A full-stack travel itinerary application with FastAPI backend and Next.js frontend for managing travel plans including accommodations, transfers, and activities
Server has 4 tools with basic schemas and descriptions, but significant gaps in parameter documentation and error handling. Tool names follow verb_noun convention (get_*, create_*), which is good. However, parameter descriptions are minimal or missing entirely. The create_itinerary tool has a complex nested schema but lacks validation guidance. No error recovery hints, no permission gates, no audit logging. Output schemas are partially documented via Pydantic models but not explicitly described in tool definitions. This is a typical community server with functional but incomplete definitions.
Create a new itinerary with accommodations, transfers, and activities
Get all itineraries with pagination support
Get a specific itinerary by ID
Get recommended itineraries for a specific duration (2-8 nights)
create_itinerary has complex nested schema (accommodations, transfers, activities arrays with nested objects) but lacks validation rules, format constraints, and error guidance. No description of what happens on validation failure or how to correct malformed input.
Parameter descriptions are minimal or absent. 'skip' and 'limit' in get_itineraries lack guidance on valid ranges; 'itinerary_id' in get_itinerary has no hint about what to do if not found. LLMs cannot infer parameter semantics from names alone.
No error handling or recovery guidance. get_itinerary returns 404 if not found, but tool description does not tell LLM what to do next (e.g., 'Try get_itineraries() to discover available IDs'). No actionable error messages.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 67 | 2026-07-28+ | v2 |
create_itinerary is a WRITE operation but has no confirmation step, dry-run option, or idempotency guarantee. Agents cannot safely retry on ambiguous failures without risking duplicate itineraries.
Output schemas are implicit (Pydantic models) and not explicitly documented in tool definitions. LLMs cannot see what fields to expect from responses, forcing them to guess or make extra discovery calls.