What are the differences between artifacts and cache in GitLab CI as job attributes?

Artifacts and cache in GitLab CI serve to speed up build and testing processes, but they have different functions and purposes.

Artifacts:

• Purpose: To save the results of a job (builds, tests, documentation, etc.) for later use or review. They can be downloaded manually or accessed via the web interface.
• Lifetime: Defined by the .gitlab-ci.yml settings. They can be deleted after a certain period or kept.
• Size: Usually larger than cache, as they contain completed builds, reports, and other results.
• Usage: To transfer results between jobs in different stages or to provide them to users.

Example configuration:

artifacts:
  paths: # paths to files and directories that will become artifacts
    - build/
    - test-results.xml
  expire_in: 1 week # artifact lifetime

Cache:

• Purpose: To save project dependencies or intermediate results to speed up subsequent runs of the same job. Cache is tied to a specific branch or tag (configurable).
• Lifetime: Managed automatically by GitLab. Old cache entries are deleted when quotas are exceeded or the expiration date is reached. Cache can be automatically restored or created on a new run.
• Size: Usually smaller than artifacts, as it contains dependencies like Maven repositories, npm packages, etc.
• Usage: To speed up dependency installation and reduce build time.

Example configuration:

cache:
  paths: # paths to directories to be cached
    - node_modules/
    - ~/.m2/repository/
  key: # cache key, can use predefined variables
    files:
      - package-lock.json # cache will update when this file changes
    prefix: ${CI_COMMIT_REF_SLUG} # adds a prefix with branch/tag name

Key differences:

Attribute	Artifacts (artifacts)	Cache (cache)
Purpose	Saving job results	Saving dependencies to speed up
Usage	Downloading, transferring between jobs (different stages)	Automatic restoration/creation
Lifetime	Explicitly set in .gitlab-ci.yml configuration	Managed by GitLab, automatic deletion
Access	Via web interface, API	Automatically used by Runner
Branch relation	No direct link (unless explicitly specified)	Usually tied to a branch for isolation

Thus, artifacts are the output data of a build, while cache is the reused components to accelerate the build itself.

job_example:
  stage: build
  script:
    - echo "Building and testing..."
    - mkdir build
    - echo "Build result" > build/result.txt
    - echo "dependency installation"
  cache:
    paths:
      - dependency_dir/ # folder with dependencies, will be cached
  artifacts:
    paths:
      - build/ # build folder will become an artifact