haive.core.models.llm.providers.google¶

Google AI Providers Module.

This module implements both Google Generative AI (Gemini) and Vertex AI providers for the Haive framework. It supports Gemini models through the standard API and enterprise Vertex AI deployments.

The providers handle API key management, model configuration, and safe imports of the langchain-google packages.

Examples

Using Gemini:

from haive.core.models.llm.providers.google import GeminiProvider

provider = GeminiProvider(
    model="gemini-1.5-pro",
    temperature=0.7
)
llm = provider.instantiate()

Using Vertex AI:

from haive.core.models.llm.providers.google import VertexAIProvider

provider = VertexAIProvider(
    model="gemini-1.5-pro",
    project="my-project",
    location="us-central1"
)
llm = provider.instantiate()

GeminiProvider(*[, requests_per_second, ...])

Google Gemini language model provider configuration.

VertexAIProvider(*[, requests_per_second, ...])

Google Vertex AI language model provider configuration.

Classes¶

GeminiProvider

Google Gemini language model provider configuration.

VertexAIProvider

Google Vertex AI language model provider configuration.

Module Contents¶

class haive.core.models.llm.providers.google.GeminiProvider(/, **data)[source]¶

Bases: haive.core.models.llm.providers.base.BaseLLMProvider

Google Gemini language model provider configuration.

This provider supports Google’s Gemini models through the Generative AI API. It’s suitable for general use with API key authentication.

Parameters:

  • data (Any)

  • requests_per_second (float | None)

  • tokens_per_second (int | None)

  • tokens_per_minute (int | None)

  • max_retries (int)

  • retry_delay (float)

  • check_every_n_seconds (float | None)

  • burst_size (int | None)

  • provider (LLMProvider)

  • model (str | None)

  • name (str | None)

  • api_key (SecretStr)

  • cache_enabled (bool)

  • cache_ttl (int | None)

  • extra_params (dict[str, Any] | None)

  • debug (bool)

  • temperature (float | None)

  • max_output_tokens (int | None)

  • top_p (float | None)

  • top_k (int | None)

  • n (int | None)

provider¶

Always LLMProvider.GEMINI

model¶

Model name (default: “gemini-1.5-pro”)

temperature¶

Sampling temperature (0-1)

max_output_tokens¶

Maximum tokens to generate

top_p¶

Nucleus sampling parameter

top_k¶

Top-k sampling parameter

n¶

Number of responses to generate

Environment Variables:

GOOGLE_API_KEY: API key for authentication GEMINI_API_KEY: Alternative API key environment variable

Model Variants:

  • gemini-1.5-pro: Most capable, 1M token context

  • gemini-1.5-flash: Faster, more efficient

  • gemini-pro: Previous generation

  • gemini-pro-vision: Multimodal support

Examples

Basic usage:

provider = GeminiProvider(
    model="gemini-1.5-pro",
    temperature=0.7,
    max_output_tokens=2048
)
llm = provider.instantiate()

With advanced sampling:

provider = GeminiProvider(
    model="gemini-1.5-flash",
    temperature=0.9,
    top_p=0.95,
    top_k=40
)

Create a new model by parsing and validating input data from keyword arguments.

Raises [ValidationError][pydantic_core.ValidationError] if the input data cannot be validated to form a valid model.

self is explicitly positional-only to allow self as a field name.

classmethod get_models()[source]¶

Get available Gemini models.

Returns:

List of available model names

Return type:

list[str]

max_output_tokens: int | None = None¶

Get maximum output tokens for this model.

class haive.core.models.llm.providers.google.VertexAIProvider(/, **data)[source]¶

Bases: haive.core.models.llm.providers.base.BaseLLMProvider

Google Vertex AI language model provider configuration.

This provider supports Google’s models through Vertex AI, suitable for enterprise deployments with project-based authentication and regional control.

Parameters:

  • data (Any)

  • requests_per_second (float | None)

  • tokens_per_second (int | None)

  • tokens_per_minute (int | None)

  • max_retries (int)

  • retry_delay (float)

  • check_every_n_seconds (float | None)

  • burst_size (int | None)

  • provider (LLMProvider)

  • model (str | None)

  • name (str | None)

  • api_key (SecretStr)

  • cache_enabled (bool)

  • cache_ttl (int | None)

  • extra_params (dict[str, Any] | None)

  • debug (bool)

  • project (str | None)

  • location (str)

  • temperature (float | None)

  • max_output_tokens (int | None)

  • top_p (float | None)

  • top_k (int | None)

provider¶

Always LLMProvider.VERTEX_AI

model¶

Model name (default: “gemini-1.5-pro”)

project¶

Google Cloud project ID

location¶

Google Cloud region (default: “us-central1”)

temperature¶

Sampling temperature (0-1)

max_output_tokens¶

Maximum tokens to generate

top_p¶

Nucleus sampling parameter

top_k¶

Top-k sampling parameter

Environment Variables:

GOOGLE_CLOUD_PROJECT: Default project ID GOOGLE_APPLICATION_CREDENTIALS: Path to service account JSON

Authentication:

Vertex AI uses Google Cloud authentication. You can authenticate by: 1. Setting GOOGLE_APPLICATION_CREDENTIALS to service account key path 2. Using gcloud auth application-default login 3. Running on Google Cloud with appropriate IAM roles

Examples

Basic usage:

provider = VertexAIProvider(
    model="gemini-1.5-pro",
    project="my-project",
    location="us-central1"
)
llm = provider.instantiate()

With custom parameters:

provider = VertexAIProvider(
    model="gemini-1.5-flash",
    project="my-project",
    location="europe-west1",
    temperature=0.5,
    max_output_tokens=1024
)

Create a new model by parsing and validating input data from keyword arguments.

Raises [ValidationError][pydantic_core.ValidationError] if the input data cannot be validated to form a valid model.

self is explicitly positional-only to allow self as a field name.

classmethod get_models()[source]¶

Get available Vertex AI models.

Returns:

List of available model names

Return type:

list[str]

max_output_tokens: int | None = None¶

Get maximum output tokens for this model.