| --- |
| 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.** |
|
|
| [](https://github.com/petter2025us/agentic-reliability-framework) |
| [](https://github.com/petter2025us/agentic-reliability-framework/blob/main/TUTORIAL.md) |
| [](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>=4.0.0` |
| - **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.* |
