Add configuration to a vector

Add vector-scoped configuration to a vector. Use it for feature flags and authored configuration values that Konfidence versions together with the vector.

Konfidence includes the configuration in the vector, so the vector ID uniquely determines its configuration. You can add it in two ways: Konfidence assembles the vector from a VectorTemplate, or you build the vector yourself with the kden command-line tool.

Prerequisites

Before you begin, make sure you meet these requirements:

• You know which path builds your vectors. See Build vectors.
• For the template path: a VectorTemplate for the vector that receives the configuration.
• For the manual path: the vector's component constructor file and the kden CLI.

Choose how the configuration enters the vector

Both paths produce the same result: an Open Component Model (OCM) resource named cloud-konfidence-vector-config on the vector. Choose the path that matches how you build vectors.

You build vectors with	Path	Where the configuration lives
A VectorTemplate and Konfidence assembly	Add the configuration to a VectorTemplate	spec.vectorConfig of the template
kden vector push from a constructor file	Add the configuration to a manually built vector	A JSON file referenced as a resource in the constructor file

Add the configuration to a VectorTemplate

1. Add spec.vectorConfig to your existing VectorTemplate manifest:

   spec:
     vectorConfig:
       features:
         enableBeta: true
         maxUsers: 150
         ratio: 4.6
         title: "TestLabel"
       authored:
         log-level: info
         database:
           host: "mysql-service"
           port: 3306

2. Apply the updated manifest:

   kubectl apply --namespace <namespace> --filename <vector-template-file>

3. Check the assembly status:

   kubectl get vectortemplate <vector-template-name> --namespace <namespace>

When the configuration changes, assembly creates a new vector. After assembly completes, READY is True, REASON is VectorCreated, and LATEST-VECTOR contains the new concrete vector reference. That vector contains the configuration as a local resource named cloud-konfidence-vector-config.

Add the configuration to a manually built vector

Store the configuration in one JSON file next to your existing vector component constructor. The examples below assume that the constructor is named component-constructor.yaml.

1. Write the configuration file next to the constructor file, for example vector-config.json:

   {
     "schemaVersion": "v1",
     "features": {
       "enableBeta": true,
       "maxUsers": 150
     },
     "authored": {
       "log-level": "info"
     }
   }

   schemaVersion must be v1. Both features and authored are optional.

2. Add the file as a local resource named cloud-konfidence-vector-config to the resources list of the vector component in component-constructor.yaml:

   components:
     - name: github.com/example/shop/vector
       version: v1.0.0
       provider:
         name: example
       componentReferences:
         - componentName: github.com/example/shop/checkout
           name: checkout
           version: v1.0.0
       resources:
         - name: cloud-konfidence-vector-config
           type: json
           version: 1.0.0
           relation: local
           input:
             type: file/v1
             path: ./vector-config.json
             mediaType: application/json

   Konfidence matches the resource by its name, not by its type. A vector carries at most one resource with this name. A second one fails the deployment.

3. Validate and push the vector:

   kden vector validate --files ./component-constructor.yaml
   kden vector push --file ./component-constructor.yaml --registry <registry>/<subpath>

   validate checks the component constructor before push writes the vector component version to the registry. After the push succeeds, that component version contains the configuration from vector-config.json as a local resource named cloud-konfidence-vector-config. See kden vector push for all flags.

Feature flags

Use the top-level features block for feature flags. The keys are flat, and values can use any JSON value type, such as Boolean, number, string, array, or object.

{
  "features": {
    "new-checkout": true,
    "max-users": 150,
    "experimental-payment-providers": ["stripe", "adyen"]
  }
}

Konfidence does not add targeting, variants, or rules inside feature flags. The vector is the targeting unit. Changing a flag creates a new vector version, which keeps the change auditable, atomic with code, and reproducible.

To read flags from your application, see Read feature flags in your application.

Authored config

Use the top-level authored block for free-form JSON. Konfidence does not impose a schema on it. This block is a contract between you and your application.

{
  "authored": {
    "ui": {
      "theme": "dark",
      "locale": "en-US"
    },
    "limits": {
      "requestTimeoutMs": 5000
    }
  }
}

Authored config is optional and immutable per vector version. Provide one authored block per vector, or none.

Authored config is available through the whole-bundle response only. Query the vector ID as the flag key and read the authored subtree from the returned vector configuration object. The single-flag and bulk endpoints resolve feature flags only.

Next steps

After you add configuration to a vector, access vector data in your application.