the file babylon60-1.0.2.tar.gz.

File metadata

  • Download URL: babylon60-1.0.2.tar.gz
  • Upload date:
  • Size: 173.1 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.11.15

File hashes

Hashes for babylon60-1.0.2.tar.gz
Algorithm Hash digest
SHA256 3e55d4eb077d89bbe73bda92ab9b87e297f5962de0591cfb0c9037003d739170
MD5 3a9e2516d87e197e388276d6fa806ee2
BLAKE2b-256 ee36865e81a2334b9a8164823ed5c6df7e13decf8a018ca996f76a04d5c6502e

See more details on using hashes here.

File details

Details for the file babylon60-1.0.2-py3-none-any.whl.

File metadata

  • Download URL: babylon60-1.0.2-py3-none-any.whl
  • Upload date:
  • Size: 185.7 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.11.15

File hashes

Hashes for babylon60-1.0.2-py3-none-any.whl
Algorithm Hash digest
SHA256 44277399fe60ebe93779dc45510233e9787d5405eafcdd7e95378d1020b3b86f
MD5 d2f13d419021f1239f42250f2c2eee17
BLAKE2b-256 43f92bb2a3783aa64815e891139cec4f74e35e28e4fb263b5fe52d636c0464c3

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page