fluid providers
List the infrastructure providers this install has registered: the deployment targets --provider and the contract's binding.platform can resolve to.
From 0.27.0, the list is read from entry-point metadata and imports no provider, so a plugin's code does not run; --debug imports each provider to report it. A provider plugin that fails to import is reported, by name and error type, the first time a command uses it.
Syntax
fluid providers [--debug]
Key options
| Option | Description |
|---|---|
--debug | Print discovery metadata for each provider: where it was registered from, its module, and the metadata it declares (display name, version, supported platforms, tags). |
Example
fluid providers
{
"providers": [
"aws",
"databricks",
"gcp",
"local",
"snowflake"
]
}
The output is already JSON; fluid providers --json is not a flag.
With --debug, each name becomes an object. This is the first entry, trimmed:
{
"providers": [
{
"name": "aws",
"source": "explicit",
"module": "fluid_build.providers.aws.provider",
"info": {
"name": "aws",
"display_name": "Amazon Web Services",
"supported_platforms": ["aws", "s3", "athena", "glue"],
"tags": ["aws", "cloud", "s3", "glue", "athena"],
...
}
},
...
]
}
source is explicit for providers the CLI registers itself and entrypoint for providers discovered through the fluid_build.providers entry-point group. In the 0.18.1 install used for this page snowflake was the one reported as entrypoint. --debug also writes discovery log lines to the terminal before the JSON.
Retired provider names
(from 0.27.0) redshift and datamesh-manager (also spelled datamesh_manager) are no longer providers. Naming either one, with --provider or FLUID_PROVIDER, exits 2 with a message that says where to go instead:
Provider 'redshift' was retired: Redshift is part of the aws provider; use --provider aws
Provider 'datamesh-manager' was retired: Data Mesh Manager is a catalog, not a provider; use fluid publish --target datamesh-manager (or fluid datamesh-manager)
- Redshift objects apply through the
awsprovider. A contract whosebuilds[].execution.runtime.platformisredshiftis served byaws.binding.platform: redshiftwas never valid in any schema; bind toaws. - Data Mesh Manager publishing is unchanged:
fluid publish --target datamesh-manager(aliasesentropy-data,dmm) andfluid datamesh-manager. - A third-party plugin cannot claim either name. Discovery skips it with a WARNING.
When a provider cannot be found
A command reaches a provider by one of two routes, and each fails with its own message.
1. You name it: --provider or FLUID_PROVIDER. (from 0.27.0) The check accepts built-in providers only. A platform served by an IaC plugin is reached through the contract's binding.platform instead. The flag is global, so it goes before the subcommand (fluid --provider gcp ...) or in the environment. The value is checked against the list above before the command runs. Anything else exits 2:
⚠️ Unknown provider 'nope' — installed providers: aws, databricks, gcp, local, snowflake (see `fluid providers`)
This check carries no ERR_ code. A subcommand that also declares its own --provider (for example fluid export) is checked the same way after parsing. --provider gcp additionally prints Provider 'gcp' requires --project to be specified unless the global --project (or FLUID_PROJECT) is set; the line is a warning, and the command carries on.
2. The contract names it. Without a flag, apply and plan read the platform from the contract: the first exposes[].binding.platform, then a top-level binding.platform, then builds[].execution.runtime.platform. They print the result as Detected provider: local. fluid policy apply reads provider from the compiled bindings file instead. If that name is not registered, or no provider can be determined, the command fails with one of the codes below. Every one prints a link to this page.
ERR_PROVIDER_UNKNOWN, exit 2
The contract (or bindings file) names a provider that is not registered. Here a bindings file for fluid policy apply whose binding names provider: azure:
❌ provider_unknown [ERR_PROVIDER_UNKNOWN]
requested: azure
available: ['aws', 'databricks', 'gcp', 'local', 'snowflake']
💡 Suggestions:
• Run 'fluid providers' to see the available providers
• Valid built-ins: aws, databricks, gcp, local, snowflake
Fix: use a name from available, or install a package that registers the provider. Through 0.26.0 the second suggestion listed azure, which was not registered. See Known gaps.
ERR_PROVIDER_NOT_SPECIFIED, exit 2
No provider was passed, none is in the environment, and none could be read from the file. Reproduced with a bindings file whose bindings carry no provider:
fluid policy apply bindings.json --mode enforce
❌ provider_not_specified [ERR_PROVIDER_NOT_SPECIFIED]
💡 Suggestions:
• Pass a provider: --provider aws|databricks|gcp|local|snowflake
• Or set the FLUID_PROVIDER environment variable
• Run 'fluid providers' to list the available providers
Fix: fluid --provider <name> policy apply ..., or export FLUID_PROVIDER=<name>.
ERR_PROVIDER_NOT_FOUND, exit 1
Raised when an action in an execution plan names a provider that is missing from the running registry. The suggestions are to check the spelling and to install the provider's extra, for example:
pip install 'data-product-forge[gcp]'
The package defines local, gcp, snowflake and aws extras, and others for catalogs and tooling; Upgrading lists them all.
ERR_PROVIDER_BLOCKED_BY_POLICY, exit 2
The provider is installed, and an operator policy removed it. FLUID_PLUGINS_BLOCKLIST names it, or FLUID_PLUGINS_ALLOWLIST is set and does not. Reproduced with the quickstart contract (binding.platform: local):
FLUID_PLUGINS_BLOCKLIST=local fluid apply contract.fluid.yaml --yes
Detected provider: local, project: gold.customer.analytics_360_v1
❌ provider_blocked_by_policy [ERR_PROVIDER_BLOCKED_BY_POLICY]
requested: local
fluid plugins --role provider shows the effective policy:
provider (4)
• aws allowed
from=data-product-forge 0.18.1
...
• local BLOCKED (allow/block policy)
from=data-product-forge 0.18.1
Fix: unset or amend the variable. See fluid plugins.
If the same blocked name arrives through --provider local, the global check reports Unknown provider 'local' instead, because a blocked provider is already absent from the registry that check reads.
ERR_RESERVED_PROVIDER_NAME, exit 1
Raised by fluid provider-init when the name you chose belongs to a built-in:
❌ reserved_provider_name [ERR_RESERVED_PROVIDER_NAME]
name: aws
Fix: choose another name. The reserved names are listed on the provider-init page.
Known gaps
- Through 0.26.0,
azureappeared in the--providerchoices offluid exportand in theERR_PROVIDER_*suggestions, but no Azure provider was registered. (since 0.27.0) It is in neither, andfluid --provider azure ...,fluid export --provider azure,fluid init --provider azureandfluid import dbt --provider azureexit 2 withUnknown provider 'azure'. confluentis accepted byfluid generate iac --provider, because it is an IaC platform, but it is not a provider: the globalfluid --provider confluent ...exits 2 withUnknown provider 'confluent'.databricksis registered, so--provider databricksis accepted.
Notes
- These are deployment targets: providers that
fluid applycan provision infrastructure against. - Providers the CLI registers itself are not all in the
providerrole thatfluid pluginslists. That role holds the providers discovered through entry points, andfluid providersholds the larger registry. - Spec-export formats (ODCS, ODPS, ODPS-Bitol) are not providers; they do not deploy infrastructure. List them with
fluid exportersinstead. - See the provider guides for target-specific documentation.