Skip to content

Google Gen AI

Installation

pip3 install genkit-plugin-google-genai

Configuration

genkit-plugin-google-genai package provides two plugins:

  1. GoogleAI - for accessing Google Gemini API models
  2. VertexAI - for accessing Gemini API in Vertex AI models

Google Gemini API

from genkit.plugins.google_genai import GoogleAI

ai = Genkit(
  plugins=[GoogleAI()],
  model='googleai/gemini-2.0-flash', # optional
)

You will need to set GEMINI_API_KEY environment variable or you can provide the API Key directly:

ai = Genkit(
  plugins=[GoogleAI(api_key='...')]
)

Gemini API in Vertex AI

from genkit.plugins.google_genai import VertexAI

ai = Genkit(
  plugins=[VertexAI()],
  model='vertexai/gemini-2.0-flash', # optional
)

you can specify location and project as well as other configuation options.

ai = Genkit(
  plugins=[VertexAI(
    location='us-east1',
    project='my-project-id',
  )],
)

API Reference

GeminiConfigSchema

Bases: google.genai.types.GenerateContentConfig

Source code in plugins/google-genai/src/genkit/plugins/google_genai/models/gemini.py 
180
181
class GeminiConfigSchema(genai_types.GenerateContentConfig):
    pass

GoogleAI

Bases: genkit.ai.Plugin

GoogleAI plugin for Genkit.

Source code in plugins/google-genai/src/genkit/plugins/google_genai/google.py 
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
class GoogleAI(Plugin):
    """GoogleAI plugin for Genkit."""

    name = GOOGLEAI_PLUGIN_NAME
    _vertexai = False

    def __init__(
        self,
        api_key: str | None = None,
        credentials: Credentials | None = None,
        debug_config: DebugConfig | None = None,
        http_options: HttpOptions | HttpOptionsDict | None = None,
    ):
        api_key = api_key if api_key else os.getenv('GEMINI_API_KEY')
        if not api_key:
            raise ValueError(
                'Gemini api key should be passed in plugin params or as a GEMINI_API_KEY environment variable'
            )

        self._client = genai.client.Client(
            vertexai=self._vertexai,
            api_key=api_key,
            credentials=credentials,
            debug_config=debug_config,
            http_options=_inject_attribution_headers(http_options),
        )

    def initialize(self, ai: GenkitRegistry) -> None:
        """Initialize the plugin by registering actions in the registry.

        Args:
            ai: the action registry.

        Returns:
            None
        """
        for version in GoogleAIGeminiVersion:
            gemini_model = GeminiModel(version, self._client, ai)
            ai.define_model(
                name=googleai_name(version),
                fn=gemini_model.generate,
                metadata=gemini_model.metadata,
                config_schema=GeminiConfigSchema,
            )

        for version in GeminiEmbeddingModels:
            embedder = Embedder(version=version, client=self._client)
            ai.define_embedder(name=googleai_name(version), fn=embedder.generate)

initialize(ai)

Initialize the plugin by registering actions in the registry.

Parameters:

Name Type Description Default
ai genkit.ai.GenkitRegistry

the action registry.

 required

Returns:

Type Description
None

None
Source code in plugins/google-genai/src/genkit/plugins/google_genai/google.py 
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
def initialize(self, ai: GenkitRegistry) -> None:
    """Initialize the plugin by registering actions in the registry.

    Args:
        ai: the action registry.

    Returns:
        None
    """
    for version in GoogleAIGeminiVersion:
        gemini_model = GeminiModel(version, self._client, ai)
        ai.define_model(
            name=googleai_name(version),
            fn=gemini_model.generate,
            metadata=gemini_model.metadata,
            config_schema=GeminiConfigSchema,
        )

    for version in GeminiEmbeddingModels:
        embedder = Embedder(version=version, client=self._client)
        ai.define_embedder(name=googleai_name(version), fn=embedder.generate)

VertexAI

Bases: genkit.ai.Plugin

VertexAI plugin for Genkit.

Source code in plugins/google-genai/src/genkit/plugins/google_genai/google.py 
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
class VertexAI(Plugin):
    """VertexAI plugin for Genkit."""

    _vertexai = True

    name = VERTEXAI_PLUGIN_NAME

    def __init__(
        self,
        credentials: Credentials | None = None,
        project: str | None = None,
        location: str | None = 'us-central1',
        debug_config: DebugConfig | None = None,
        http_options: HttpOptions | HttpOptionsDict | None = None,
    ):
        self._client = genai.client.Client(
            vertexai=self._vertexai,
            api_key=None,
            credentials=credentials,
            project=project,
            location=location,
            debug_config=debug_config,
            http_options=_inject_attribution_headers(http_options),
        )

    def initialize(self, ai: GenkitRegistry) -> None:
        """Initialize the plugin by registering actions in the registry.

        Args:
            ai: the action registry.

        Returns:
            None
        """
        for version in VertexAIGeminiVersion:
            gemini_model = GeminiModel(version, self._client, ai)
            ai.define_model(
                name=vertexai_name(version),
                fn=gemini_model.generate,
                metadata=gemini_model.metadata,
                config_schema=GeminiConfigSchema,
            )

        for version in VertexEmbeddingModels:
            embedder = Embedder(version=version, client=self._client)
            ai.define_embedder(name=vertexai_name(version), fn=embedder.generate)

        for version in ImagenVersion:
            imagen_model = ImagenModel(version, self._client)
            ai.define_model(name=vertexai_name(version), fn=imagen_model.generate, metadata=imagen_model.metadata)

initialize(ai)

Initialize the plugin by registering actions in the registry.

Parameters:

Name Type Description Default
ai genkit.ai.GenkitRegistry

the action registry.

 required

Returns:

Type Description
None

None
Source code in plugins/google-genai/src/genkit/plugins/google_genai/google.py 
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
def initialize(self, ai: GenkitRegistry) -> None:
    """Initialize the plugin by registering actions in the registry.

    Args:
        ai: the action registry.

    Returns:
        None
    """
    for version in VertexAIGeminiVersion:
        gemini_model = GeminiModel(version, self._client, ai)
        ai.define_model(
            name=vertexai_name(version),
            fn=gemini_model.generate,
            metadata=gemini_model.metadata,
            config_schema=GeminiConfigSchema,
        )

    for version in VertexEmbeddingModels:
        embedder = Embedder(version=version, client=self._client)
        ai.define_embedder(name=vertexai_name(version), fn=embedder.generate)

    for version in ImagenVersion:
        imagen_model = ImagenModel(version, self._client)
        ai.define_model(name=vertexai_name(version), fn=imagen_model.generate, metadata=imagen_model.metadata)