A webhook-driven API that automatically generates rich, multi-section travel content for cruises and excursions in Latvian—combining n8n workflow automation with GPT to cut copywriting time by 85%.
Project Details
AI Travel Content Engine is an asynchronous webhook API built for a Baltic travel portal that needed to scale content production across hundreds of cruise itineraries and excursion packages. The system accepts a cruise URL or excursion document (DOCX/PDF), scrapes or parses the source material, and generates up to 15 distinct content sections—hero copy, detailed descriptions, segmented traveler reviews, FAQs, cabin breakdowns, destination guides, pricing tables, and AI-generated imagery—all in publication-ready Latvian.
The architecture is built around n8n workflow automation orchestrating OpenAI GPT calls in a multi-stage pipeline. Each generation request becomes an asynchronous task that processes in parallel, with a polling endpoint for status tracking. Developers can request individual sections or batch multiple sections in a single call (e.g., 'hero,faq,about_cruise'), and pass through structured data like pricing and rules that get woven into the AI-generated narrative. The entire API is documented with an OpenAPI 3.1.1 specification and served via a containerized Scalar API reference portal.
Project Research
Research began by analyzing the client's content bottleneck: their travel portal listed 300+ cruise itineraries and 150+ excursion packages, each requiring hero copy, detailed descriptions, traveler reviews across 5 segments (couples, families, solo travelers, groups, seniors), FAQs, and pricing breakdowns—all in Latvian. A single listing took a copywriter 4–6 hours to produce. At their catalog growth rate, manual production was unsustainable.
We mapped the content taxonomy across cruise lines (Norwegian, MSC, Royal Caribbean) and excursion operators, identifying common structural patterns that could be templatized. We tested GPT's ability to generate culturally appropriate Latvian travel copy, fine-tuning prompts for tone, accuracy, and SEO keyword density. The async architecture was chosen after benchmarking: full multi-section generation averages 25–30 minutes per listing due to sequential AI calls and image generation, making synchronous responses impractical.
Project Results
The AI Travel Content Engine transformed the client's content operations from a manual bottleneck into an automated pipeline. Content production time dropped from 4–6 hours per listing to under 40 minutes of hands-off processing—an 85% reduction. The editorial team shifted from writing content to reviewing and approving AI-generated drafts, increasing their effective throughput by 6x.
Within the first month, the system generated content for 200+ cruise listings and 80+ excursion packages, achieving a 93% first-draft acceptance rate from editors. The multi-section architecture allowed the team to regenerate individual sections without re-running the entire pipeline, saving additional time on revisions. The OpenAPI-first documentation approach enabled two partner agencies to integrate within a week, extending the platform's reach across the Baltic travel market. The containerized Scalar docs portal reduced developer onboarding from days to hours.
