🚀 AURA-Engine: Multimodal Agentic RAG & Incident Operations Engine

---

🖥️ System Architecture & Interface

Figure 1: AURA-Engine Streamlit dashboard displaying query analysis, Qdrant vector retrieval confidence scoring, raw log context auditability, and Human-in-the-Loop (HITL) action plan approval.

---

💡 System Overview

AURA-Engine is an enterprise-grade, agentic Retrieval-Augmented Generation (RAG) platform built to automate IT incident diagnosis and operational troubleshooting safely.

Unlike passive search engines or simple wrapper APIs, AURA-Engine combines high-dimensional vector retrieval, agentic graph routing, and deterministic safety controls to analyze system logs, trace error cascades, and recommend human-verified remediation steps.

---

🔥 Key Technical Innovations

• Agentic LangGraph Routing: Dynamically directs incoming queries through state-machine nodes to handle log parsing, vector lookup, contextual retrieval, and action planning.
• Qdrant Vector Engine: Real-time context retrieval leveraging dense vector embeddings with metadata filtering and confidence scoring.
• Human-in-the-Loop (HITL) Safety: Prevents risky automated execution by requiring explicit human verification and state interruption prior to triggering remediation workflows.
• Adversarial Security & Guardrails: Enforces structured Pydantic schemas and XML context separation to isolate untrusted user inputs from system agent execution logic.
• Non-Blocking Ingestion: Optimized multipart/form-data endpoint handling for high-throughput log file ingestion.
• Container-Ready Stack: Includes Dockerfile and docker-compose.yml configurations for multi-service deployment.

---

📂 Project Structure

aura-engine/
├── app/                  # Backend core modules & LangGraph agent logic
├── assets/               # System screenshots, architecture diagrams, and media
│   └── dashboard.png     # Application UI screenshot
├── frontend/             # Streamlit interactive dashboard
│   └── app.py            # Dashboard user interface
├── qdrant_db/            # Local vector database storage index
├── docker-compose.yml    # Multi-service container config
├── Dockerfile            # Production container build file
├── main.py               # FastAPI server entry point & REST endpoints
├── test_graph.py         # Test suite for LangGraph state machine flow
├── test_vector.py        # Vector embedding & search benchmarking scripts
└── requirements.txt      # Python dependencies

---

🛠️ Quickstart Guide

1. Prerequisites

• Python 3.9+
• Virtual environment (venv)

2. Local Setup

Clone the repository and set up your virtual environment:

git clone [https://github.com/mdrafique15/aura-engine.git](https://github.com/mdrafique15/aura-engine.git)
cd aura-engine/aura-engine
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

3. Running the Stack

Step 1: Start Backend (FastAPI)

uvicorn main:app --reload --port 8000

Step 2: Start Frontend (Streamlit)

Open a separate terminal window or tab, activate the virtual environment, and run:

source venv/bin/activate
streamlit run frontend/app.py

---

🔒 Security Architecture

1. Input Sanitization: Strict Pydantic validation on all incoming REST JSON payloads.
2. Context Isolation: Embeds user query payloads inside explicit XML tag context wrappers to mitigate indirect prompt injection threats.
3. Deterministic Interrupts: LangGraph agent state pauses at designated approval nodes before enacting operational changes.