petter2025 commited on
Commit
c90025f
·
verified ·
1 Parent(s): 216a183

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +45 -38
README.md CHANGED
@@ -9,66 +9,73 @@ license: apache-2.0
9
  short_description: Bayesian governance for agentic systems
10
  ---
11
 
12
- # Agentic Reliability Framework (ARF)
13
 
14
- [![GitHub](https://img.shields.io/badge/github-arf--foundation-blue?logo=github)](https://github.com/arf-foundation/agentic-reliability-framework)
15
- [![License](https://img.shields.io/badge/license-Apache%202.0-green)](LICENSE)
16
- [![Python Versions](https://img.shields.io/badge/python-3.10%2B-blue)](https://pypi.org/project/agentic-reliability-framework/)
17
- [![Hugging Face Spaces](https://img.shields.io/badge/🤗-spaces-orange)](https://huggingface.co/A-R-F)
18
-
19
- **Open‑source advisory engine for cloud infrastructure governance.**
20
- ARF provides **provably safe, mathematically grounded recommendations**—approve, deny, or escalate—when users request provisioning, configuration, or access changes.
21
 
22
  ---
23
 
24
- ## 🔍 What We Do
25
 
26
- - **Bayesian Online Learning** – Fast, conjugate updates for per‑category risk using beta‑binomial models.
27
- - **Offline Pattern Discovery** – Hamiltonian Monte Carlo (HMC/NUTS) logistic regression captures complex interactions (time‑of‑day, user role, environment).
28
- - **Composable Policy Algebra** – Build fine‑grained rules with AND/OR/NOT combinators.
29
- - **Semantic Memory** – FAISS‑based retrieval of similar past incidents for context‑aware decisions.
30
- - **Deterministic Probability Thresholds (DPT)** – Clear approve/deny/escalate decisions based on calibrated failure probabilities.
31
 
32
- ---
 
 
 
 
 
33
 
34
- ## 📊 Key Mathematical Insights
35
 
36
- | Concept | Implementation |
37
- |--------|----------------|
38
- | **Conjugate Priors** | Per‑category Beta priors, updated online with outcomes. |
39
- | **HMC Sampling** | Logistic regression with NUTS, serialized to JSON for hot‑loading. |
40
- | **Risk Fusion** | Dynamic weighted combination of conjugate, hyperprior, and HMC estimates. |
41
- | **DPT** | `Approve` if P(failure) < 0.2; `Deny` if > 0.8; otherwise `Escalate`. |
 
 
 
 
 
 
42
 
43
  ---
44
 
45
- ## 🚀 Quick Links
46
 
47
- - 📦 **[GitHub Repository](https://github.com/arf-foundation/agentic-reliability-framework)**Source code, issues, contributions.
48
- - 🤗 **[ARF v4 Demo Space](https://huggingface.co/spaces/A-R-F/Agentic-Reliability-Framework-v4)**Interactive Bayesian risk dashboard.
49
- - 🔧 **[Legacy API Demo](https://huggingface.co/spaces/A-R-F/Agentic-Reliability-Framework-API)**v3.3.9 API for backward compatibility.
50
- - 📚 **[Documentation](https://arf-foundation.github.io/arf-spec/)** – Full specs, tutorials, and API reference.
51
 
52
  ---
53
 
54
- ## 🧪 Try It Now
55
 
56
- Click the Spaces below to interact with live demos. The v4 Space showcases the full Bayesian engine with real‑time risk scoring and policy evaluation.
 
57
 
58
  ---
59
 
60
- ## ��� Contributing
61
 
62
- ARF is open source under the Apache 2.0 license. We welcome contributions of all kinds—code, documentation, ideas, or feedback.
 
 
 
 
63
 
64
- - [Contribution Guidelines](https://github.com/arf-foundation/agentic-reliability-framework/blob/main/CONTRIBUTING.md)
65
- - [Open an Issue](https://github.com/arf-foundation/agentic-reliability-framework/issues)
66
- - [Start a Discussion](https://github.com/arf-foundation/agentic-reliability-framework/discussions)
67
 
68
  ---
69
 
70
- ## 📫 Contact
 
 
 
71
 
72
- - Email: [petter2025us@outlook.com](mailto:petter2025us@outlook.com)
73
- - LinkedIn: [petterjuan](https://linkedin.com/in/petterjuan)
74
- - Book a call: [Calendly](https://calendly.com/petter2025us/30min)
 
9
  short_description: Bayesian governance for agentic systems
10
  ---
11
 
12
+ # Agentic Reliability Framework (ARF) – Technical Overview
13
 
14
+ > 🔒 **The core ARF engine is access‑controlled and not open source.**
15
+ > Available to qualified pilots under outcome‑based pricing.
16
+ > The **public specification** and **demo UI** are Apache 2.0.
 
 
 
 
17
 
18
  ---
19
 
20
+ ## 🧠 Core Engine (Protected)
21
 
22
+ **Bayesian Risk Fusion** – Conjugate Beta priors (online) + HMC logistic regression (offline) + optional hierarchical hyperpriors.
 
 
 
 
23
 
24
+ \[
25
+ \text{risk} = w_{\text{conj}}\cdot\frac{\alpha}{\alpha+\beta} \;+\; w_{\text{hmc}}\cdot p_{\text{hmc}} \;+\; w_{\text{hyper}}\cdot \mu_{\text{hyper}}
26
+ \]
27
+
28
+ - Weights are dynamic: \(w_{\text{hmc}} = \min(0.6, n/n_0)\), \(w_{\text{hyper}} = \min(0.3, \text{base}\cdot(1-w_{\text{hmc}}))\)
29
+ - Posterior variance → 90% HDI for uncertainty quantification.
30
 
31
+ **Expected Loss Minimisation** Chooses `APPROVE`, `DENY`, or `ESCALATE` by minimising:
32
 
33
+ \[
34
+ \begin{aligned}
35
+ L_{\text{approve}} &= \text{COST\_FP}\cdot R + \text{COST\_IMPACT}\cdot b_{\text{mean}} + \dots \\
36
+ L_{\text{deny}} &= \text{COST\_FN}\cdot(1-R) + \text{COST\_OPP}\cdot v_{\text{mean}} \\
37
+ L_{\text{escalate}} &= \text{COST\_REVIEW} + \text{COST\_UNCERTAINTY}\cdot\psi
38
+ \end{aligned}
39
+ \]
40
+
41
+ **Execution Ladder (Rust)** – Mechanical gates: `license`, `confidence`, `risk`, `rollback`, `causal`.
42
+ **Lyapunov stability** – Quadratic candidate \(V(x,r) = \alpha r^2 + \beta\|x-x_{\text{des}}\|^2\) ensures healing actions converge.
43
+
44
+ **Cryptographic signing** – Ed25519 signatures for `HealingIntent` (Python `cryptography`, Rust `ed25519_dalek`).
45
 
46
  ---
47
 
48
+ ## 📄 Public Specification (Open Source)
49
 
50
+ - [Data models & API contracts](https://github.com/arf-foundation/arf-spec) – `InfrastructureIntent`, `HealingIntent`, `RiskScore`, `GovernanceLoop`
51
+ - [Mathematics](https://arf-foundation.github.io/arf-spec/mathematics/) – Full Bayesian derivations, Lyapunov proof sketch
52
+ - [Governance loop](https://arf-foundation.github.io/arf-spec/governance/) – Step‑by‑step flow with constants (`COST_FP`, `COST_FN`, `EPISTEMIC_ESCALATION_THRESHOLD`)
 
53
 
54
  ---
55
 
56
+ ## 🧪 Live Demos (Mock Data Only)
57
 
58
+ - [Interactive Risk Demo](https://huggingface.co/spaces/A-R-F/Agentic-Reliability-Framework-v4) Adjust priors, see HMC simulation, semantic memory retrieval.
59
+ - [Sandbox API](https://sandbox.arf.dev/docs) – Mock endpoint (returns simulated responses, no real Bayesian engine).
60
 
61
  ---
62
 
63
+ ## 🔗 Public Repos
64
 
65
+ | Repo | Description |
66
+ |------|-------------|
67
+ | [`arf-spec`](https://github.com/arf-foundation/arf-spec) | Canonical specification (Apache 2.0) |
68
+ | [`arf-frontend`](https://github.com/arf-foundation/arf-frontend) | Demo UI (Next.js, Tailwind) |
69
+ | [`pitch-deck`](https://github.com/arf-foundation/pitch-deck) | Public overview |
70
 
71
+ Private repos (`agentic_reliability_framework`, `arf-api`, `enterprise`) are **pilot/enterprise only**.
 
 
72
 
73
  ---
74
 
75
+ ## ✈️ Pilot Access
76
+
77
+ Request access via email: **petter2025us@outlook.com** (include use case, volume, cloud environment).
78
+ Outcome‑based pricing – pay for risk reduction, not API calls.
79
 
80
+ - [Live spec site](https://arf-foundation.github.io/arf-spec/)
81
+ - [Demo dashboard](https://arf-frontend-sandy.vercel.app)