GitOps (Pipeline-as-Code)

VectorFlow supports pipeline-as-code workflows where pipeline configurations are stored in a Git repository and kept in sync between VectorFlow and your version control system.

Supported Git Providers

VectorFlow supports the following Git hosting providers:

The provider is auto-detected from the repository URL domain (e.g., github.com, gitlab.com, bitbucket.org). For self-hosted instances (e.g., gitlab.internal.corp), you can explicitly set the provider in the Git Integration settings.

Modes

Each environment can operate in one of four GitOps modes:

Setting up Push Only

Push-only mode commits pipeline YAML files to your Git repository every time you deploy or delete a pipeline.

From this point forward, every pipeline deploy writes the generated YAML to {environment-name}/{pipeline-name}.yaml in the configured repository, and every pipeline deletion removes the file.

Setting up Bi-directional GitOps

Bi-directional mode adds a webhook so that pushes to the Git repository automatically import or update pipelines in VectorFlow.

How the import works

When a push event arrives:

1. VectorFlow verifies the webhook signature using the appropriate method for the Git provider.
2. It checks that the push targets the configured branch.
3. For each added or modified .yaml / .yml file in the push, it fetches the file content via the provider's API.
4. The pipeline name is derived from the filename (e.g., production/my-pipeline.yaml becomes my-pipeline).
5. If a pipeline with a matching gitPath or name already exists in the environment, its graph is replaced. Otherwise, a new pipeline is created.
6. If the environment has Require Deploy Approval enabled, imported pipelines are saved as drafts with a pending deploy request instead of being deployed immediately.

Pipeline Name / Filename Decoupling

VectorFlow uses a stable gitPath field to track the file path in Git for each pipeline. This means:

• Renaming a pipeline in VectorFlow does not change its filename in Git. The original path is preserved.
• First sync automatically assigns a gitPath based on the environment and pipeline name slugs.
• Webhook imports match files by gitPath first, then by name as a fallback.

This prevents broken sync when pipelines are renamed after initial setup.

Sync Status and Retries

The Git Sync Status section on the environment detail page shows:

• Health badge -- Green (healthy), yellow (pending retries), or red (failed).
• Last successful sync timestamp.
• Recent sync jobs with status, attempt count, and per-job retry buttons.
• Import errors from webhook events (YAML parse failures, invalid filenames, etc.).

Failed sync operations are automatically retried up to 3 times with exponential backoff (30 seconds, 2 minutes, 10 minutes). After all retries are exhausted, a git_sync_failed alert is fired (if subscribed). You can also manually retry failed jobs from the UI.

The environment list page shows a warning badge when an environment has unresolved sync failures.

File layout

VectorFlow expects pipeline YAML files to follow the standard Vector configuration format:

repo-root/
  environment-name/
    pipeline-a.yaml
    pipeline-b.yaml
  other-environment/
    pipeline-c.yaml

The directory name should match the slugified environment name. Files must have a .yaml or .yml extension.

Disabling GitOps

To disable GitOps, set the mode back to Off and click Save. The webhook secret is cleared, and incoming webhook requests will be rejected.