Providers¶

Providers in Goose mean "LLM providers" that Goose can interact with. Providers are defined in the Exchange library for the most part, but you can define your own.

As you configure your chosen provider, you add the models you want to use to the ~/.config/goose/profiles.yaml file and you can set any necessary environment variables or API keys in your terminal. For example:

export PROVIDER_API_KEY="your_api_key_here"

Currently Available Providers¶

Anthropic¶

To use Anthropic, you need an API key, which you can obtain by signing up or logging into Anthropic's platform. Once you have your API key and your profiles.yaml file updated to the provider, you can set the ANTHROPIC_API_KEY environment variable in your shell using:

export ANTHROPIC_API_KEY="your_api_key_here"`.

profiles.yaml

default:
  provider: anthropic
  processor: claude-3-5-sonnet-20241022
  accelerator: claude-3-5-sonnet-20241022
  moderator: synopsis
  toolkits:
  - name: synopsis
    requires: {}

Important Anthropic performs the best when both the synopsis toolkit and moderator are enabled.

Azure¶

Azure AI services provide API keys through the Azure Portal. Visit the Azure Portal to create a resource and obtain your key. You will need to configure Goose by updating your profile and setting appropriate environment variables.

profiles.yaml

default:
  provider: azure
  processor: azure-gpt-4
  accelerator: azure-gpt-3

Bedrock¶

More information can be found at AWS Bedrock. You need to set up your AWS credentials and configure Bedrock access accordingly in your Goose profile.

profiles.yaml

default:
  provider: bedrock
  processor: titan-llm
  accelerator: titan-llm-lite

Databricks¶

To use Databricks, sign up or log into Databricks and generate a personal access token via the user settings. Configure Goose by setting the DATABRICKS_HOST and DATABRICKS_TOKEN environment variables.

profiles.yaml

default:
  provider: databricks
  processor: databricks-meta-llama-3-1-70b-instruct
  accelerator: databricks-meta-llama-3-1-70b-instruct

Google¶

Google Cloud AI services require you to set up a project in the Google Cloud Console. After enabling the relevant APIs, you should generate an API key or set up a service account. Ensure your application can access these credentials.

profiles.yaml

default:
  provider: google
  processor: gemini-1.5-flash
  accelerator: gemini-1.5-flash
  moderator: truncate
  toolkits:
  - name: developer
    requires: {}

Important Gemini performs the best when both the developer toolkit and truncate moderator are enabled.

Ollama¶

For Ollama, refer to the setup process on Ollama's site for obtaining necessary credentials. Make sure your environment has all the required tokens set up.

profiles.yaml

default:
  provider: ollama
  processor: ollama-pro
  accelerator: ollama-lite

OpenAI¶

Register at OpenAI's platform to obtain an API key. Configure Goose by updating your profiles.yaml file and setting the OPENAI_API_KEY in your terminal:

export OPENAI_API_KEY="your_api_key_here"

profiles.yaml

default:
  provider: openai
  processor: gpt-4
  accelerator: gpt-3.5-turbo