8/27/2024

Integrating Ollama with Bitbucket Pipelines

In the fast-paced world of software development, Continuous Integration and Continuous Delivery (CI/CD) practices have become pivotal. Today, we will dive deep into integrating Ollama with Bitbucket Pipelines, a feature that allows you to automate code deployments and streamline your workflow by running your builds in isolated environments. This integration will help boost your deployment process and enable you to leverage the power of Ollama in your CI/CD pipelines.

What is Ollama?

Ollama is an innovative open-source framework showcasing how you can harness the power of containerization for developing, testing, and deploying machine learning models. By utilizing Ollama in your Bitbucket Pipelines, you can automate your model deployment with seamless integration alongside your existing workflows. The setup is straightforward, which makes it ideal for developers looking to quickly implement AI functionalities into their applications.

Why Use Bitbucket Pipelines?

Bitbucket Pipelines is a powerful CI/CD tool that helps teams to automatically build, test, and deploy code based on a configuration file called

bitbucket-pipelines.yml

. By using this, developers can ensure that every piece of code works as intended before it hits production. Integrating this with Ollama means developers can also ensure their machine learning models are functioning correctly after every commit, which is crucial when the models evolve rapidly.

Setting Up Your Bitbucket Pipeline with Ollama

The first step in integrating Ollama into your Bitbucket Pipeline is to ensure your Bitbucket repository is correctly set up for Docker support. Here's how to do it:

Step 1: Enable Docker Service in Bitbucket Pipelines

You’ll need to add a

docker

service to your

bitbucket-pipelines.yml

file to leverage Docker in your Bitbucket Pipelines.

Your configuration file should look something like this:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
image: maven:3.6.1

pipelines:
  default:
    - step:
        name: Build & Test
        services:
          - docker
        script:
          - export TESTCONTAINERS_RYUK_DISABLED=true
          - mvn clean install

definitions:
  services:
    docker:
      memory: 2048

Step 2: Pulling a Model Using Ollama

Once Docker is set up, you can create a pipeline that pulls a model using Ollama. Here's how to pull a model within your pipeline:

Create a custom step to pull models using the Ollama container. You will be utilizing the

OllamaContainer

class to create a new container instance and then executing commands within that context. Here’s how this can be done:

1
2
OllamaContainer ollama = new OllamaContainer("ollama/ollama:0.1.26");
ollama.execInContainer("ollama", "pull", "all-minilm");

In this example, replace

all-minilm

with your desired model name. This command pulls the specified model into your Docker container.

Step 3: Commit the Model to Image

After running your model, the next step is to commit changes in the instance to an image that can be used in future steps or deployments. Here’s the code to do so:

1
ollama.commitToImage("newImageName");

Replace

newImageName

with something meaningful to you so you can reference it in subsequent deployments. You'll then leverage this image later in your pipeline.

Step 4: Update Your Pipeline Configuration File

Once you have the necessary scripts to pull the models and commit them, integrate these commands into your

bitbucket-pipelines.yml

. Here’s an example of how those scripts would fit into your pipeline:

1
2
3
4
5
6
7
pipelines:
  default:
    - step:
        name: Pull Model
        script:
          - export TESTCONTAINERS_RYUK_DISABLED=true
          - java -cp target/yourproject.jar com.yourpackage.YourMainClass

Make sure to adjust the paths and class names properly.

Handling Dependencies

If you’re using libraries such as Maven or Gradle, ensure that you add the relevant dependencies for Ollama in your project file. For Maven, add this to your

pom.xml

1
2
3
4
5
<dependency>
    <groupId>org.testcontainers</groupId>
    <artifactId>ollama</artifactId>
    <version>1.20.1</version>
</dependency>

For Gradle, add this line in your

build.gradle

1
testImplementation "org.testcontainers:ollama:1.20.1"

Debugging Common Issues

Memory Limitations

When running this integration, you might hit memory limits. Bitbucket pipelines provide a default memory limit for Docker but you can increase it in the

docker

service definition:

1
2
3
services:
  docker:
    memory: 4096

Ryuk Configuration

Make sure to set the

TESTCONTAINERS_RYUK_DISABLED

environment variable to avoid errors when starting privileged containers. This is critical as Bitbucket does not permit them by default.

Arsturn to Boost Your Deployment Strategy

While integrating Ollama with Bitbucket Pipelines can elevate your CI/CD game, adding tools like Arsturn can push your deployments further. Arsturn is an AI chatbot builder that enables you to create INTUITIVE conversational agents capable of engaging users on your website. With its seamless integration of ChatGPT, you can automate customer interactions, enhance user engagement, and personalize your offerings effortlessly. The process is as simple as designing the chatbot's functions and training it with your data. You can tailor it to cater to your needs, ensuring a unique experience for your audience. Join the THOUSANDS already utilizing Arsturn to boost their branding and engagement.

Find out more at Arsturn.com and start building connections that convert!

Testing Your Integration

Once everything is set up, commit your

bitbucket-pipelines.yml

file and push it to your repository. Check the pipelines tab in Bitbucket to monitor your builds, ensuring the model pulls correctly and runs without errors.

Conclusion

Integrating Ollama with Bitbucket Pipelines is a fantastic way to automate your deployment process, ensure the health of your models, and streamline your CI/CD workflow. This brings a solid approach to scaling your AI deployment efforts effectively. Using an intuitive tool like Arsturn on top of this adds a layer of user engagement that’s invaluable in today’s digital landscape. By utilizing these technologies, you will unleash the potential of your code and engage your audience effectively.

Let’s get started!