Recipe Agent

Agent that suggests recipes from the pantry list from a specified file such as files/ingredients.txt or a specified ingredients such as find me chicken and rice recipes

Project Description

Key Features

• Import pantry list or specify ingredients in prompt
• Searches recipes on google based on key ingredients
• Extracts ingredients from recipe website
• Calculates matching and missing ingredients
• Summarises ingredients you have (core ingredients) and ingredients you need (other ingredients)

Architecture

Agents

• input_agent – resolves file path vs free text
• main_ingredients_agent – extracts key ingredients from user text for search
• recipe_search_agent – finds recipe URLs via Google search based on key ingredients
• url_critic_agent – validates search agent URLs and can call exit_loop
• recipe_verification_loop – LoopAgent that iterates search and critiques search
• recipe_ingredient_agent – fetches the recipe page and extracts ingredients
• recipe_ranking_agent – compares user ingredients vs recipe ingredients
• display_agent - summarises the data as markdown output for the user

Tools

• resolve_input – extracts user input when it is a file path
• get_match_outcomes – computes matched and missing ingredients by comparing sets
• google_search
• fetch_webpage_content – fetches page HTML/text from a URL

Examples

See Examples Here

Example Ingredient lists

Setup

Prerequisites: Python, Pip

Install the ADK package

pip install google-adk

Create and activate a virtual environment from project root directory

python -m venv .venv && source .venv/bin/activate On Windows: .venv\Scripts\activate)

Install dependencies

pip install -r requirements.txt

Configure API key

You can manage your API key on google cloud studio here

cd app

echo "GOOGLE_GENAI_USE_VERTEXAI=True" >> app/.env

echo "GOOGLE_API_KEY=YOUR_AI_STUDIO_API_KEY" >> app/.env

Your API key should be in the .env file in app/

Run agent

From root directory

Run Locally

adk run app

In web ui

adk web --port 8000

Makefile

Refer to makefile for further commands

Further Resources

Project Link

Docs reference

Video demo

Future Features

• Run pipeline in parallel to suggest and rank multiple recipes.
• Store recipe text content in model context to allow users to ask follow up questions on recipe.
• Deploy to API with JSON output for application integration.
• Integrate agent with a shopping list application.