HeASe is a Python library designed to suggest alternative recipes for healthier or more sustainable options. Leveraging machine learning and natural language processing, it compares nutritional profiles and semantic similarities to provide recipe recommendations.
The library consists of several modules:
- Data Preprocessing: Remove unnecessary recipe data.
- Transformer Embeddings: Generates text embeddings for recipe titles using a transformer model.
- Nutritional Vector Space: Maps recipes into a vector space based on nutritional content.
- Similarity Search: Executes cosine similarity searches for finding matching recipes.
- Sort results: Sorts recipes based on healthiness or sustainability metrics.
Install HeASe using pip:
pip install git+https://github.com/xxxxInitializes the system by loading the data and preparing the embeddings.
def __init__(self, recipes_df, nutrients, load=False,
transformer_name='davanstrien/autotrain-recipes-2451975973'):
"""
Initializes the system by loading the data and preparing the embeddings.
:param recipe_df: Datframe containing the recipes.
:param nutrients: List of nutrient names to use.
:param transformer_name: Name of the transformer model to use for embeddings.
"""
from HeASe import HeASe
sm = HeASe(
recipes_df="recipes_df",
load=False,
nutrients=['calories', 'fat', 'protein', ...],
transformer_name='your_transformer_model'
) Default Transformer used davanstrien/autotrain-recipes-2451975973
Use the find_similar_recipes function to get the alternative recipes based on macronutrients similarity based on macronutrients similarity (Step 1 & 2 of the Architecture Diagram) .
def find_similar_recipes(self, input_text, k, acceptable_tags, match_all_tags, check_sustainability=False, j=5):
"""
Finds recipes similar to the given input text.
:param input_text: The input text to find similar recipes for.
:param k: Number of similar recipes to return.
:param acceptable_tags: List of tags considered acceptable for filtering recipes.
:param match_all_tags: Matching strategy
:param check_sustainability: check if the desired recipe is already sustainable
:param j: number of recipes to consider in the centroid computation
:return: A list of tuples with similar recipes and their similarity scores.
"""
similar_recipes = sm.find_similar_recipes(
input_text='Creamy Lemon Asparagus Risotto',
k=10,
acceptable_tags=['appetizers', 'main-dish', 'side-dishes', 'fruits', 'desserts',
'breakfast', 'pasta-rice-and-grains', 'beverages', 'drinks'],
match_all_tags=True
)Use the order_recipe_by_healthiness to sort the recipes (Step 3 of the Architecture Diagram) .
def order_recipe_by_healthiness(self, nearest_recipes=None, score='who_score'):
"""
Order the recipes obtained previously.
:param (optional) nearest_recipes: Dataframe to order, if none the dataframe computed by find_similar_recipes will be used.
:param score: The column name used as the primary sorting criterion.
:return: A DataFrame of recipes ordered by the specified score.
"""
- Sorts the DataFrame produced by 'find_similar_recipes', assuming it has been executed before.
healthier_recipes = sm.order_recipe_by_healthiness(score="who_score")- Sort a different dataframe than the one computed by 'find_similar_recipes' (for example is useful for reordering a subset of the previous dataframe )
healthier_recipes = sm.order_recipe_by_healthiness(
nearest_recipes=your_df,
score="who_score"
)Use the order_recipe_by_sustainability to sort the recipes (Step 3 of the Architecture Diagram) .
def order_recipe_by_sustainability(self, nearest_recipes=None, score='sustainability_label',
secondary_sort_field='who_score'):
"""
Order the recipes obtained previously.
:param (optional) nearest_recipes: Dataframe to order , if none the dataframe computed by find_similar_recipes will be used.
:param score: The column name used as the primary sorting criterion.
:param secondary_sort_field: The column name used as the secondary sorting criterion.
:return: A Dataframe with recipes ordered by the given metric.
"""
- Sorts the DataFrame produced by 'find_similar_recipes', assuming it has been executed before.
order_by_sus_recipes = sm.order_recipe_by_sustainability(score='sustainability_label')- Sort a different dataframe than the one computed by 'find_similar_recipes' (for example is useful for reordering a subset of the previous dataframe )
order_by_sus_recipes = sm.order_recipe_by_sustainability(
nearest_recipes=your_df,
score="sustainability_label"
)Use the order_recipe_by_HeASe to sort the recipes.
def order_recipe_by_HeASe(self, nearest_recipes=None, alpha=0.7, beta=0.3):
"""
Order the recipes obtained previously.
:param (optional) nearest_recipes: Dataframe to order , if none the dataframe computed by find_similar_recipes will be used.
:return: A Dataframe with recipes ordered by the given metric.
:param alpha: weight for sustainability score
:param beta: weight for healthiness score
"""
- Sorts the DataFrame produced by 'find_similar_recipes', assuming it has been executed before.
order_by_sus_recipes = sm.order_recipe_by_HeASe()- Sort a different dataframe than the one computed by 'find_similar_recipes' (for example is useful for reordering a subset of the previous dataframe )
order_by_sus_recipes = sm.order_recipe_by_HeASe(
nearest_recipes=your_df)Use an LLM to choose the best recipe from a list ordered by HeASe score.
def choose_best_recipe_with_gpt(self, custom_prompt=None, nearest_recipes=None, alpha=0.7, beta=0.3):
"""
:param custom_prompt: Custom prompt to override the first section of the prompt
:param nearest_recipes: Optional DataFrame of recipes to order. If None, use the DataFrame from find_similar_recipes.
:param alpha: Weight for sustainability score.
:param beta: Weight for healthiness score.
:return: The name of the best recipe chosen by GPT-3.5.
"""
sm.setup_key("YOUR_OPEN_AI_KEY")
sm.choose_best_recipe_with_gpt(custom_prompt):Custom prompt will override the first instructions
if custom_prompt is not None:
prompt = custom_prompt
else:
prompt = "Using your knowledge please rank (if necessary) the following recipes from most to least recommended based on a balance of sustainability and healthiness:\n\n"
prompt += "\n".join([
f"{idx + 1}. Recipe: {row['title']}"
for idx, row in ordered_recipes.iterrows()])
prompt += "\n\nWhich one should I choose? Return just the name"It is possible to initialize a primordial version of an agent
sm.setup_key("YOUR_OPEN_AI_KEY")
sm.create_agent()
sm.agent_ask("Can you suggest an alternative recipe to Lasagna?")The HeASe Command Line Interface (CLI) provides an interface to perform various operations on your HeASe library. Below are the different operations available and how to use them.
HeASe --loadThis command loads processed data from previously saved files.
HeASe --recipes_csv path/to/your/file.csvThis command loads data from a CSV file containing recipes.
HeASe find_similar RECIPE_NAME --k 5 --acceptable_tags tag1 tag2 --match_all_tags --check_sustainabilityFind recipes similar to the specified one, with options to limit results based on tags, match all tags, and check sustainability.
HeASe order_by_healthiness RECIPE_NAME --k 5 --acceptable_tags tag1 tag2 --match_all_tags --check_sustainability --score who_scoreOrder recipes by healthiness using a specific score.
HeASe order_by_sustainability RECIPE_NAME --k 5 --acceptable_tags tag1 tag2 --match_all_tags --check_sustainability --score sustainability_score --secondary_sort_field who_scoreOrder recipes by sustainability using a specific primary score and a secondary sort field.
HeASe order_by_HeASe RECIPE_NAME --k 5 --acceptable_tags tag1 tag2 --match_all_tags --check_sustainability --alpha 0.7 --beta 0.3Order recipes using the HeASe score with specific weights for sustainability and healthiness.
HeASe --loadHeASe find_similar "Pasta with Tomato" --k 5 --acceptable_tags main-dish --check_sustainabilityHeASe order_by_sustainability "Quinoa Salad" --k 5 --acceptable_tags side-dishes --check_sustainability --score sustainability_score --secondary_sort_field who_score