Google AI

The Google AI provider uses provider key google-ai and the HPD-Agent.Providers.GoogleAI package. ModelName is a Gemini model id.

Set either supported API key alias:

export GOOGLE_API_KEY="..."
# or
export GEMINI_API_KEY="..."

Use fluent setup first:

using HPD.Agent;
using HPD.Agent.Providers.GoogleAI;

var agent = await new AgentBuilder()
    .WithGoogleAI(model: "gemini-2.0-flash")
    .BuildAsync();

var result = await agent.RunAsync("Write one sentence about Google AI setup.");
Console.WriteLine(result.Text);

Runtime chat behavior such as temperature, top-p, top-k, max output tokens, stop sequences, JSON output, tools, and reasoning belongs in ChatDefaults or per-run AgentRunConfig.Chat:

var agent = await new AgentBuilder()
    .WithGoogleAI(model: "gemini-2.0-flash")
    .WithChatDefaults(chat =>
    {
        chat.Temperature = 0.2;
        chat.TopP = 0.95;
        chat.MaxOutputTokens = 4096;
    })
    .BuildAsync();

GoogleAIProviderConfig is for Google-native provider construction:

var agent = await new AgentBuilder()
    .WithGoogleAI(
        model: "gemini-2.0-flash",
        configure: google =>
        {
            google.ApiVersion = "v1beta";
        })
    .BuildAsync();

For Vertex AI:

var agent = await new AgentBuilder()
    .WithGoogleAI(
        model: "gemini-2.0-flash",
        configure: google =>
        {
            google.Platform = GoogleAIPlatform.VertexAI;
            google.ProjectId = "my-project";
            google.Region = "us-central1";
            google.ApiVersion = "v1beta1";
            google.CredentialsFile = "/path/to/application_default_credentials.json";
        })
    .BuildAsync();

Use ExpressMode = true for Vertex AI Express mode. Express mode uses API-key authentication, so configure an API key the same way as the Gemini Developer API path.

Provider Options

GoogleAIProviderConfig is for Google-native client construction and authentication setup:

Option	Purpose
Platform	Selects the Google adapter. GeminiDeveloperApi uses the Gemini Developer API with API-key auth; VertexAI uses Vertex AI.
ApiVersion	Selects the Google API version, such as v1, v1beta, or v1beta1.
ProjectId	Google Cloud project id for Vertex AI.
Region	Google Cloud region for Vertex AI, such as us-central1 or global.
ExpressMode	Enables Vertex AI Express mode. Express mode uses API-key authentication.
CredentialsFile	Path to an Application Default Credentials JSON file for Vertex AI.
ValidateAccessToken	Controls whether the underlying adapter validates supplied access tokens. Defaults to true.

The equivalent Clients.Chat shape is:

{
  "Clients": {
    "Chat": {
      "ProviderKey": "google-ai",
      "ModelName": "gemini-2.0-flash"
    }
  }
}

With provider options:

{
  "Clients": {
    "Chat": {
      "ProviderKey": "google-ai",
      "ModelName": "gemini-2.0-flash",
      "ProviderOptions": {
        "platform": "VertexAI",
        "projectId": "my-project",
        "region": "us-central1",
        "apiVersion": "v1beta1",
        "credentialsFile": "/path/to/application_default_credentials.json"
      }
    }
  }
}

Prefer GOOGLE_API_KEY or GEMINI_API_KEY for Gemini Developer API credentials. GOOGLE_AI_API_KEY is also registered as the HPD provider-key convention alias.