🤖 Veridian Corp AI IT Support Agent

An intelligent, agentic IT support system designed for Veridian Corp. This project demonstrates a professional, full-stack AI agent capable of understanding employee issues, retrieving relevant company policies, making reasoned decisions based on historical precedents, and managing an integrated IT ticketing system.

🚀 Project Overview

The system acts as an automated first-line IT support engineer. Unlike a standard chatbot, it employs an Agentic Workflow to ensure that every response is grounded in company policy and consistent with previous decisions.

🔄 Agentic Workflow

Employee Request $\rightarrow$ Intent Classification $\rightarrow$ Hybrid Policy Retrieval $\rightarrow$ Reasoning (with History) $\rightarrow$ Decision $\rightarrow$ Resolution / Escalation.

✨ Key Features

• Natural Language Understanding: Automatically classifies user requests into specific IT categories using LLM-based intent detection.
• Grounded Reasoning (RAG): Implements a Retrieval-Augmented Generation (RAG) pattern using a curated Knowledge Base (KB) to prevent AI hallucinations.
• Context-Aware Decisions: Analyzes historical ticket data to maintain consistency in approvals, rejections, and resolution paths.
• Secure Admin Dashboard: A protected management area for IT admins to track tickets, review full audit trails, and manage policy grounding.
• Automated Ticketing: Intelligently decides when a request can be self-serviced vs. when it requires a structured IT ticket for human intervention.
• Professional Full-Stack UI: Modern web frontend built with HTML, CSS, and JavaScript, powered by a high-performance FastAPI backend.

---

📁 File Structure & Component Map

⚙️ Backend (Python/FastAPI)

File	Description
main.py	Entry Point. FastAPI server handling routing, static file serving, and API endpoints for support and admin management.
agent.py	The Brain. Orchestrates the AI workflow. Handles LLM-based classification, decision-making, and response generation via the Ollama API.
retriever.py	Knowledge Fetcher. Implements a hybrid retrieval strategy (Category Mapping $\rightarrow$ Keyword Scoring) to fetch the most relevant policy from policies.json.
database.py	Data Layer. Manages the SQLite database (veridian_it.db) for tickets, audit logs, and system settings.
auth_utils.py	Security. Provides secure password hashing and verification using the bcrypt library.
set_admin_password.py	Setup Utility. Standalone script to securely initialize or update the admin dashboard password.
seed_tickets.py	Data Utility. Populates the database with historical ticket data from tickets.json to provide context for the AI.

📚 Data & Configuration

File	Description
policies.json	Knowledge Base. The "Ground Truth" containing official Veridian Corp IT policies (KB-01 to KB-10).
tickets.json	Historical Data. A dataset of past tickets used to seed the database and provide precedent for the AI's reasoning.
requirements.txt	Dependencies. Required Python libraries (FastAPI, Uvicorn, Requests, Bcrypt, Python-Dotenv).
.env	Environment Variables. Configuration for the Ollama API (Base URL, Model, and API Key).

🎨 Frontend (Static Assets)

File	Description
static/index.html	UI Structure. The main interface featuring the chat portal and the Admin Dashboard.
static/style.css	Styling. Modern corporate aesthetic with a responsive layout.
static/script.js	Frontend Logic. Handles API communication, session management, and dynamic DOM updates.

---

🛠️ Installation & Setup

1. Clone and Install

# Install required dependencies
pip install -r requirements.txt

2. Configure AI Model

Create a .env file in the root directory. The system defaults to llama3 via Ollama:

OLLAMA_BASE_URL=http://localhost:11434
OLLAMA_API_KEY=your_api_key_here (optional)
OLLAMA_MODEL=llama3

3. Initialize System

Before starting the server, you must set the admin password and seed the database with historical context:

# Set the password for the Admin Dashboard
python set_admin_password.py

# Populate the database with historical precedents
python seed_tickets.py

4. Run the Server

python main.py

Visit http://127.0.0.1:8000 in your browser to access the portal.

---

🛡️ Admin Access & API

Admin Dashboard

To manage the system:

1. Navigate to the Admin tab in the web portal.
2. Enter the password created during the setup step.
3. Manage Tickets, review the Audit Trail, and browse the Knowledge Base.

Key API Endpoints

Endpoint	Method	Description	Access
/api/support/request	POST	Submit a support request to the AI Agent	Public
/api/support/my-tickets	GET	Get tickets for a specific employee email	Public
/api/admin/login	POST	Authenticate admin and receive session token	Public
/api/support/tickets	GET	List all system tickets	Admin
/api/support/audit	GET	Retrieve the full agent audit trail	Admin
/api/support/policies	GET	Retrieve the current knowledge base	Admin