Enterprise h2oGPTe Integration

H2O Driverless AI features integration with Enterprise h2oGPTe that lets you optionally generate and view dataset or experiment results summaries using a GPT model. These options let you easily explain the results of complex experiments and datasets and help you share the insights and implications of your experiments and data in a more convenient manner.

참고

For information about H2O.ai’s enterprise generative AI solution, see the Enterprise h2oGPTe documentation and Enterprise h2oGPTe on H2O.ai’s Platform.

The following sections describe how to enable and use GPT-related options in H2O Driverless AI:

Description of relevant configuration options

The following are descriptions of the relevant configuration options when enabling Enterprise h2oGPTe integration.

  • enable_gpt: Specifies whether to enable GPT functionality. By default, this option is set to False. This option must be set to True to use any GPT features, regardless of whether you configure Enterprise h2oGPTe or OpenAI.

  • h2ogpt_url: Specifies your Enterprise h2oGPTe endpoint. For some Enterprise h2oGPTe URLs, you must also set the h2ogpt_key configuration option to enable authorized access for GPT-related tasks. (Optional: OpenAI can also be used by setting the openai_api_secret_key configuration option instead.)

    참고

    Driverless AI checks h2ogpt_url first. If set, it uses Enterprise h2oGPTe. If h2ogpt_url is not set, Driverless AI checks openai_api_secret_key and uses OpenAI if configured.

  • h2ogpt_key: Specifies an Enterprise h2oGPTe key. For some Enterprise h2oGPTe URLs, this key is required to enable authorized access for GPT-related tasks.

  • h2ogpt_model_name: Specifies a model to use based on its name from the list provided by the Enterprise h2oGPTe client. If the provided name doesn’t match any available model, the first model in the list is used by default.

  • openai_api_secret_key: Specifies your OpenAI API key to use OpenAI models instead of Enterprise h2oGPTe. This is only used when h2ogpt_url is not configured.

  • openai_api_model: (Required when using OpenAI) Specifies the OpenAI model to use. By default, this option is set to gpt-4. If you don’t have access to this model, you’ll need to specify a different model that your account has access to.

참고

Configuration requirements for using OpenAI:

  1. Set both openai_api_secret_key and openai_api_model.

  2. The model name must exactly match an available model in your OpenAI account.

경고

If you specify a model that your OpenAI account does not have access to, you may receive the following error:

Error code: 404 - {"error":{"message":"The model 'gpt-4' does not exist or you do not have access to it."}}

GPT summaries

The following subsections describe how to generate and view summaries for datasets and experiment results.

Summarize datasets

  1. Ensure that GPT functionality is enabled by setting enable_gpt=true.

    • For Enterprise h2oGPTe: Configure h2ogpt_url (and h2ogpt_key if required)

    • For OpenAI: Configure both openai_api_secret_key and openai_api_model

  2. Ensure that the relevant configuration options have been set correctly. For more information, see Description of relevant configuration options.

  3. In the main navigation, click Datasets to go to the Datasets page.

  4. Click Click for Actions next to the dataset you want to summarize.

  5. From the list of options, select Summarize.

Summarize experiment results

  1. Ensure that GPT functionality is enabled by setting enable_gpt=true.

  2. Ensure that the relevant configuration options have been set correctly. For more information, see Description of relevant configuration options.

  3. In the main navigation, click Experiments to go to the Experiments page.

  4. For the completed experiment you want to summarize, click Extra > Summarize from the list of experiment options.