Configuration #

Global configuration #

The fabric configuration block serves as the global configuration for Fabric, offering a centralized space for defining essential aspects such as plugin dependencies and local directory paths.

fabric {
  ...
}

Within the codebase, only one fabric block can be defined.

Supported arguments #

plugin_versions: (required) a map that aligns namespaced plugin names with version constraints in SemVer (refer to Terraform version constraint syntax).
cache_dir: (optional) a path to a directory on the local file system. The default value is .fabric directory in the current folder. If the directory doesn’t exist, Fabric creates it upon the first run.

Supported nested blocks #

plugin_registry: (optional) a block defines available plugin registries and can include the following arguments:
```
plugin_registry {
  base_url = "<url>"
  mirror_dir = "<path>"
}
```
- base_url: (optional) the base URL of the plugin registry. Default value: https://registry.blackstork.io
- mirror_dir: (optional) the path to a directory on the local filesystem containing plugin binaries.

Example #

fabric {

  cache_dir = "./.fabric"

  plugins_registry {
    mirror_dir = "/tmp/local-mirror/plugins"
  }

  plugin_versions = {
    "blackstork/elastic" = "1.2.3"
    "blackstork/openai" = "=11.22.33"
  }
}

Data source configuration #

A Fabric plugin can include one or more data sources. For example, blackstork/github plugin includes github_issues data source.

A data source within Fabric serves the purpose of loading data from either a local or an external data store, platform, or service.

Configuration for data sources is accomplished through the use of the config block:

config data <data-source-name> "<name>" {
  ...
}

If the block is named (<name> is provided), the config block can be referenced in a config argument inside data blocks. This is helpful if there is a need to have more than one configuration for the same data source.

If <name> isn’t provided, the configuration acts as a default configuration for a specified data source.

Supported arguments #

The arguments allowed in the configuration block depend on the data source. See Plugins for the details on the configuration parameters supported.

Supported nested blocks #

Nested blocks aren’t supported inside the config blocks.

Example #


config data csv {
  delimiter = ";"
}

data csv "events_a" {
  path = "/tmp/events-a.csv"
}

document "test-document" {

  data ref {
    base = data.csv.events_a
  }

  data csv "events_b" {
    config {
      delimiter = ",";
    }

    path = "/tmp/events-b.csv"
  }
}

Content provider configuration #

A Fabric plugin can include one or more content providers. For example, blackstork/openai plugin includes openai_text content provider. Content providers generate Markdown content that Fabric will include into the rendered document. The provider can either render content locally or use an external API (like OpenAI API).

Content providers can be configured using config block:

config content <content-provider-name> "<name>" {
  ...
}

If the block is named (<name> is provided), the config block can be explicitly referenced in a config argument inside content. This is helpful if there is a need to have more than one configuration for the same content provider.

If <name> isn’t provided, the configuration acts as a default configuration for a specified content provider.

Supported arguments #

The arguments allowed in the configuration block depend on the content provider. See Plugins for the details on the configuration parameters supported.

Supported nested blocks #

Nested blocks aren’t supported inside the config blocks.

Example #

config content openai_text {
  api_key = 'some-openai-api-key'

  system_prompt = 'You are the best at saying Hi!'
}

document "test-document" {

  content openai_text {
    prompt = "Say hi!"
  }
}

Next steps #

See Documents to learn how to build document templates in Fabric configuration language.