LLM Itinerary Logic Optimization
Budget: $15 – $25 USD
I run an AI-driven travel planner that talks to travellers through OpenRouter and returns day-by-day itineraries. The conversational layer works, but the moment a user adds a third or fourth city the plan starts to drift—duplicate attractions, impossible transfers, or dropped preferences. I want that fixed.
Your focus will be the itinerary-generation logic itself. I need every response to stay accurate while juggling multiple destinations, dates, and personal tastes. To get there you’ll refine prompts, extend our existing JSON schema, and wire the whole flow back into our REST API so the conversation remains seamless.
Key deliverables
• Prompt templates and system messages that reliably handle multi-city trips
• A concise, validated JSON schema for itinerary objects (cities, days, transport, costs)
• Revised inference flow integrated with OpenRouter endpoints (Python/TypeScript SDKs are in place)
• Quick test harness or Postman collection that proves accuracy gains against sample conversations
Acceptance criteria
An itinerary request covering at least three distinct cities over seven days must return:
✔ No conflicting dates or duplicate sights
✔ Travel segments under the user’s stated time limits
✔ Consistent JSON matching the agreed schema, ready for frontend rendering
Experience in prompt engineering, schema design, and API integration is essential. If you’ve already solved multi-destination travel, even better—share a quick example in your bid and let’s get this planner truly smart.
Your focus will be the itinerary-generation logic itself. I need every response to stay accurate while juggling multiple destinations, dates, and personal tastes. To get there you’ll refine prompts, extend our existing JSON schema, and wire the whole flow back into our REST API so the conversation remains seamless.
Key deliverables
• Prompt templates and system messages that reliably handle multi-city trips
• A concise, validated JSON schema for itinerary objects (cities, days, transport, costs)
• Revised inference flow integrated with OpenRouter endpoints (Python/TypeScript SDKs are in place)
• Quick test harness or Postman collection that proves accuracy gains against sample conversations
Acceptance criteria
An itinerary request covering at least three distinct cities over seven days must return:
✔ No conflicting dates or duplicate sights
✔ Travel segments under the user’s stated time limits
✔ Consistent JSON matching the agreed schema, ready for frontend rendering
Experience in prompt engineering, schema design, and API integration is essential. If you’ve already solved multi-destination travel, even better—share a quick example in your bid and let’s get this planner truly smart.