Auditable cloud governance powered by Bayesian intelligence. Build reliable, observable, and self‑healing AI systems for real‑world infrastructure.

🔐 The core ARF engine is access‑controlled and not open source.
It is available only to qualified pilots and enterprise customers under outcome‑based pricing.

👉 ARF Control Center (public demo UI)

ARF makes AI operations provably safe, auditable, and transparent.
We provide a mathematically rigorous governance layer for deterministic and probabilistic decision‑making in production AI systems.

• ✅ Enable provably safe AI operations in cloud, hybrid, and multi‑agent environments.
• 🧮 Deliver expected loss minimisation and hybrid Bayesian inference for calibrated risk scoring.
• 🔍 Offer full traceability through auditable logs, semantic memory, and transparent decision records.
• 🧭 Steward the framework – not a free‑for‑all, but a protected, pilot‑first product.

🔒 Private repositories (core engine, API control plane, enterprise code) are not listed here. Access is restricted to pilots and enterprise customers.

• Bayesian Risk Scoring – Conjugate priors + HMC for calibrated uncertainty.
• Semantic Memory – FAISS‑based retrieval of similar past incidents.
• Expected Loss Minimisation – Chooses approve/deny/escalate by balancing risk, cost, and uncertainty.
• Multi‑Agent Orchestration – Automated anomaly detection, root cause analysis, forecasting.
• Policy Composability – AND/OR/NOT combinators for complex rules.
• Traceability & Audit – Every decision is fully auditable and queryable.

⚠️ These capabilities are implemented in the protected core engine, not in public demo assets.

• UI Concept Demo – Hugging Face Space – Interactive risk dashboard (mock data only).
• Sandbox API – Mock endpoint on Hugging Face – Returns mock responses, not real Bayesian inference. Interactive docs at /docs.

Example sandbox call (returns mock data):

curl -X POST https://a-r-f-arf-sandbox-api.hf.space/v1/evaluate \
  -H "Content-Type: application/json" \
  -d '{"service_name":"api","event_type":"latency","severity":"high"}'

The real engine is not publicly accessible.

We accept limited contributions to public repositories (arf-spec, arf-frontend, pitch-deck) – bug fixes, documentation, demo improvements.
We do not accept pull requests against private core repositories.

1. Open an issue describing your proposed change.
2. Wait for a maintainer to assign the issue.
3. Sign a Contributor License Agreement (CLA) if requested.
4. Submit a pull request referencing the issue.

All changes are reviewed and merged at the founder’s discretion.

For questions about the protected engine or pilot access, please do not open issues – use the contact details below.

The core ARF engine is not open source. To request pilot access (time‑limited free trial) or enterprise licensing, contact us directly:

When requesting access, please provide:

• Your full name and organisation
• Use case description
• Expected monthly incident volume
• Cloud environment (AWS, Azure, GCP, on‑prem)

• Public repositories (arf-spec, arf-frontend, pitch-deck) are licensed under Apache 2.0.
• The core engine and all private repositories are proprietary – no license is granted for public use.
• See the NOTICE file for full details.

All modules, dashboards, and APIs that are public are open source (Apache 2.0).
The core Bayesian engine is not open source – it is protected and access‑controlled.

Stewarded by the founder – pilot‑first, outcome‑based pricing.