Dynamic and Runtime Application Configuration

Configuration for Spin application features such as application variables, key value storage, SQL storage and Serverless AI can be supplied dynamically, i.e. during the application runtime, requiring no changes to the application code itself.

This runtime configuration data is stored in the runtime-config.toml file and passed in via the --runtime-config-file flag when invoking the spin up command.

Let’s look at each configuration category in-depth below.

Application Variables Runtime Configuration

Application Variables values may be set at runtime by providers. Currently, there are three application variable providers: the environment-variable provider, the Vault provider and the Azure Key Vault provider.

Multiple application variable providers can be configured in Spin. Providers are prioritized top-down in the runtime configuration file, with higher-listed providers taking precedence. The environment variable provider always has the highest priority.

The provider examples below show how to use or configure each provider. For examples on how to access these variables values within your application, see Using Variables from Applications.

Environment Variable Provider

The environment variable provider gets variable values from the spin process’s environment (not the component environment). Variable keys are translated to environment variables by upper-casing and prepending with SPIN_VARIABLE_:

$ export SPIN_VARIABLE_API_KEY="1234"  # Sets the `api_key` value.
$ spin up

Vault Application Variable Provider

The Vault application variable provider gets secret values from HashiCorp Vault. Currently, only the KV Secrets Engine - Version 2 is supported. You can set up the v2 kv secret engine at any mount point and provide Vault information in the runtime configuration file:

[[config_provider]]
type = "vault"
url = "http://127.0.0.1:8200"
token = "root"
mount = "secret"

Vault Application Variable Provider Example

  1. Install Vault.
  2. Start Vault:
$ vault server -dev -dev-root-token-id root
  1. Set a password:
$ export VAULT_TOKEN=root
$ export VAULT_ADDR=http://127.0.0.1:8200
$ vault kv put secret/secret value="test_password"
$ vault kv get secret/secret
  1. Go to the Vault variable provider example application.
  2. Build and run the vault-provider app:
$ spin build
$ spin up --runtime-config-file runtime_config.toml
  1. Test the app:
$ curl localhost:3000 --data "test_password"
{"authentication": "accepted"}

$ curl localhost:3000 --data "wrong_password"
{"authentication": "denied"}

Azure Key Vault Application Variable Provider

The Azure Key Vault application variable provider gets secret values from Azure Key Vault.

Currently, only receiving the latest version of a secret is supported.

For authenticating against Azure Key Vault, you must use the client credentials flow. To do so, create a Service Principal (SP) within your Microsoft Entra ID (previously known as Azure Active Directory) and assign the Key Vault Secrets User role to the SP on the scope of your Azure Key Vault instance.

You can set up Azure Key Vault application variable provider in the runtime configuration file:

[[config_provider]]
type = "azure_key_vault"
vault_url = "https://spin.vault.azure.net/"
client_id = "12345678-1234-1234-1234-123456789012"
client_secret = "some.generated.password"
tenant_id = "12345678-1234-1234-1234-123456789012"
authority_host = "AzurePublicCloud"

Azure Key Vault Application Variable Provider Example

  1. Deploy Azure Key Vault:
# Variable Definition
$ KV_NAME=spin123
$ LOCATION=germanywestcentral
$ RG_NAME=rg-spin-azure-key-vault

# Create an Azure Resource Group and an Azure Key Vault
$ az group create -n $RG_NAME -l $LOCATION
$ az keyvault create -n $KV_NAME \
  -g $RG_NAME \
  -l $LOCATION \
  --enable-rbac-authorization true

# Grab the Azure Resource Identifier of the Azure Key Vault instance
$ KV_SCOPE=$(az keyvault show -n $KV_NAME -g $RG_NAME -otsv --query "id")
  1. Add a Secret to the Azure Key Vault instance:
# Grab the ID of the currently signed in user in Azure CLI
$ CURRENT_USER_ID=$(az ad signed-in-user show -otsv --query "id")

# Make the currently signed in user a "Key Vault Secrets Officer"
# on the scope of the new Azure Key Vault instance
$ az role assignment create --assignee $CURRENT_USER_ID \
  --role "Key Vault Secrets Officer" \
  --scope $KV_SCOPE

# Create a test secret called `secret` in the Azure Key Vault instance
$ az keyvault secret set -n secret --vault-name $KV_NAME --value secret_value --onone
  1. Create a Service Principal and Role Assignment for Spin:
$ export SP_NAME=sp-spin

# Create the SP
$ SP=$(az ad sp create-for-rbac -n $SP_NAME -ojson)

# Populate local shell variables from the SP JSON
$ CLIENT_ID=$(echo $SP | jq -r '.appId')
$ CLIENT_SECRET=$(echo $SP | jq -r '.password')
$ TENANT_ID=$(echo $SP | jq -r '.tenant')

# Assign the "Key Vault Secrets User" role to the SP
# allowing it to read secrets from the Azure Key Vault instance
$ az role assignment create --assignee $CLIENT_ID \
  --role "Key Vault Secrets User" \
  --scope $KV_SCOPE
  1. Go to the Azure Key Vault variable provider example application.
  2. Replace Tokens in runtime_config.toml.

The azure-key-vault-provider application contains a runtime_config.toml file. Replace all tokens (e.g. $KV_NAME$) with the corresponding shell variables you created in the previous steps.

  1. Build and run the azure-key-vault-provider app:
$ spin build
$ spin up --runtime-config-file runtime_config.toml
  1. Test the app:
$ curl localhost:3000
Loaded Secret from Azure Key Vault: secret_value

Key Value Store Runtime Configuration

Spin provides built-in key-value storage. This storage is backed by an SQLite database embedded in Spin by default. However, the Spin runtime configuration file (runtime-config.toml) can be updated to not only modify the SQLite configuration but also choose to use a different backing store. The available store options are the embedded SQLite database, an external Redis database or Azure CosmosDB.

Redis Key Value Store Provider

The following is an example of how an application’s runtime-config.toml file can be configured to use Redis instead. Note the type and url values, which are set to redis and the URL of the Redis host, respectively:

[key_value_store.default]
type = "redis"
url = "redis://localhost"

Azure CosmosDB Key Value Store Provider

Similarly, to implement Azure CosmosDB as a backend for Spin’s key/value store, change the type to azure_cosmos and specify your database account details:

[key_value_store.default]
type = "azure_cosmos"
key = "<key>"
account = "<cosmos-account>"
database = "<cosmos-database>"
container = "<cosmos-container>"

Note: The CosmosDB container must be created with the default partition key, /id.

Whilst a single default store may be sufficient for certain application use cases, each Spin application can be configured to support multiple stores of any type, as shown in the runtime-config.toml file below:

Note: At present, when deploying an application to Fermyon Cloud only the single “default” key-value store is supported. To see more about Spin support on Fermyon Cloud, visit the limitations documentation:

# This defines a new store named user_data
[key_value_store.user_data]
type = "spin" 
path = ".spin/user_data.db"

# This defines a new store named other_data backed by a Redis database
[key_value_store.other_data]
type = "redis"
url = "redis://localhost"

You must individually grant each component access to the stores that it needs to use. To do this, use the component.key_value_stores entry in the component manifest within spin.toml. See Spin Key Value Store for more details.

SQLite Storage Runtime Configuration

Spin provides built-in SQLite storage. By default, this is backed by a database that Spin creates for you underneath your application directory (in the .spin subdirectory). However, you can use the Spin runtime configuration file (runtime-config.toml) to add and customize SQLite databases.

The following example runtime-config.toml tells Spin to map the default database to an SQLite database elsewhere in the file system:

[sqlite_database.default]
type = "spin"
path = "/planning/todo.db"

If you need more than one database, you can configure multiple databases, each with its own name:

# This defines a new store named todo
[sqlite_database.todo]
type = "spin"
path = "/planning/todo.db"

# This defines a new store named finance
[sqlite_database.finance]
type = "spin"
path = "/super/secret/monies.db"

Spin creates any database files that don’t exist. However, it is up to you to delete them when you no longer need them.

LibSQL Storage Provider

Spin can also use libSQL databases accessed over HTTPS. libSQL is fully compatible with SQLite but provides additional features including remote, distributed databases.

Spin does not provide libSQL access to file-based databases, only databases served over HTTPS. Specifically, Spin supports the sqld libSQL server.

To use libSQL, set type = "libsql" in your runtime-config.toml entry. You must then provide a url and authentication token instead of a file path. For example, this entry tells Spin to map the default database to a libSQL service running on libsql.example.com:

# This tells Spin to use the remote host as its default database
[sqlite_database.default]
type = "libsql"
url = "https://sensational-penguin-ahacker.libsql.example.com"
token = "a secret"

Spin does not create libSQL databases. Use your hosting service’s tools to create them (or sqld if you are self-hosting) . You can still set up tables and data in a libSQL database via spin up --sqlite.

You must include the scheme in the url field. The scheme must be http or https. Non-HTTP libSQL protocols are not supported.

The default database will still be defined, even if you add other databases.

By default, components will not have access to any of these databases (even the default one). You must grant each component access to the databases that it needs to use. To do this, use the component.sqlite_databases entry in the component manifest within spin.toml. See SQLite Database for more details.

LLM Runtime Configuration

Spin provides a Large Language Model interface for interacting with LLMs for inferencing and embedding. The default host implementation is to use local CPU/GPU compute. However, the Spin runtime configuration file (runtime-config.toml) can be updated to enable Spin to use remote compute using HTTP requests.

Remote Compute Provider

The following is an example of how an application’s runtime-config.toml file can be configured to use the remote compute option. Note the type, url and auth_token are set to remote_http, URL of the server and the auth token for the server. 

[llm_compute]
type = "remote_http"
url = "http://example.com"
auth_token = "<auth_token>"

Currently, the remote compute option requires an user to deploy their own LLM proxy service. Fermyon Cloud users can do this using the cloud-gpu plugin. If you prefer to create and deploy your own proxy service, you can find a reference implementation of the proxy protocol in the spin-cloud-gpu plugin repository.

By default, components will not have access to the LLM models unless granted explicit access through the component.ai_models entry in the component manifest within spin.toml. See Serverless AI for more details.