Orchestrate Cohere LLMs with Apache Airflow

Info

This page has not yet been updated for Airflow 3. The concepts shown are relevant, but some code may need to be updated. If you run any examples, take care to update import statements and watch for any other breaking changes.

Cohere is a natural language processing (NLP) platform that provides an API to access cutting-edge large language models (LLMs). The Cohere Airflow provider offers modules to easily integrate Cohere with Airflow.

In this tutorial, you use Airflow and the Cohere Airflow provider to generate recipe suggestions based on a list of ingredients and countries of recipe origin. Additionally, you create embeddings of the recipes and perform dimensionality reduction using principal component analysis (PCA) to plot recipe similarity in two dimensions.

Why use Airflow with Cohere?

Cohere provides highly specialized out-of-the box and custom LLMs. Countless applications use these models for both user-facing needs, such as to moderate user-generated content, and internal purposes, like providing insight into customer support tickets.

Integrating Cohere with Airflow into one end-to-end machine learning pipeline allows you to:

• Use Airflow’s data-driven scheduling to run operations with Cohere LLM endpoints based on upstream events in your data ecosystem, like when new user input is ingested or a new dataset is available.
• Send several requests to a model endpoint in parallel based on upstream events in your data ecosystem or user input with Airflow params.
• Add Airflow features like retries and alerts to your Cohere operations. This is critical for day 2 MLOps operations, for example, to handle model service outages.
• Use Airflow to orchestrate the creation of vector embeddings with Cohere models, which is especially useful for very large datasets that cannot be processed automatically by vector databases.

Time to complete

This tutorial takes approximately 15 minutes to complete (cooking your recommended recipe not included).

Assumed knowledge

To get the most out of this tutorial, make sure you have an understanding of:

• The basics of the Cohere API. See Cohere Documentation.
• The basics of vector embeddings. See the Cohere Embeddings guide.
• Airflow fundamentals, such as writing DAGs and defining tasks. See Get started with Apache Airflow.
• Airflow operators. See Operators 101.
• Airflow hooks. See Hooks 101.

Prerequisites

• The Astro CLI.
• A Cohere API key. You can generate an API key in the Cohere dashboard, accessible with a Cohere account. A free tier API key is sufficient for this tutorial.

Step 1: Configure your Astro project

1. Create a new Astro project:

   1
   $ mkdir astro-cohere-tutorial && cd astro-cohere-tutorial
   2
   $ astro dev init

2. Add the following lines to your requirements.txt file to install the Cohere Airflow provider and other supporting packages:

   apache-airflow-providers-cohere==1.0.0
   matplotlib==3.8.1
   seaborn==0.13.0
   scikit-learn==1.3.2
   pandas==1.5.3
   numpy==1.26.2
   adjustText==0.8

3. To create an Airflow connection to Cohere, add the following environment variables to your .env file. Make sure to provide <your-cohere-api-key>.

   AIRFLOW_CONN_COHERE_DEFAULT='{
       "conn_type": "cohere",
       "password": "<your-cohere-api-key>"
   }'

Step 2: Create your DAG

1. In your dags folder, create a file called recipe_suggestions.py.

2. Copy the following code into the file.

   1
   """
   2
   ## Get recipe suggestions using Cohere's LLMs, embed and visualize the results
   3
   4
   This DAG shows how to use the Cohere Airflow provider to interact with the Cohere API.
   5
   The DAG generates recipes based on user input via Airflow params, embeds the
   6
   responses using Cohere embeddings, and visualizes them in 2 dimensions using PCA,
   7
   matplotlib and seaborn.
   8
   """
   9
   10
   from airflow.decorators import dag, task
   11
   from airflow.models.param import Param
   12
   from airflow.models.baseoperator import chain
   13
   from airflow.providers.cohere.hooks.cohere import CohereHook
   14
   from airflow.providers.cohere.operators.embedding import CohereEmbeddingOperator
   15
   from sklearn.metrics.pairwise import euclidean_distances
   16
   from sklearn.decomposition import PCA
   17
   from adjustText import adjust_text
   18
   from pendulum import datetime
   19
   import matplotlib.pyplot as plt
   20
   import seaborn as sns
   21
   import pandas as pd
   22
   import numpy as np
   23
   24
   25
   COHERE_CONN_ID = "cohere_default"
   26
   IMAGE_PATH = "include/recipe_plot.png"
   27
   28
   29
   @dag(
   30
       start_date=datetime(2023, 11, 1),
   31
       schedule=None,
   32
       catchup=False,
   33
       params={
   34
           "countries": Param(
   35
               ["Switzerland", "Norway", "New Zealand", "Cameroon", "Bhutan", "Chile"],
   36
               type="array",
   37
               title="Countries of recipe origin",
   38
               description="Enter from which countries you would like to get recipes."
   39
               + "List at least two countries.",
   40
           ),
   41
           "pantry_ingredients": Param(
   42
               ["gruyere", "olives", "potatoes", "onions", "pineapple"],
   43
               type="array",
   44
               description="List the ingredients you have in your pantry, you'd like to use",
   45
           ),
   46
           "type": Param(
   47
               "vegetarian",
   48
               type="string",
   49
               enum=["vegan", "vegetarian", "omnivore"],
   50
               description="Select the type of recipe you'd like to get.",
   51
           ),
   52
           "max_tokens_recipe": Param(
   53
               500,
   54
               type="integer",
   55
               description="Enter the max number of tokens the model should generate.",
   56
           ),
   57
           "randomness_of_recipe": Param(
   58
               25,
   59
               type="integer",
   60
               description=(
   61
                   "Enter the desired randomness of the recipe on a scale"
   62
                   + "from 0 (no randomness) to 50 (full randomness). "
   63
                   + "This setting corresponds to 10x the temperature setting in the Cohere API."
   64
               ),
   65
           ),
   66
       },
   67
   )
   68
   def recipe_suggestions():
   69
       @task
   70
       def get_countries_list(**context):
   71
           "Pull the list of countries from the context."
   72
           countries = context["params"]["countries"]
   73
           return countries
   74
   75
       @task
   76
       def get_ingredients_list(**context):
   77
           "Pull the list of ingredients from the context."
   78
           ingredients = context["params"]["pantry_ingredients"]
   79
           return ingredients
   80
   81
       @task
   82
       def get_a_recipe(
   83
           cohere_conn_id: str, country: str, ingredients_list: list, **context
   84
       ):
   85
           "Get recipes from the Cohere API for your pantry ingredients for a given country."
   86
           type = context["params"]["type"]
   87
           max_tokens_answer = context["params"]["max_tokens_recipe"]
   88
           randomness_of_answer = context["params"]["randomness_of_recipe"]
   89
           co = CohereHook(conn_id=cohere_conn_id).get_conn
   90
   91
           response = co.generate(
   92
               model="command",
   93
               prompt=f"Please provide a delicious {type} recipe from {country} "
   94
               + f"that uses as many of these ingredients: {', '.join(ingredients_list)} as possible, "
   95
               + "if you can't find a recipe that uses all of them, suggest an additional desert."
   96
               + "Bonus points if it's a traditional recipe from that country, "
   97
               + "you can name the city or region it's from and you can provide "
   98
               + "vegan alternatives for the ingredients."
   99
               + "Provide the full recipe with all steps and ingredients.",
   100
               max_tokens=max_tokens_answer,
   101
               temperature=randomness_of_answer / 10,
   102
           )
   103
   104
           recipe = response.generations[0].text
   105
   106
           print(f"Your recipe from {country}")
   107
           print(f"for the ingredients {', '.join(ingredients_list)} is:")
   108
           print(recipe)
   109
   110
           with open(f"include/{country}_recipe.txt", "w") as f:
   111
               f.write(recipe)
   112
   113
           return recipe
   114
   115
       countries_list = get_countries_list()
   116
       ingredients_list = get_ingredients_list()
   117
       recipes_list = get_a_recipe.partial(
   118
           cohere_conn_id=COHERE_CONN_ID, ingredients_list=ingredients_list
   119
       ).expand(country=countries_list)
   120
   121
       get_embeddings = CohereEmbeddingOperator.partial(
   122
           task_id="get_embeddings",
   123
           conn_id=COHERE_CONN_ID,
   124
       ).expand(input_text=recipes_list)
   125
   126
       @task
   127
       def plot_embeddings(embeddings, text_labels, file_name="embeddings_plot.png"):
   128
           "Plot the embeddings of the recipes."
   129
   130
           embeddings = [x[0] for x in embeddings]
   131
           print(text_labels)
   132
   133
           pca = PCA(n_components=2)
   134
           reduced_embeddings = pca.fit_transform(embeddings)
   135
   136
           plt.figure(figsize=(10, 8))
   137
           df_embeddings = pd.DataFrame(reduced_embeddings, columns=["PC1", "PC2"])
   138
           sns.scatterplot(
   139
               df_embeddings, x="PC1", y="PC2", s=100, color="gold", edgecolor="black"
   140
           )
   141
   142
           font_style = {"color": "black"}
   143
           texts = []
   144
           for i, label in enumerate(text_labels):
   145
               texts.append(
   146
                   plt.text(
   147
                       reduced_embeddings[i, 0],
   148
                       reduced_embeddings[i, 1],
   149
                       label,
   150
                       fontdict=font_style,
   151
                       fontsize=15,
   152
                   )
   153
               )
   154
   155
           # prevent overlapping labels
   156
           adjust_text(texts, arrowprops=dict(arrowstyle="->", color="red"))
   157
   158
           distances = euclidean_distances(reduced_embeddings)
   159
           np.fill_diagonal(distances, np.inf)  # exclude cases where the distance is 0
   160
   161
           n = distances.shape[0]
   162
           distances_list = [
   163
               (distances[i, j], (i, j)) for i in range(n) for j in range(i + 1, n)
   164
           ]
   165
   166
           distances_list.sort(reverse=True)
   167
   168
           legend_handles = []
   169
           for dist, (i, j) in distances_list:
   170
               (line,) = plt.plot(
   171
                   [reduced_embeddings[i, 0], reduced_embeddings[j, 0]],
   172
                   [reduced_embeddings[i, 1], reduced_embeddings[j, 1]],
   173
                   "gray",
   174
                   linestyle="--",
   175
                   alpha=0.3,
   176
               )
   177
               legend_handles.append(line)
   178
   179
           legend_labels = [
   180
               f"{text_labels[i]} - {text_labels[j]}: {dist:.2f}"
   181
               for dist, (i, j) in distances_list
   182
           ]
   183
   184
           for i in range(len(reduced_embeddings)):
   185
               for j in range(i + 1, len(reduced_embeddings)):
   186
                   plt.plot(
   187
                       [reduced_embeddings[i, 0], reduced_embeddings[j, 0]],
   188
                       [reduced_embeddings[i, 1], reduced_embeddings[j, 1]],
   189
                       "gray",
   190
                       linestyle="--",
   191
                       alpha=0.5,
   192
                   )
   193
   194
           plt.legend(
   195
               legend_handles,
   196
               legend_labels,
   197
               title="Distances",
   198
               loc="center left",
   199
               bbox_to_anchor=(1, 0.5),
   200
           )
   201
   202
           plt.tight_layout()
   203
           plt.title(
   204
               "2D Visualization of recipe similarities", fontsize=16, fontweight="bold"
   205
           )
   206
           plt.xlabel("PCA Component 1", fontdict=font_style)
   207
           plt.ylabel("PCA Component 2", fontdict=font_style)
   208
   209
           plt.savefig(file_name, bbox_inches="tight")
   210
           plt.close()
   211
   212
       chain(
   213
           get_embeddings,
   214
           plot_embeddings(
   215
               get_embeddings.output,
   216
               text_labels=countries_list,
   217
               file_name=IMAGE_PATH,
   218
           ),
   219
       )
   220
   221
   222
   recipe_suggestions()

   This DAG consists of five tasks to make a simple MLOps pipeline.

   • The get_ingredients task fetches the list of ingredients that the user found in their pantry and wants to use in their recipe. The input pantry_ingredients param is provided by Airflow params when you run the DAG.
   • The get_countries task uses Airflow params to retrieve the list of user-provided countries to get recipes from.
   • The get_a_recipe task uses the CohereHook to connect to the Cohere API and use the /generate endpoint to get a tasty recipe suggestion based on the user’s pantry ingredients and one of the countries they provided. This task is dynamically mapped over the list of countries to generate one task instance per country. The recipes are saved as .txt files in the include folder.
   • The get_embeddings task is defined using the CohereEmbeddingOperator to generate vector embeddings of the recipes generated by the upstream get_a_recipe task. This task is dynamically mapped over the list of recipes to retrieve one set of embeddings per recipe. This pattern allows for efficient parallelization of the vector embedding generation.
   • The plot_embeddings task takes the embeddings created by the upstream task and performs dimensionality reduction using PCA to plot the embeddings in two dimensions.

   

Step 3: Run your DAG

1. Run astro dev start in your Astro project to start Airflow and open the Airflow UI at localhost:8080.

2. In the Airflow UI, run the recipe_suggestions DAG by clicking the play button. Then, provide Airflow params for:

   • Countries of recipe origin: A list of the countries you want to get recipe suggestions from. Make sure to create one line per country and to provide at least two countries.
   • pantry_ingredients: A list of the ingredients you have in your pantry and want to use in the recipe. Make sure to create one line per ingredient.
   • type: Select your preferred recipe type.
   • max_tokens_recipe: The maximum number of tokens available for the recipe.
   • randomness_of_recipe: The randomness of the recipe. The value provided is divided by 10 and given to the temperature parameter of the of the Cohere API. The scale for the param ranges from 0 to 50, with 0 being the most deterministic and 50 being the most random.

   

3. Go to the include folder to view the image file created by the plot_embeddings task. The image should look similar to the one below.

   

Step 4: (Optional) Cook your recipe

1. Choose one of the recipes in the include folder.
2. Navigate to your kitchen and cook the recipe you generated using Cohere with Airflow.
3. Enjoy!

Conclusion

Congratulations! You used Airflow and Cohere to get recipe suggestions based on your pantry items. You can now use Airflow to orchestrate Cohere operations in your own machine learning pipelines.