Skip to content

arf-foundation/.github

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

34 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Agentic Reliability Framework (ARF) – Stewarded Governance for AI Systems

ARF is a production‑ready reliability and governance layer for AI systems, combining Bayesian risk scoring, semantic memory, auditability, and self‑healing loops.
The framework is founder‑controlled and pilot‑first – the core engine is protected, access‑gated, and offered to selected enterprises under outcome‑based pricing.

⚠️ Important – The core ARF engine (agentic_reliability_framework, arf-api, enterprise) is not open source. It is proprietary, access‑controlled, and available only to qualified pilots and enterprise customers.


What ARF does

  • Observes agent behaviour and system health
  • Detects anomalies and failure patterns using Bayesian inference
  • Applies governance and policy controls (approve/deny/escalate)
  • Supports adaptive healing and recovery (self‑healing loops)
  • Stores incidents, outcomes, and learnings for future decisions (semantic memory)

Public vs. Private – Repository Access

Repository Status Description
arf-spec Public Canonical specification – data models, API contracts, decision rules (Apache 2.0)
arf-frontend Public Demo dashboard UI (sanitised, uses mock data)
pitch-deck Public Public overview and vision
agentic_reliability_framework Private Protected core engine – Bayesian risk scoring, expected loss minimisation
arf-api Private Production control plane – FastAPI service
enterprise Private Customer‑specific adaptations, proprietary heuristics
research Private Experimental methods and internal models

Access & Pilots

The core ARF engine is not publicly available. Access is granted only to:

  • Selected pilot customers – time‑limited free trial under a mutual qualification agreement
  • Enterprise licensees – outcome‑based pricing tied to measurable risk reduction

To request pilot access, email [email protected] with:

  • Use case description
  • Expected monthly incident volume
  • Cloud environment (AWS, Azure, GCP, on‑prem)

Monetisation – Outcome‑Based Pricing

ARF pricing is not per‑seat or per‑request. It is directly tied to the amount of operational risk the system removes from your AI workflows.
We measure risk reduction via auditable pre/post Bayesian scores – you pay only for verified improvement.


Contributing (Public Repositories Only)

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.


License

  • Public repositories (arf-spec, arf-frontend, pitch-deck) are licensed under Apache 2.0.
  • The core engine and all private repositories are proprietary – not licensed for public use.

Community & Contact


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

About

Stewarded, pilot‑first reliability framework for AI systems. Core engine is access‑controlled.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors