Building an AI-Powered HR Assistant with OpenAI Agents SDK
Introduction
In the era of digital transformation, automating HR queries and processes can significantly enhance employee experience and operational efficiency. In this project, I built an intelligent HR Assistant using the OpenAI Agents SDK that can handle a variety of HR-related queries, triage requests, and even trigger backend actions such as leave requests or complaint registration. The assistant features a conversational UI built with Gradio, making it accessible and user-friendly.
Key Features
- Multi-Agent Architecture: Specialized agents handle different HR domains (leaves, benefits, policies, complaints).
- Guardrails: Input guardrails ensure only HR-related queries are processed, improving safety and relevance.
- Tool Integration: Backend functions (tools) are registered with agents, allowing the assistant to perform actions like creating leave requests or registering complaints.
- Context Management: The assistant maintains conversation history, enabling context-aware, multi-turn conversations.
- User Interface: A Gradio-based web UI provides a seamless chat experience.
Project Structure & Important Aspects
1. Agent Design
Each HR domain is managed by a dedicated agent:
- Leaves Agent: Handles leave policies and requests.
- Benefits Agent: Answers questions about employee benefits.
- Policy Agent: Provides information on company policies.
- Complaints Agent: Registers and tracks complaints.
A Triage Agent routes user queries to the appropriate specialist agent.
2. Guardrails
A guardrail agent checks if the user’s query is HR-related before passing it to the triage agent, ensuring the assistant stays on topic.
Tools are Python functions registered with agents. For example:
@function_tool
def create_leave_request(employee_id: str, department: str, leave_type: str, start_date: str, end_date: str):
return f"Leave request created for {employee_id} in {department} from {start_date} to {end_date} for {leave_type}."
Agents invoke these tools when appropriate, enabling real actions beyond just answering questions.
4. Context Management
The assistant maintains a persistent conversation history, which is included in each prompt to ensure context continuity across multiple turns. This is crucial for handling follow-up questions and multi-step workflows.
5. Gradio UI
A simple yet effective chat interface was built using Gradio, allowing users to interact with the HR assistant in real time.
Example Queries
- “I need to take a leave next week.”
- “What is the company’s life insurance policy?”
- “How many sick leaves do I have left?”
- “I want to register a complaint about my manager.”
- “Tell me about the gym membership benefit.”
Demo
Try it Yourself
The full project is available on GitHub:
GitHub Repository
Demo Queries
Try these in the chat UI:
- “Can I get parental leave?”
- “How do I register a complaint?”
- “What are the food coupon benefits?”
- “Tell me about the weekend shift policy.”
- “I want to apply for sick leave from tomorrow.”
Conclusion
This project demonstrates how the OpenAI Agents SDK can be leveraged to build robust, context-aware, and action-oriented assistants for enterprise use cases. With modular agents, tool integration, and a user-friendly UI, the HR Assistant is a step towards smarter, more efficient HR operations.
For more details and to contribute, visit the GitHub repository.