README.md

---
title: ARF v4 – Reliability Lab
emoji: 🧠
colorFrom: blue
colorTo: purple
sdk: docker
pinned: false
license: apache-2.0
short_description: ARF v4 – Bayesian reliability demo
---

# 🧠 ARF v4 – Reliability Lab

This Space hosts a live, interactive demo of the **Agentic Reliability Framework v4 (OSS edition)**. It showcases the core intelligence engine – a hybrid Bayesian + Hamiltonian Monte Carlo (HMC) system that evaluates infrastructure incidents and produces advisory healing recommendations.

**All outputs are advisory only – no execution.**

[![GitHub](https://img.shields.io/badge/GitHub-Repository-blue?logo=github)](https://github.com/petter2025us/agentic-reliability-framework)
[![Tutorial](https://img.shields.io/badge/📘-Tutorial-green)](https://github.com/petter2025us/agentic-reliability-framework/blob/main/TUTORIAL.md)
[![Contact](https://img.shields.io/badge/📧-Email-yellow)](mailto:petter2025us@outlook.com)

---

## 🚀 How It Works

The demo uses the `EnhancedReliabilityEngine` from the ARF v4 package. When you submit telemetry (component, latency, error rate, etc.), the engine:

1. **Runs three specialised agents** in parallel:
   - **Detective** – anomaly detection and pattern recognition
   - **Diagnostician** – root cause analysis
   - **Predictive** – forecasting and trend detection

2. **Computes a risk score** using:
   - Online **Bayesian conjugate priors** (Beta‑Binomial) per action category
   - Offline **Hamiltonian Monte Carlo (HMC)** with NUTS (if trained) for complex patterns
   - A weighted blend of both for the final score

3. **Applies deterministic policy thresholds** (DPT) to recommend: `APPROVE`, `DENY`, or `ESCALATE`.

4. **Returns a JSON** containing the risk score, agent insights, healing actions, and (if configured) a Claude‑generated executive summary.

---

## 🧪 Try It Yourself

Just fill in the form on the left and click **Analyze**. The output will appear as a formatted JSON.

Example input:
- **Component**: `api-service`
- **Latency P99**: `250 ms`
- **Error Rate**: `0.08`
- **Throughput**: `1000 req/s`
- **CPU Utilization**: `0.7`
- **Memory Utilization**: `0.6`

Expected risk score: ~0.12 (low) → `APPROVE`.

---

## 📦 How This Space Is Built

- **Base image**: `python:3.10` (via Dockerfile)
- **Dependencies**:
  - `git+https://github.com/petter2025us/agentic-reliability-framework.git@v4.0.0`
  - `gradio>=6.10.0`
- **Source code**: a minimal `app.py` that imports and runs `EnhancedReliabilityEngine`.

All code is open source and available in the [main repository](https://github.com/petter2025us/agentic-reliability-framework).

---

## 🏃 Run Locally

You can run the exact same demo on your own machine:

```bash
git clone https://github.com/petter2025us/agentic-reliability-framework.git
cd agentic-reliability-framework
python -m venv venv
source venv/bin/activate
pip install -e .
pip install gradio
python examples/app.py   # or copy the Space's app.py
```
Then open `http://localhost:7860`.

---

## 📚 Learn More

- 📘 [Full Tutorial](https://github.com/petter2025us/agentic-reliability-framework/blob/main/TUTORIAL.md)
- 🐙 [GitHub Repository](https://github.com/petter2025us/agentic-reliability-framework)
- 📖 [Contributing Guidelines](https://github.com/petter2025us/agentic-reliability-framework/blob/main/CONTRIBUTING.md)
- 💼 [Enterprise Inquiries](mailto:petter2025us@outlook.com)

---

## 📬 Contact

- **Email**: [petter2025us@outlook.com](mailto:petter2025us@outlook.com)
- **LinkedIn**: [petterjuan](https://linkedin.com/in/petterjuan)
- **Book a call**: [Calendly – 30 min](https://calendly.com/petter2025us/30min)

---


*Powered by ARF v4 – Bayesian reliability for autonomous infrastructure.*