Recipe Recall

This was the result of a one-hour AI hackathon organised by Steel-AI (not including the 3 hours of debugging and documentation afterwards!).

Don’t you just hate the time it takes to write a food shopping list every week to make the recipes you want to make? Use this project if you want hallucination-free ingredients lists from a reliable source called directly from your preferred LLM. Also refer to this for inspiration for your own agentic AI tools configured using a locally hosted MCP server.

Python scraper and fastmcp script built with help from GPT-5 mini, switching to Claude Haiku 4.5 later on (which also helped build the first draft of this readme). MCP server tested on both Qwen3:8b and Llama 3.2:3b locally hosted via Ollama.

Implementation summary

1. bbcgoodfood_scraper.py

   • Standalone Python script that searches BBC Good Food recipes
   • Extracts ingredients from ul.ingredients-list.list elements
   • Parses ingredients into quantity/unit/ingredient_name components
   • Handles text formatting quirks (e.g., “50gdried” → “50g dried”)
   • Fully functional when run directly

2. fastmcp_quickstart.py

   • MCP server exposing scraper and utilities as callable tools
   • Tools: add_note, add, test_network, search_recipes
   • Uses stdio transport for subprocess communication
   • File logging to recipe_recall.log with timestamps (on the function-in-tool branch)