HelloFresh Meal Recommendation Agent with Qdrant

Home cooks and meal planners using HelloFresh who want quick, personalized recipe suggestions from the weekly menu based on tastes, dietary needs, or available ingredients. This workflow scrapes HelloFresh's weekly menu, extracts and embeds recipe data into a Qdrant vector store for semantic search, stores full details in SQLite, and powers an AI chat agent that recommends recipes using user preferences via Qdrant's recommend API with positive/negative queries. 1. Create Qdrant collection 'hello_fresh' with 1024-dim vectors (Cosine distance). 2. Set up Mistral Cloud and Qdrant API credentials in n8n. 3. Ensure SQLite DB 'hello_fresh\_1.db' is accessible via Code nodes. 4. Activate workflow and use Chat Trigger for interactions or Manual Trigger to populate data. n8n instance with LangChain nodes; Mistral Cloud API key; Qdrant server (local or cloud); Python in Code nodes for SQLite; Internet access for scraping HelloFresh. Modify AI Agent's system prompt for tone or rules; Adjust embedding model or chunk size in splitter; Add more filters in Qdrant queries; Extend tool schema for additional user inputs like budget or time. Key Technologies: n8n Workflow Automation,HelloFresh Web Scraping (HTTP + HTML Extract),Mistral AI Embeddings and Chat Model,Qdrant Vector Database,SQLite for Structured Storage,LangChain Integration for AI Agent and Tools Automation Value: Automates weekly menu fetching to keep recommendations fresh without manual checks. Enables semantic recipe matching beyond keywords, improving personalization (e.g., 'spicy vegetarian' avoids non-matches). Reduces decision fatigue by providing tailored suggestions in chat, saving users 10-20 minutes per meal plan. Scales to handle preferences like allergens or cuisine themes, enhancing user satisfaction. Builds a reusable RAG system for future expansions like nutritional analysis or shopping lists. Best Practices: Use sticky notes for documentation to guide setup and troubleshooting. Implement rate limiting and waits to handle API constraints gracefully. Filter searches by week to keep recommendations relevant and avoid stale data.