Luna AI Shopping Assistant Development
Budget: $8 – $15 USD
I have the core SmartGuide AI Agent & Management Suite running—modern UI, admin dashboard, asynchronous API engine, and the Model Context Protocol server are already in place. What I now need is for you to bring the personalized shopping-assistance layer to life, with Luna—the enthusiastic smartphone advisor—as the primary conversational persona.
Your mandate is to wire Luna into the existing stack so she can:
• Offer context-aware product recommendations drawn from our smartphone catalogue
• Query the live stock service in real time
• Create and confirm customer booking reservations on the spot
The conversational logic must leverage Generative AI with Retrieval-Augmented Generation; our product data, benchmark feeds, and availability endpoints are already exposed through a FastAPI gateway. You will design the prompt strategy, RAG pipeline, and any embedding or vector-store choices, then connect them to the MCP node so Luna stays reactive and persona-consistent.
Acceptance criteria
1. Recommendation accuracy ≥ 90 % against our evaluation set
2. Stock checks resolve in under two seconds end-to-end
3. Confirmed bookings appear correctly in the reservations table and trigger the existing webhook
4. Deployment scripts (Docker/K8s) and unit tests cover the new services at ≥ 80 %
All code should be clean Python (or TypeScript if you prefer for the UI hook-ins) with clear README instructions. I’ll provide API keys, schema docs, and a staging cluster the moment we kick off. Let me know your approach to persona design, RAG configuration, and how you plan to measure accuracy so we can align quickly and move into implementation.
Your mandate is to wire Luna into the existing stack so she can:
• Offer context-aware product recommendations drawn from our smartphone catalogue
• Query the live stock service in real time
• Create and confirm customer booking reservations on the spot
The conversational logic must leverage Generative AI with Retrieval-Augmented Generation; our product data, benchmark feeds, and availability endpoints are already exposed through a FastAPI gateway. You will design the prompt strategy, RAG pipeline, and any embedding or vector-store choices, then connect them to the MCP node so Luna stays reactive and persona-consistent.
Acceptance criteria
1. Recommendation accuracy ≥ 90 % against our evaluation set
2. Stock checks resolve in under two seconds end-to-end
3. Confirmed bookings appear correctly in the reservations table and trigger the existing webhook
4. Deployment scripts (Docker/K8s) and unit tests cover the new services at ≥ 80 %
All code should be clean Python (or TypeScript if you prefer for the UI hook-ins) with clear README instructions. I’ll provide API keys, schema docs, and a staging cluster the moment we kick off. Let me know your approach to persona design, RAG configuration, and how you plan to measure accuracy so we can align quickly and move into implementation.