Intelligent Hospitality Agent for Personalized Guest Services
This challenge involves building an advanced agentic system designed to enhance guest experiences and streamline hotel operations. You will leverage the OpenAI Agents SDK to create a multi-faceted AI assistant capable of personalized guest interactions, dynamic service recommendations, and backend system automation. The agent will manage guest requests, integrate with existing hotel management systems (simulated), and learn from guest preferences stored in a vector database to provide truly bespoke service. This project emphasizes modern agent design, tool utilization, memory management, and robust observability for production-grade AI.
What you are building
The core problem, expected build, and operating context for this challenge.
Develop an intelligent, tool-using agent with OpenAI Agents SDK to provide personalized hospitality services, integrating memory and workflow automation.
How work is evaluated
The agent will be evaluated on its ability to accurately understand and fulfill complex guest requests, its effective use of tools and memory, and the overall coherence and personalization of its responses.
Challenge datasets
Public datasets, plus the private datasets you have access to.
How submissions are scored
These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.
Correct tool usage
Agent must correctly identify and call the necessary tools for a given request.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
Personalization via Milvus
Agent must demonstrate retrieval and use of personalized information from Milvus in at least 70% of relevant scenarios.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
Request Fulfillment Accuracy
Percentage of requests correctly understood and acted upon by the agent (0-1). • target: 0.9 • range: 0.75-1
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
Response Personalization
Score reflecting how well responses are tailored using guest data (0-1). • target: 0.85 • range: 0.7-1
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
Langfuse Trace Completeness
Percentage of agent interactions fully traceable in Langfuse (0-1). • target: 1 • range: 0.9-1
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Master the OpenAI Agents SDK for building robust, autonomous agents, focusing on defining tools, assistant configuration, and thread management for multi-turn conversations.
Implement long-term memory for the agent using Milvus, storing and retrieving vectorized guest preferences, past requests, and feedback for personalized interactions.
Design and register custom tools (functions) within the OpenAI Agents SDK for simulating hotel operations like 'book_restaurant', 'request_concierge_service', 'check_in_guest', and 'retrieve_guest_profile'.
Utilize Adept's workflow automation capabilities (simulated through API calls) to trigger backend actions based on agent decisions, such as updating booking systems or scheduling staff tasks.
Orchestrate complex agent behaviors to handle nuanced guest requests, involving multiple tool calls, information synthesis from Milvus, and decision-making by the OpenAI o3 model.
Integrate Langfuse for end-to-end observability, tracing agent decisions, tool calls, LLM inputs/outputs, and intermediate thoughts, enabling effective debugging and performance monitoring.
Reference links and supporting material
Sample data for 2 tasks
Challenge setup
The agent will be evaluated on its ability to accurately understand and fulfill complex guest requests, its effective use of tools and memory, and the overall coherence and personalization of its responses.
Challenge input
{ "guest_profile": { "id": "string", "preferences": ["string"], "past_stays": ["string"] }, "request": "string" }
Agent execution
No agent tool is listed for this challenge.
Evaluated output
{ "agent_actions": [{"tool": "string", "args": {}}, ...], "final_response": "string", "personalization_score": float }
- Agent must correctly identify and call the necessary tools for a given request.
- Agent must demonstrate retrieval and use of personalized information from Milvus in at least 70% of relevan...
- Percentage of requests correctly understood and acted upon by the agent (0-1). • target: 0.9 • range: 0.75-1
- Request Fulfillment Accuracy target: 0.9
- 2 public reference cases
- Python execution harness
Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.
View technical recipe
Configured tools
No tool records are attached.
Evaluation contract
- Correct tool usage · Weight 1
- Personalization via Milvus · Weight 1
- Request Fulfillment Accuracy · Weight 1
- Response Personalization · Weight 1
- Langfuse Trace Completeness · Weight 1
Metrics
- Request Fulfillment Accuracy · Target 0.9
- Response Personalization · Target 0.85
- Langfuse Trace Completeness · Target 1
Reference cases
- 2 public reference cases
Workflow tasks
- Task 1
Input: { "guest_profile": { "id": "string", "preferences": ["string"], "past_stays": ["string"] }, "request": "string" }
Output: { "agent_actions": [{"tool": "string", "args": {}}, ...], "final_response": "string", "personalization_score": float }
Execution harness: Python
- Task 2
Input: { "guest_profile_id": "string", "query": "string" }
Output: { "retrieved_info": ["string", ...], "synthesized_response": "string", "relevance_score": float }
Execution harness: Python
Recipe freeze marker
No recipe freeze time is recorded.