Aura-Ops: Edge-Native Incident Copilot

Aura-Ops is a stateful, edge-native incident detection and debugging engine. Built entirely on Cloudflare Workers and Durable Objects, it shifts observability from a centralized, high-latency model to an atomic, near-zero-latency architecture at the edge.

⚡ Live Demo (Test in 30 Seconds)

To see the Durable Object state coordination and Workers AI analysis in action without touching a terminal:

Trigger Incident: Click here to simulate a 6-error failure on 'auth-api'
- Logic: This programmatically sends 6 logs, triggers the anomaly threshold, and invokes the AI SRE analysis instantly.

Chat with the Engine: Ask the Copilot about the system state using the chat endpoint:

curl -X POST https://cf_ai_aura-ops.sakshamgupta-ai.workers.dev/chat/auth-api \
  -H "Content-Type: application/json" \
  -d '{"message": "What happened to the auth-api and how do I fix it?"}'

🏗️ Architecture

Aura-Ops avoids the "chat with PDF" trap by treating AI as a deterministic component of a distributed system.

graph TD
    A[Logs API Ingest] -->|Route to DO| B(Incident Engine - Durable Object)
    B -->|State Management| C{Anomaly Detector}
    C -->|Trigger| D[Workers AI - Llama 3.1]
    D -->|Contextual Analysis| E[Active Incident State]
    E -->|API / Chat| F[Aura-Ops Copilot]

Core Innovations

Atomic State Coordination: Uses Durable Objects to maintain per-service incident state. This ensures that log aggregation is consistent and race-condition-free, even in a globally distributed environment.
Edge-Native Intelligence: Logic and AI inference occur within the same Cloudflare PoP as the incoming logs, eliminating cross-region egress latency.
Type-Strict Infrastructure: Built with strict TypeScript contracts (no any), mirroring the rigor of production-grade systems engineering.

🛠️ Proof of Concept

The engine processes incidents in real-time by receiving raw logs, performing anomaly detection, and triggering the Workers AI Copilot to provide actionable fixes.

1. Incident Detection & Analysis

2. Contextual AI Chat (RAG)

Sent via Bruno (see tests/ folder for collection):

🚀 Getting Started

Install Dependencies:
```
bun install
```
Deploy:
```
bun run deploy
```
Project Documentation:
- Engineering Journal & AI Prompts: A detailed log of architectural decisions, AI prompts, and debugging hurdles (TLS mismatches, Stream consumption fixes).

🛤️ Roadmap & Future Improvements

Aura-Ops is a high-performance PoC. Future architecture upgrades include:

Persistent Storage Hook: Implement R2 integration to archive incident snapshots for long-term audit trails.
Alerting Integrations: Build a dedicated webhooks service to push active alerts to Slack/PagerDuty.
Vector Store Integration: Leverage Vectorize to perform RAG on historical logs for long-term pattern recognition.

⚖️ Built with

Cloudflare Workers (Compute)
Durable Objects (Stateful Memory)
Workers AI (Llama 3.1-8b-instruct)
Hono (Routing)
TypeScript (Safety)

Built with love ❤️ by Saksham Gupta

FilesExpand file tree

README.md

Latest commit

History