Model Providers

Model providers allow you to use OpenAI-compatible API endpoints with cecli. You can define custom providers that point to any OpenAI-compatible API, including self-hosted models, proprietary endpoints, or third-party services.

Built-in providers are defined in providers.json and are automatically available. Use the --model-providers option to add your own.

Configuration Format

Each provider is defined as a JSON object with the following keys:

Configuration File Usage

You can also define model providers in your ~/.cecli/conf.yml or .cecli.conf.yml configuration file:

model-providers:
  my-provider:
    api_base: "https://api.myprovider.com/v1"
    api_key_env:
      - "MY_PROVIDER_API_KEY"
    display_name: "My Provider"

Command Line Usage

Use the --model-providers option to define custom providers as a JSON string:

cecli --model-providers '{
  "my-provider": {
    "api_base": "https://api.myprovider.com/v1",
    "api_key_env": ["MY_PROVIDER_API_KEY"],
    "display_name": "My Provider"
  }
}'

Using Custom Provider Models

Once a provider is configured, you can use its models by referencing them with the provider slug as a prefix:

cecli --model my-provider/model-name

For example, if you configure provider my-provider and it serves a model called gpt-4o-mini, you would use:

cecli --model my-provider/gpt-4o-mini

Setting API Keys

Set the environment variable(s) specified in api_key_env before running cecli:

export MY_PROVIDER_API_KEY="sk-your-key-here"
cecli --model my-provider/gpt-4o-mini

Advanced Configuration

Custom Model List Endpoint

If your provider uses a different endpoint for listing models, specify models_url:

model-providers:
  my-provider:
    api_base: "https://api.myprovider.com/v1"
    api_key_env: ["MY_PROVIDER_API_KEY"]
    models_url: "https://api.myprovider.com/v1/models/list"

Static Model Definitions

If your provider does not expose a /models endpoint, define models statically:

model-providers:
  my-provider:
    api_base: "https://api.myprovider.com/v1"
    api_key_env: ["MY_PROVIDER_API_KEY"]
    static_models:
      - id: "gpt-4o"
        max_input_tokens: 128000
        mode: "chat"
      - id: "gpt-4o-mini"
        max_input_tokens: 128000
        mode: "chat"

Default Headers

Add default HTTP headers for every request:

model-providers:
  my-provider:
    api_base: "https://api.myprovider.com/v1"
    api_key_env: ["MY_PROVIDER_API_KEY"]
    default_headers:
      X-Custom-Header: "value"
      X-Organization: "my-org"

HuggingFace Namespace

For providers that serve HuggingFace models, enable the hf_namespace flag:

model-providers:
  my-hf-provider:
    api_base: "https://api.myprovider.com/v1"
    api_key_env: ["MY_PROVIDER_API_KEY"]
    hf_namespace: true

This will prefix model names with hf: (e.g., hf:meta-llama/Llama-2-7b).

Built-in Providers

Cecli ships with several built-in providers defined in providers.json. These are automatically available without any configuration:

Troubleshooting

“Provider not found”

Ensure the provider slug is correctly spelled and that you are using it as a prefix in the model name:

cecli --model my-provider/model-name

“Missing API key”

Set the appropriate environment variable defined in api_key_env before running cecli:

export MY_PROVIDER_API_KEY="sk-..."

“Connection refused” or timeout errors

Verify that the api_base URL is correct and accessible from your network. Some providers may require VPN access or specific network configuration.