DocOps Enterprise 🚀

DocOps is an advanced, AI-powered industrial documentation and diagnostics platform. It bridges the gap between massive, unstructured industrial manuals (P&ID diagrams, spec sheets, safety procedures) and the operational workforce on the ground.

🎯 Purpose of the Project

In heavy industries (oil & gas, manufacturing, chemical plants), operators and maintenance engineers lose hundreds of hours flipping through dense PDFs to find torque specifications, shutdown procedures, or RCA (Root Cause Analysis) history.

DocOps acts as a smart industrial assistant. It ingests these massive documents, vectorizes the knowledge, and provides a real-time, context-aware Copilot. Furthermore, it integrates live IoT telemetry from the plant floor and offers AR barcode scanning, bridging the physical equipment with the digital documentation.

---

✅ Tasks Completed (MVP Level: 100%)

We have successfully built a massive, feature-complete Enterprise MVP (v2.4). The core architecture is fully functional across three tiers: Frontend (Next.js), Gateway (Node.js), and AI Engine (Python/FastAPI).

Key Features Implemented:

1. AI Document Ingestion Pipeline: Extracts text and structures knowledge from PDFs into a local Qdrant Vector Database.
2. Interactive AI Copilot (RAG): Uses LangGraph and Gemini to provide semantic search, compliance verification, and 5-Why RCA based strictly on ingested documents.
3. Live IoT Telemetry Dashboard: Simulates and visualizes live pump (P-101) vibration and temperature data via WebSockets, including critical anomaly alerts.
4. AR Asset Scanner: In-browser camera overlay allowing operators to scan physical QR/Barcodes on machinery to instantly query the AI Copilot.
5. True Offline AI (WebLLM): Browser-based WebGPU Llama-3 integration allowing the LLM to run entirely on the local device without any external network dependency.
6. Air-Gapped Sync: Ability for Plant Managers to export the Qdrant vault as a .zip for deployment on disconnected, ruggedized field tablets.
7. Role-Based Authentication: JWT-based access control distinguishing between operator and plant_manager roles.

---

⚠️ Gaps & Limitations Found

While the MVP is robust, there are several "gaps" representing mock systems or unoptimized paths:

1. Mock IoT Data: The live telemetry is currently simulated via a setInterval in the Node.js Gateway. It needs to be hooked up to an actual MQTT broker or industrial historian (like OSIsoft PI).
2. Local File Storage: The Qdrant database and the JSON user store (data/users.json) are entirely local. This won't scale in a distributed cloud environment.
3. Hardcoded Fallbacks: If the Gemini API key fails, the backend falls back to a mock mode rather than graceful degradation or a localized Python LLM equivalent.
4. WebLLM Initial Load: Downloading the 4GB Llama-3 model to the browser cache takes significant time and bandwidth on the first run, which may hang lower-end devices.

---

🔮 What to Do Next (Roadmap)

1. Database Migration: Replace the local users.json with a managed PostgreSQL instance and migrate Qdrant to Qdrant Cloud.
2. Live Vision Integration: Expand the AR scanner to not just read barcodes, but use computer vision to identify gauge readings (e.g., analog pressure gauges).
3. Dockerization: Wrap the Frontend, Gateway, and Python Backend in a single docker-compose.yml for effortless 1-click deployments on-premise.
4. Agentic Workflows: Allow the Copilot to actually execute actions, such as automatically generating a PDF maintenance ticket and emailing it to the shift supervisor.

---

📖 User Manual

1. Prerequisites

• Node.js (v18+)
• Python (3.10+)
• Gemini API Key

2. Starting the Services

You must start all three services in separate terminals:

Terminal 1: AI Engine (Backend)

cd backend
python -m venv venv
venv\Scripts\activate
pip install -r requirements.txt
python main.py

(Runs on http://localhost:8000)

Terminal 2: Gateway (Middleware)

cd gateway
npm install
node server.js

(Runs on http://localhost:3001)

Terminal 3: Frontend (UI)

cd frontend
npm install
npm run dev

(Runs on http://localhost:3000)

3. Using the App

• Navigate to http://localhost:3000.
• Log in using the test credentials below.
• Go to Documents to upload a PDF manual (e.g., pump maintenance guide).
• Go to the Dashboard to see the live telemetry pulsing.
• Use the Scan Asset button in the sidebar to simulate an AR tag scan.
• Chat with the AI Copilot to extract knowledge from your uploaded PDFs.

---

🔑 Testing Login Credentials

Use the following credentials to access the system:

Role	Username	Password
Plant Manager	admin	admin123
Operator	operator1	ops123

(Note: Plant Managers have access to the "Export Knowledge Vault" and Document Deletion features).