feat(py): VertexAI options

py/packages/genkit/src/genkit/ai/model.py

@@ -2,7 +2,13 @@
 # SPDX-License-Identifier: Apache-2.0
 
 from collections.abc import Callable
+from pydantic import BaseModel
+from genkit.core.schemas import GenerateRequest, GenerateResponse, ModelInfo
 
-from genkit.core.schemas import GenerateRequest, GenerateResponse
 
 ModelFn = Callable[[GenerateRequest], GenerateResponse]
+
+
+class ModelReference(BaseModel):
+  name: str
+  info: ModelInfo

py/plugins/vertex-ai/src/genkit/plugins/vertex_ai/__init__.py

@@ -2,68 +2,9 @@
 # SPDX-License-Identifier: Apache-2.0
 
 
-"""
-Google Cloud Vertex AI Plugin for Genkit.
-"""
+"""Google Cloud Vertex AI Plugin for Genkit."""
+from genkit.plugins.vertex_ai.plugin_api import vertexAI, gemini
+from genkit.plugins.vertex_ai.options import PluginOptions
 
-from collections.abc import Callable
 
-import vertexai
-from genkit.core.schemas import (
-    GenerateRequest,
-    GenerateResponse,
-    Message,
-    Role,
-    TextPart,
-)
-from genkit.veneer.veneer import Genkit
-from vertexai.generative_models import Content, GenerativeModel, Part
-
-
-def package_name() -> str:
-    return 'genkit.plugins.vertex_ai'
-
-
-def vertex_ai(project_id: str | None = None) -> Callable[[Genkit], None]:
-    def plugin(ai: Genkit) -> None:
-        vertexai.init(location='us-central1', project=project_id)
-
-        def handle_gemini_request(request: GenerateRequest) -> GenerateResponse:
-            gemini_msgs: list[Content] = []
-            for m in request.messages:
-                gemini_parts: list[Part] = []
-                for p in m.content:
-                    if p.root.text is not None:
-                        gemini_parts.append(Part.from_text(p.root.text))
-                    else:
-                        raise Exception('unsupported part type')
-                gemini_msgs.append(
-                    Content(role=m.role.value, parts=gemini_parts)
-                )
-            model = GenerativeModel('gemini-1.5-flash-002')
-            response = model.generate_content(contents=gemini_msgs)
-            return GenerateResponse(
-                message=Message(
-                    role=Role.model, content=[TextPart(text=response.text)]
-                )
-            )
-
-        ai.define_model(
-            name='vertexai/gemini-1.5-flash',
-            fn=handle_gemini_request,
-            metadata={
-                'model': {
-                    'label': 'banana',
-                    'supports': {'multiturn': True},
-                }
-            },
-        )
-
-    return plugin
-
-
-def gemini(name: str) -> str:
-    return f'vertexai/{name}'
-
-
-__all__ = ['package_name', 'vertex_ai', 'gemini']
+__all__ = ['package_name', 'vertexAI', 'gemini', 'PluginOptions']

py/plugins/vertex-ai/src/genkit/plugins/vertex_ai/constants.py

@@ -0,0 +1,10 @@
+# Copyright 2025 Google LLC
+# SPDX-License-Identifier: Apache-2.0
+
+"""Constants for VertexAI plugin."""
+GCLOUD_PROJECT = 'GCLOUD_PROJECT'
+GCLOUD_LOCATION = 'GCLOUD_LOCATION'
+GCLOUD_PLATFORM_OAUTH_SCOPE = 'https://www.googleapis.com/auth/cloud-platform'
+GCLOUD_SERVICE_ACCOUNT_CREDS = 'GCLOUD_SERVICE_ACCOUNT_CREDS'
+FIREBASE_CONFIG = 'FIREBASE_CONFIG'
+DEFAULT_REGION = 'us-central1'

py/plugins/vertex-ai/src/genkit/plugins/vertex_ai/gemini.py

@@ -0,0 +1,140 @@
+# Copyright 2025 Google LLC
+# SPDX-License-Identifier: Apache-2.0
+
+"""Processes Gemini request."""
+import logging
+
+from vertexai.generative_models import Content, GenerativeModel, Part
+
+from genkit.ai.model import ModelReference
+from genkit.core.schemas import (
+    GenerateRequest,
+    GenerateResponse,
+    Message,
+    ModelInfo,
+    Role,
+    Supports,
+    TextPart,
+)
+
+LOG = logging.getLogger(__name__)
+
+
+# Deprecated on 2/15/2025
+SUPPORTED_V1_MODELS = {
+  'gemini-1.0-pro': ModelReference(
+        name='vertexai/gemini-1.0-pro',
+        info=ModelInfo(
+    versions=['gemini-1.0-pro-001', 'gemini-1.0-pro-002'],
+    label='Vertex AI - Gemini Pro',
+    supports=Supports(
+        multiturn=True,
+        media=False,
+        tools=True,
+        systemRole=True
+    )
+))
+}
+
+SUPPORTED_V15_MODELS = {
+    'gemini-1.5-pro': ModelReference(
+        name='vertexai/gemini-1.5-pro',
+        info=ModelInfo(
+        versions=['gemini-1.5-pro-001', 'gemini-1.5-pro-002'],
+        label='Vertex AI - Gemini 1.5 Pro',
+        supports=Supports(
+            multiturn=True,
+            media=True,
+            tools=True,
+            systemRole=True
+        )
+    )),
+    'gemini-1.5-flash': ModelReference(
+        name='vertexai/gemini-1.5-flash',
+        info=ModelInfo(
+        versions=['gemini-1.5-flash-001', 'gemini-1.5-flash-002'],
+        label='Vertex AI - Gemini 1.5 Flash',
+        supports=Supports(
+            multiturn=True,
+            media=True,
+            tools=True,
+            systemRole=True
+        )
+    )),
+    'gemini-2.0-flash-001': ModelReference(
+        name='vertexai/gemini-2.0-flash-001',
+        info=ModelInfo(
+        versions=[],
+        label='Vertex AI - Gemini 2.0 Flash 001',
+        supports=Supports(
+            multiturn=True,
+            media=True,
+            tools=True,
+            systemRole=True
+        )
+    )),
+    'gemini-2.0-flash-lite-preview-02-05': ModelReference(
+        name='vertexai/gemini-2.0-flash-lite-preview-02-05',
+        info=ModelInfo(
+        versions=[],
+        label='Vertex AI - Gemini 2.0 Flash Lite Preview 02-05',
+        supports=Supports(
+            multiturn=True,
+            media=True,
+            tools=True,
+            systemRole=True
+        )
+    )),
+    'gemini-2.0-pro-exp-02-05': ModelReference(
+        name='vertexai/gemini-2.0-pro-exp-02-05',
+        info=ModelInfo(
+        versions=[],
+        label='Vertex AI - Gemini 2.0 Flash Pro Experimental 02-05',
+        supports=Supports(
+            multiturn=True,
+            media=True,
+            tools=True,
+            systemRole=True
+        )
+    )),
+}
+
+SUPPORTED_MODELS = SUPPORTED_V1_MODELS | SUPPORTED_V15_MODELS
+
+
+def nearest_gemini_model(model_name):
+    model = SUPPORTED_MODELS.get(model_name)
+    if model:
+        return model
+    return ModelReference(
+        name=f'vertexai/{model_name}',
+        info=ModelInfo(
+        versions=[],
+        label='Vertex AI - Gemini',
+        supports=Supports(
+            multiturn=True,
+            media=True,
+            tools=True,
+            systemRole=True
+            )
+        )
+    )
+
+
+def execute_gemini_request(request: GenerateRequest) -> GenerateResponse:
+    messages: list[Content] = []
+    for msg in request.messages:
+        parts: list[Part] = []
+        for part in msg.content:
+            if hasattr(part, "text") and part.text:
+                parts.append(Part.from_text(part.text))
+            else:
+                LOG.error("Unsupported message type.")
+        messages.append(Content(role=msg.role.value, parts=parts))
+    model = GenerativeModel('gemini-1.5-flash-001')
+    response = model.generate_content(contents=messages)
+    return GenerateResponse(
+        message=Message(
+            role=Role.model, content=[TextPart(text=response.text)]
+        )
+    )

py/plugins/vertex-ai/src/genkit/plugins/vertex_ai/models/__init__.py → py/plugins/vertex-ai/src/genkit/plugins/vertex_ai/modelgarden/__init__.py

@@ -8,7 +8,7 @@
 
 
 def package_name() -> str:
-    return 'genkit.plugins.vertex_ai.models'
+    return 'genkit.plugins.vertex_ai.modelgarden'
 
 
 __all__ = ['package_name']

py/plugins/vertex-ai/src/genkit/plugins/vertex_ai/options.py

@@ -0,0 +1,67 @@
+# Copyright 2025 Google LLC
+# SPDX-License-Identifier: Apache-2.0
+
+"""Common options for plugin configuration."""
+import dataclasses
+import json
+import logging
+import os
+
+import google.oauth2.credentials as oauth2_creds
+from google.auth import credentials as auth_credentials
+
+from genkit.core.schemas import ModelInfo
+from genkit.plugins.vertex_ai import constants as const
+
+LOG = logging.getLogger(__name__)
+
+
+@dataclasses.dataclass(slots=True)
+class PluginOptions:
+    project_id: str | None = None
+    location: str | None = None
+    google_auth: auth_credentials.Credentials | None = None
+    models: list[str | ModelInfo] = dataclasses.field(default_factory=list)
+
+
+def get_project_from_firebase_config() -> str  | None:
+    config = os.getenv(const.FIREBASE_CONFIG)
+    if config:
+        try:
+            project_id = json.loads(config)['projectId']
+            return project_id
+        except json.JSONDecodeError:
+            LOG.error('Invalid JSON syntax in %s environment variable',
+                      const.FIREBASE_CONFIG)
+        except KeyError:
+            LOG.error('projectId key is not in %s environment variable',
+                      const.FIREBASE_CONFIG)
+
+    return None
+
+
+def get_plugin_parameters(options: PluginOptions | None):
+    project_id = options.project_id
+    if not project_id:
+        # The project_id retrieval order:
+        # - defined in a code
+        # - defined in GOOGLE_CLOUD_PROJECT env variable
+        # - defined in firebase config variable
+        # - defined by gcloud auth application-default login
+        #   (by VertexAI Python library)
+        project_id = (os.getenv(const.GCLOUD_PROJECT)
+                      or get_project_from_firebase_config())
+
+    location = (options.location
+                or os.getenv(const.GCLOUD_LOCATION)
+                or const.DEFAULT_REGION)
+
+    credentials = options.google_auth
+
+    sa_env = os.getenv(const.GCLOUD_SERVICE_ACCOUNT_CREDS)
+    if not credentials and sa_env:
+        # Credentials from oauth2 inherit from auth module credentials
+        credentials = oauth2_creds.Credentials.from_authorized_user_file(
+            sa_env, scopes=[const.GCLOUD_PLATFORM_OAUTH_SCOPE])
+
+    return project_id, location, credentials

py/plugins/vertex-ai/src/genkit/plugins/vertex_ai/plugin_api.py

@@ -0,0 +1,42 @@
+# Copyright 2025 Google LLC
+# SPDX-License-Identifier: Apache-2.0
+
+"""Google Cloud Vertex AI Plugin for Genkit."""
+import logging
+from collections.abc import Callable
+
+import vertexai
+
+from genkit.plugins.vertex_ai.gemini import execute_gemini_request
+from genkit.plugins.vertex_ai.options import (
+    PluginOptions,
+    get_plugin_parameters,
+)
+from genkit.veneer.veneer import Genkit
+
+LOG = logging.getLogger(__name__)
+
+
+def vertexAI(options: PluginOptions | None) -> Callable[[Genkit], None]:
+
+    def plugin(ai: Genkit) -> None:
+        project_id, location, credentials = get_plugin_parameters(options)
+        vertexai.init(project=project_id,
+                      location=location,
+                      credentials=credentials)
+
+        ai.define_model(
+            name=gemini('gemini-1.5-flash'),
+            fn=execute_gemini_request,
+            metadata={
+                'model': {
+                    'supports': {'multiturn': True},
+                }
+            },
+        )
+
+    return plugin
+
+
+def gemini(name: str) -> str:
+    return f'vertexai/{name}'