Agentic IT Helpdesk Pipeline

A beginner-friendly Retrieval-Augmented Generation (RAG) and Tool-Calling project.

This project demonstrates how to build an AI support agent that can read a static knowledge base (like an IT policy document) to answer user questions, and use basic tools to interact with a local ticket database.

---

Key Features

• Basic RAG (Retrieval-Augmented Generation): The agent reads from policy.txt using ChromaDB to understand company rules before answering support questions.
• Simple Tool Calling: The agent can execute basic Python functions to interact with a local SQLite database (tickets.db), allowing it to check ticket statuses or log new issues.
• Agent Memory: Uses local storage (agent_storage.db) to remember conversation history and user details across sessions.
• Fast Orchestration: Powered by the Groq Inference API for near-instantaneous LLM responses.

---

Tech Stack

• Framework: Python 3
• AI Orchestration: Agno (formerly Phidata)
• LLM: Llama 3 (via Groq)
• Vector Database: ChromaDB (Local)
• Relational Database: SQLite (Local)
• Package Manager: uv

---

Project Structure

AGENTIC_HELPDESK/
├── policy.txt            # The text file serving as the Knowledge Base for RAG
├── agent.py              # Base agent setup file
├── agent_v2.py           # Iteration 2 of the agent setup
├── final_helpdesk.py     # Finalized agent implementation script
├── finance_agent.py      # Additional test agent for finance queries
├── helpdesk_app.py       # Main application script to run the helpdesk
├── main.py               # Alternative entry point / API testing
├── playground.py         # Agno Playground UI script
├── team_agent.py         # Additional test agent for team queries
├── ticket_system.py      # Python tools (SQL CRUD) for the agent to use
├── pyproject.toml        # Project metadata and uv build configuration
├── uv.lock               # Locked dependencies for perfect reproducibility
├── requirement.txt       # Python dependencies list
└── .env                  # Environment variables (API Keys - Git Ignored)

The vector database(s) will also be stored within this project folder.

Setup & Installation

This project utilizes uv for lightning-fast package management.

1. Clone the Repository

git clone [https://github.com/YOUR_USERNAME/agentic-it-helpdesk.git](https://github.com/YOUR_USERNAME/agentic-it-helpdesk.git)
cd agentic-it-helpdesk

2. Configure Environment Variables

Create a .env file in the root directory. You will need a free API key from Groq.

Ini, TOML

GROQ_API_KEY=gsk_your_groq_api_key_here

3. Install Dependencies

Set up the virtual environment and install the required packages:

Bash

# Using uv (Recommended)
uv venv
source .venv/bin/activate  # On Windows: .venv\Scripts\activate
uv pip install -r requirement.txt

(Note: standard python -m venv venv and pip install -r requirement.txt works as well if you are not using uv).

Usage Guide

1. Run the Main Helpdesk Agent

To interact with the agent in your terminal or via the Agno UI, run the primary application file:

python helpdesk_app.py

2. Test the Playground (UI)

If you want to view the agent in a web browser using the Phidata/Agno playground:

python playground.py

The terminal will provide a localhost link (typically http://localhost:7777). Open this in your browser.

Example Prompts to Try

Once the agent is running, try asking it a combination of knowledge-based and action-based questions:

1. RAG Test: "My laptop screen is broken. What is the IT policy for a replacement?"
2. Tool Test: "Can you check the status of ticket 102?"
3. Combined Test: "Please create a new high-priority ticket for my broken laptop. My manager is John Doe."