litellm/terraform/provider

LiteLLM Terraform Provider

This Terraform provider allows you to manage LiteLLM resources through Infrastructure as Code. It provides support for managing models, teams, team members, API keys, users, organizations, budgets, tags, projects, guardrails, prompts, agents, search tools, access groups, fallbacks, MCP servers, credentials and vector stores via the LiteLLM REST API, along with read-only data sources for each of them.

Source of truth

This directory (terraform/provider/ in BerriAI/litellm) is the source of truth for the provider. BerriAI/terraform-provider-litellm is a thin release mirror that the public Terraform Registry ingests from; do not open PRs there. Changes land here, where CI builds the provider, runs its tests, and statically audits every endpoint the provider calls against the proxy's generated OpenAPI schema (tools/endpointaudit/), so the provider cannot drift from the LiteLLM API silently. The same audit runs in reverse as a coverage gate: every management endpoint in the schema must be covered by a resource or data source, or carry a documented entry in tools/endpointaudit/coverage_allowlist.txt, and stale allowlist entries fail CI. Releases are published by mirroring this directory into the split repo and tagging it, which triggers the goreleaser workflow there (see RELEASING.md)

Versioning

The provider version is the LiteLLM version. Every LiteLLM release (dev, rc and stable) publishes the provider at the same version as the proxy, built from the same commit, so 1.99.0 of the provider is the one that shipped with 1.99.0 of the proxy and was audited against that proxy's API. Pin the provider to the line your proxy runs:

version = "~> 1.99.0"

Pre-release versions (1.99.0-rc.1, 1.99.0-dev.1) are published too; Terraform only selects one when it is pinned exactly.

Versions 0.1.0 through 0.4.0 predate this scheme and sit on their own line. They stay in the registry, but a ~> 0.4 constraint will never pick up another release: re-pin to the LiteLLM version to keep receiving updates.

Features

• Manage LiteLLM model configurations
• Associate models with specific teams
• Create and manage teams
• Configure team members and their permissions
• Set usage limits and budgets
• Control access to specific models
• Specify model modes (e.g., completion, embedding, image generation)
• Manage API keys with fine-grained controls
• Support for reasoning effort configuration in the model resource

Requirements

• Terraform >= 0.13.x
• Go >= 1.16 (for development)

Using the Provider

To use the LiteLLM provider in your Terraform configuration, you need to declare it in the terraform block:

terraform {
  required_providers {
    litellm = {
      source  = "BerriAI/litellm"
      version = "~> 1.99.0" # the LiteLLM version your proxy runs
    }
  }
}

provider "litellm" {
  api_base = var.litellm_api_base
  api_key  = var.litellm_api_key
}

Then, you can use the provider to manage LiteLLM resources. Here's an example of creating a model configuration:

resource "litellm_model" "gpt4" {
  model_name          = "gpt-4-proxy"
  custom_llm_provider = "openai"
  model_api_key       = var.openai_api_key
  model_api_base      = "https://api.openai.com/v1"
  base_model          = "gpt-4"
  tier                = "paid"
  mode                = "chat"
  reasoning_effort    = "medium"  # Optional: "low", "medium", or "high"
  
  input_cost_per_million_tokens  = 30.0
  output_cost_per_million_tokens = 60.0
}

For full details on the litellm_model resource, see the model resource documentation.

Here's an example of creating an API key with various options:

resource "litellm_key" "example_key" {
  key_type             = "llm_api"
  models               = ["gpt-4", "claude-3.5-sonnet"]
  max_budget           = 100.0
  user_id              = "user123"
  team_id              = "team456"
  max_parallel_requests = 5
  tpm_limit            = 1000
  rpm_limit            = 60
  budget_duration      = "monthly"
  key_alias            = "prod-key-1"
  duration             = "30d"
  metadata             = {
    environment = "production"
  }
  allowed_cache_controls = ["no-cache", "max-age=3600"]
  soft_budget          = 80.0
  aliases              = {
    "gpt-4" = "gpt4"
  }
  config               = {
    default_model = "gpt-4"
  }
  permissions          = {
    can_create_keys = "true"
  }
  model_max_budget     = jsonencode({
    "gpt-4" = {
      budget_limit = 50.0
      time_period  = "30d"
    }
  })
  model_rpm_limit      = {
    "claude-3.5-sonnet" = 30
  }
  model_tpm_limit      = {
    "gpt-4" = 500
  }
  guardrails           = ["content_filter", "token_limit"]
  blocked              = false
  tags                 = ["production", "api"]
}

The litellm_key resource supports the following options:

• key_type: Choose the key's default route access
• models: List of allowed models for this key
• max_budget: Maximum budget for the key
• user_id and team_id: Associate the key with a user and team
• max_parallel_requests: Limit concurrent requests
• tpm_limit and rpm_limit: Set tokens and requests per minute limits
• budget_duration: Specify budget duration (e.g., "monthly", "weekly")
• key_alias: Set a friendly name for the key
• duration: Set the key's validity period
• metadata: Add custom metadata to the key
• allowed_cache_controls: Specify allowed cache control directives
• soft_budget: Set a soft budget limit
• aliases: Define model aliases
• config: Set configuration options
• permissions: Specify key permissions
• model_max_budget, model_rpm_limit, model_tpm_limit: Set per-model limits
• guardrails: Apply specific guardrails to the key
• blocked: Flag to block/unblock the key
• tags: Add tags for organization and filtering

For full details on the litellm_key resource, see the key resource documentation.

Available Resources

• litellm_model: Manage model configurations. Documentation
• litellm_team: Manage teams. Documentation
• litellm_team_member: Manage team members. Documentation
• litellm_team_member_add: Add multiple members to teams. Documentation
• litellm_key: Manage API keys. Documentation
• litellm_mcp_server: Manage MCP (Model Context Protocol) servers. Documentation
• litellm_credential: Manage credentials for secure authentication. Documentation
• litellm_vector_store: Manage vector stores for embeddings and RAG. Documentation
• litellm_jwt_key_mapping: Map JWT claim values to virtual keys for per-client budgets and limits. Documentation

Available Data Sources

• litellm_credential: Retrieve information about existing credentials. Documentation
• litellm_vector_store: Retrieve information about existing vector stores. Documentation

Development

Project Structure

The project is organized as follows:

terraform-provider-litellm/
├── litellm/
│   ├── provider.go
│   ├── resource_model.go
│   ├── resource_model_crud.go
│   ├── resource_team.go
│   ├── resource_team_member.go
│   ├── resource_key.go
│   ├── resource_key_utils.go
│   ├── types.go
│   └── utils.go
├── main.go
├── go.mod
├── go.sum
├── Makefile
└── ...

Building the Provider

1. Clone the repository:

git clone https://github.com/your-username/terraform-provider-litellm.git

2. Enter the repository directory:

cd terraform-provider-litellm

3. Build and install the provider:

make install

Development Commands

The Makefile provides several useful commands for development:

• make build: Builds the provider
• make install: Builds and installs the provider
• make test: Runs the test suite
• make fmt: Formats the code
• make vet: Runs go vet
• make lint: Runs golangci-lint
• make clean: Removes build artifacts and installed provider

Testing

To run the tests:

make test

Contributing

Contributions are welcome! Please read our contributing guidelines first.

License

This project is licensed under the Apache License 2.0 - see the LICENSE file for details.

Notes

• Always use environment variables or secure secret management solutions to handle sensitive information like API keys and AWS credentials.
• Refer to the comprehensive documentation in the docs/ directory for detailed usage examples and configuration options.
• Keep the provider version in step with the LiteLLM version your proxy runs; see Versioning.
• The provider now supports AWS cross-account access with aws_session_name and aws_role_name parameters in the model resource.
• All example configurations have been consolidated into the documentation for better organization and maintenance.