Local Attraction Recommender Workflow

This n8n workflow is designed to create a sophisticated system for recommending local attractions based on user input. It integrates webhooks, language models, vector similarity search, and data logging to deliver personalized suggestions and insights. The process begins with a webhook trigger, allowing users to submit queries or preferences. The input is then processed and split into manageable text chunks using a character-based text splitter.

These chunks are transformed into embeddings using the Cohere API, enabling semantic understanding. The embeddings are stored in a Pinecone vector database for efficient similarity searches. When a user makes a new query, the workflow searches relevant vectors in Pinecone to find similar attractions or information.

The retrieved data is further analyzed by a language model (using the Anthropic API) to generate detailed, human-like responses or recommendations. An agent node defines the task to extract meaningful insights or define the query. Finally, the workflow logs all interactions and generated responses into a Google Sheet for record-keeping and analysis.

This workflow is particularly useful for tourism businesses, travel websites, or local guides aiming to provide personalized attraction recommendations and maintain a log of user interactions for continual improvement.