Gemini: Add option to specify config_entry in generate_content service

homeassistant/components/google_generative_ai_conversation/__init__.py

@@ -23,6 +23,7 @@
     ConfigEntryError,
     ConfigEntryNotReady,
     HomeAssistantError,
+    ServiceValidationError,
 )
 from homeassistant.helpers import config_validation as cv
 from homeassistant.helpers.issue_registry import IssueSeverity, async_create_issue
@@ -35,10 +36,12 @@
     RECOMMENDED_CHAT_MODEL,
     TIMEOUT_MILLIS,
 )
+from .model_setup import get_content_config
 
 SERVICE_GENERATE_CONTENT = "generate_content"
 CONF_IMAGE_FILENAME = "image_filename"
 CONF_FILENAMES = "filenames"
+CONF_CONFIG_ENTRY = "config_entry"
 
 CONFIG_SCHEMA = cv.config_entry_only_config_schema(DOMAIN)
 PLATFORMS = (Platform.CONVERSATION,)
@@ -64,11 +67,31 @@
                 translation_key="deprecated_image_filename_parameter",
             )
 
-        prompt_parts = [call.data[CONF_PROMPT]]
+        config_entry: GoogleGenerativeAIConfigEntry
+        if CONF_CONFIG_ENTRY in call.data:
+            entry_id = call.data[CONF_CONFIG_ENTRY]
+            found_entry = hass.config_entries.async_get_entry(entry_id)
+            if found_entry is None or found_entry.domain != DOMAIN:
+                raise ServiceValidationError(
+                    translation_domain=DOMAIN,
+                    translation_key="invalid_config_entry",
+                    translation_placeholders={"config_entry": entry_id},
+                )
+            config_entry = found_entry
+        else:
+            # Deprecated in 2025.6, to remove in 2025.10
+            async_create_issue(
+                hass,
+                DOMAIN,
+                "missing_config_entry_parameter",
+                breaks_in_ha_version="2025.10.0",
+                is_fixable=False,
+                severity=IssueSeverity.WARNING,
+                translation_key="missing_config_entry_parameter",
+            )
+            config_entry = hass.config_entries.async_loaded_entries(DOMAIN)[0]
 
-        config_entry: GoogleGenerativeAIConfigEntry = (
-            hass.config_entries.async_loaded_entries(DOMAIN)[0]
-        )
+        prompt_parts = [call.data[CONF_PROMPT]]
 
         client = config_entry.runtime_data
 
@@ -93,9 +116,15 @@
 
         await hass.async_add_executor_job(append_files_to_prompt)
 
+        model_name = config_entry.options.get(CONF_CHAT_MODEL, RECOMMENDED_CHAT_MODEL)
+
         try:
             response = await client.aio.models.generate_content(
-                model=RECOMMENDED_CHAT_MODEL, contents=prompt_parts
+                model=model_name,
+                # Features like tools and custom prompts are powered by
+                # HA's Conversation infra, so we cannot use them here.
+                config=get_content_config(config_entry),
+                contents=prompt_parts,
             )
         except (
             APIError,
@@ -120,6 +149,7 @@
         schema=vol.Schema(
             {
                 vol.Required(CONF_PROMPT): cv.string,
+                vol.Optional(CONF_CONFIG_ENTRY): cv.string,
                 vol.Optional(CONF_IMAGE_FILENAME, default=[]): vol.All(
                     cv.ensure_list, [cv.string]
                 ),

homeassistant/components/google_generative_ai_conversation/conversation.py

@@ -12,11 +12,8 @@
     AutomaticFunctionCallingConfig,
     Content,
     FunctionDeclaration,
-    GenerateContentConfig,
     GoogleSearch,
-    HarmCategory,
     Part,
-    SafetySetting,
     Schema,
     Tool,
 )
@@ -32,25 +29,13 @@
 
 from .const import (
     CONF_CHAT_MODEL,
-    CONF_DANGEROUS_BLOCK_THRESHOLD,
-    CONF_HARASSMENT_BLOCK_THRESHOLD,
-    CONF_HATE_BLOCK_THRESHOLD,
-    CONF_MAX_TOKENS,
     CONF_PROMPT,
-    CONF_SEXUAL_BLOCK_THRESHOLD,
-    CONF_TEMPERATURE,
-    CONF_TOP_K,
-    CONF_TOP_P,
     CONF_USE_GOOGLE_SEARCH_TOOL,
     DOMAIN,
     LOGGER,
     RECOMMENDED_CHAT_MODEL,
-    RECOMMENDED_HARM_BLOCK_THRESHOLD,
-    RECOMMENDED_MAX_TOKENS,
-    RECOMMENDED_TEMPERATURE,
-    RECOMMENDED_TOP_K,
-    RECOMMENDED_TOP_P,
 )
+from .model_setup import get_content_config
 
 # Max number of back and forth with the LLM to generate a response
 MAX_TOOL_ITERATIONS = 10
@@ -371,50 +356,16 @@ async def _async_handle_message(
 
         if tool_results:
             messages.append(_create_google_tool_response_content(tool_results))
-        generateContentConfig = GenerateContentConfig(
-            temperature=self.entry.options.get(
-                CONF_TEMPERATURE, RECOMMENDED_TEMPERATURE
-            ),
-            top_k=self.entry.options.get(CONF_TOP_K, RECOMMENDED_TOP_K),
-            top_p=self.entry.options.get(CONF_TOP_P, RECOMMENDED_TOP_P),
-            max_output_tokens=self.entry.options.get(
-                CONF_MAX_TOKENS, RECOMMENDED_MAX_TOKENS
-            ),
-            safety_settings=[
-                SafetySetting(
-                    category=HarmCategory.HARM_CATEGORY_HATE_SPEECH,
-                    threshold=self.entry.options.get(
-                        CONF_HATE_BLOCK_THRESHOLD, RECOMMENDED_HARM_BLOCK_THRESHOLD
-                    ),
-                ),
-                SafetySetting(
-                    category=HarmCategory.HARM_CATEGORY_HARASSMENT,
-                    threshold=self.entry.options.get(
-                        CONF_HARASSMENT_BLOCK_THRESHOLD,
-                        RECOMMENDED_HARM_BLOCK_THRESHOLD,
-                    ),
-                ),
-                SafetySetting(
-                    category=HarmCategory.HARM_CATEGORY_DANGEROUS_CONTENT,
-                    threshold=self.entry.options.get(
-                        CONF_DANGEROUS_BLOCK_THRESHOLD, RECOMMENDED_HARM_BLOCK_THRESHOLD
-                    ),
-                ),
-                SafetySetting(
-                    category=HarmCategory.HARM_CATEGORY_SEXUALLY_EXPLICIT,
-                    threshold=self.entry.options.get(
-                        CONF_SEXUAL_BLOCK_THRESHOLD, RECOMMENDED_HARM_BLOCK_THRESHOLD
-                    ),
-                ),
-            ],
-            tools=tools or None,
-            system_instruction=prompt if supports_system_instruction else None,
-            automatic_function_calling=AutomaticFunctionCallingConfig(
-                disable=True, maximum_remote_calls=None
-            ),
-        )
 
-        if not supports_system_instruction:
+        generateContentConfig = get_content_config(self.entry)
+        # Set additional config options that are only supported in conversation.
+        generateContentConfig.tools = tools
+        generateContentConfig.automatic_function_calling = (
+            AutomaticFunctionCallingConfig(disable=True, maximum_remote_calls=None)
+        )
+        if supports_system_instruction:
+            generateContentConfig.system_instruction = prompt
+        else:
             messages = [
                 Content(role="user", parts=[Part.from_text(text=prompt)]),
                 Content(role="model", parts=[Part.from_text(text="Ok")]),

homeassistant/components/google_generative_ai_conversation/model_setup.py

@@ -0,0 +1,64 @@
+"""Generic helper functions for setting up Gemini models."""
+
+from google.genai import Client
+from google.genai.types import GenerateContentConfig, HarmCategory, SafetySetting
+
+from homeassistant.config_entries import ConfigEntry
+
+from .const import (
+    CONF_DANGEROUS_BLOCK_THRESHOLD,
+    CONF_HARASSMENT_BLOCK_THRESHOLD,
+    CONF_HATE_BLOCK_THRESHOLD,
+    CONF_MAX_TOKENS,
+    CONF_SEXUAL_BLOCK_THRESHOLD,
+    CONF_TEMPERATURE,
+    CONF_TOP_K,
+    CONF_TOP_P,
+    RECOMMENDED_HARM_BLOCK_THRESHOLD,
+    RECOMMENDED_MAX_TOKENS,
+    RECOMMENDED_TEMPERATURE,
+    RECOMMENDED_TOP_K,
+    RECOMMENDED_TOP_P,
+)
+
+type GoogleGenerativeAIConfigEntry = ConfigEntry[Client]
+
+
+def get_content_config(
+    entry: GoogleGenerativeAIConfigEntry,
+) -> GenerateContentConfig:
+    """Create parameters for Gemini model inputs from a config entry."""
+
+    return GenerateContentConfig(
+        temperature=entry.options.get(CONF_TEMPERATURE, RECOMMENDED_TEMPERATURE),
+        top_k=entry.options.get(CONF_TOP_K, RECOMMENDED_TOP_K),
+        top_p=entry.options.get(CONF_TOP_P, RECOMMENDED_TOP_P),
+        max_output_tokens=entry.options.get(CONF_MAX_TOKENS, RECOMMENDED_MAX_TOKENS),
+        safety_settings=[
+            SafetySetting(
+                category=HarmCategory.HARM_CATEGORY_HATE_SPEECH,
+                threshold=entry.options.get(
+                    CONF_HATE_BLOCK_THRESHOLD, RECOMMENDED_HARM_BLOCK_THRESHOLD
+                ),
+            ),
+            SafetySetting(
+                category=HarmCategory.HARM_CATEGORY_HARASSMENT,
+                threshold=entry.options.get(
+                    CONF_HARASSMENT_BLOCK_THRESHOLD,
+                    RECOMMENDED_HARM_BLOCK_THRESHOLD,
+                ),
+            ),
+            SafetySetting(
+                category=HarmCategory.HARM_CATEGORY_DANGEROUS_CONTENT,
+                threshold=entry.options.get(
+                    CONF_DANGEROUS_BLOCK_THRESHOLD, RECOMMENDED_HARM_BLOCK_THRESHOLD
+                ),
+            ),
+            SafetySetting(
+                category=HarmCategory.HARM_CATEGORY_SEXUALLY_EXPLICIT,
+                threshold=entry.options.get(
+                    CONF_SEXUAL_BLOCK_THRESHOLD, RECOMMENDED_HARM_BLOCK_THRESHOLD
+                ),
+            ),
+        ],
+    )

homeassistant/components/google_generative_ai_conversation/services.yaml

@@ -1,5 +1,10 @@
 generate_content:
   fields:
+    config_entry:
+      required: true
+      selector:
+        config_entry:
+          integration: google_generative_ai_conversation
     prompt:
       required: true
       selector:

homeassistant/components/google_generative_ai_conversation/strings.json

@@ -54,6 +54,10 @@
       "name": "Generate content",
       "description": "Generate content from a prompt consisting of text and optionally images",
       "fields": {
+        "config_entry": {
+          "name": "Config entry",
+          "description": "The config entry specifying model settings to use for this action"
+        },
         "prompt": {
           "name": "Prompt",
           "description": "The prompt",
@@ -76,6 +80,15 @@
     "deprecated_image_filename_parameter": {
       "title": "Deprecated 'image_filename' parameter",
       "description": "The 'image_filename' parameter in Google Generative AI actions is deprecated. Please edit scripts and automations to use 'filenames' instead."
+    },
+    "missing_config_entry_parameter": {
+      "title": "Missing 'Config entry' parameter",
+      "description": "The 'Config entry' parameter in Google Generative AI actions is required. Please edit scripts and automations to specify a config entry."
+    }
+  },
+  "exceptions": {
+    "invalid_config_entry": {
+      "message": "Invalid config entry `{config_entry}`"
     }
   }
 }

tests/components/google_generative_ai_conversation/snapshots/test_init.ambr

@@ -6,6 +6,7 @@
       tuple(
       ),
       dict({
+        'config': GenerateContentConfig(http_options=None, system_instruction=None, temperature=1.0, top_p=0.95, top_k=64.0, candidate_count=None, max_output_tokens=1500, stop_sequences=None, response_logprobs=None, logprobs=None, presence_penalty=None, frequency_penalty=None, seed=None, response_mime_type=None, response_schema=None, routing_config=None, safety_settings=[SafetySetting(method=None, category=<HarmCategory.HARM_CATEGORY_HATE_SPEECH: 'HARM_CATEGORY_HATE_SPEECH'>, threshold=<HarmBlockThreshold.BLOCK_MEDIUM_AND_ABOVE: 'BLOCK_MEDIUM_AND_ABOVE'>), SafetySetting(method=None, category=<HarmCategory.HARM_CATEGORY_HARASSMENT: 'HARM_CATEGORY_HARASSMENT'>, threshold=<HarmBlockThreshold.BLOCK_MEDIUM_AND_ABOVE: 'BLOCK_MEDIUM_AND_ABOVE'>), SafetySetting(method=None, category=<HarmCategory.HARM_CATEGORY_DANGEROUS_CONTENT: 'HARM_CATEGORY_DANGEROUS_CONTENT'>, threshold=<HarmBlockThreshold.BLOCK_MEDIUM_AND_ABOVE: 'BLOCK_MEDIUM_AND_ABOVE'>), SafetySetting(method=None, category=<HarmCategory.HARM_CATEGORY_SEXUALLY_EXPLICIT: 'HARM_CATEGORY_SEXUALLY_EXPLICIT'>, threshold=<HarmBlockThreshold.BLOCK_MEDIUM_AND_ABOVE: 'BLOCK_MEDIUM_AND_ABOVE'>)], tools=None, tool_config=None, labels=None, cached_content=None, response_modalities=None, media_resolution=None, speech_config=None, audio_timestamp=None, automatic_function_calling=None, thinking_config=None),
         'contents': list([
           'Describe this image from my doorbell camera',
           b'some file',
@@ -23,10 +24,11 @@
       tuple(
       ),
       dict({
+        'config': GenerateContentConfig(http_options=None, system_instruction=None, temperature=1.0, top_p=0.95, top_k=64.0, candidate_count=None, max_output_tokens=1500, stop_sequences=None, response_logprobs=None, logprobs=None, presence_penalty=None, frequency_penalty=None, seed=None, response_mime_type=None, response_schema=None, routing_config=None, safety_settings=[SafetySetting(method=None, category=<HarmCategory.HARM_CATEGORY_HATE_SPEECH: 'HARM_CATEGORY_HATE_SPEECH'>, threshold=<HarmBlockThreshold.BLOCK_MEDIUM_AND_ABOVE: 'BLOCK_MEDIUM_AND_ABOVE'>), SafetySetting(method=None, category=<HarmCategory.HARM_CATEGORY_HARASSMENT: 'HARM_CATEGORY_HARASSMENT'>, threshold=<HarmBlockThreshold.BLOCK_MEDIUM_AND_ABOVE: 'BLOCK_MEDIUM_AND_ABOVE'>), SafetySetting(method=None, category=<HarmCategory.HARM_CATEGORY_DANGEROUS_CONTENT: 'HARM_CATEGORY_DANGEROUS_CONTENT'>, threshold=<HarmBlockThreshold.BLOCK_MEDIUM_AND_ABOVE: 'BLOCK_MEDIUM_AND_ABOVE'>), SafetySetting(method=None, category=<HarmCategory.HARM_CATEGORY_SEXUALLY_EXPLICIT: 'HARM_CATEGORY_SEXUALLY_EXPLICIT'>, threshold=<HarmBlockThreshold.BLOCK_MEDIUM_AND_ABOVE: 'BLOCK_MEDIUM_AND_ABOVE'>)], tools=None, tool_config=None, labels=None, cached_content=None, response_modalities=None, media_resolution=None, speech_config=None, audio_timestamp=None, automatic_function_calling=None, thinking_config=None),
         'contents': list([
           'Write an opening speech for a Home Assistant release party',
         ]),
-        'model': 'models/gemini-2.0-flash',
+        'model': 'fake-test-model',
       }),
     ),
   ])
@@ -38,6 +40,7 @@
       tuple(
       ),
       dict({
+        'config': GenerateContentConfig(http_options=None, system_instruction=None, temperature=1.0, top_p=0.95, top_k=64.0, candidate_count=None, max_output_tokens=1500, stop_sequences=None, response_logprobs=None, logprobs=None, presence_penalty=None, frequency_penalty=None, seed=None, response_mime_type=None, response_schema=None, routing_config=None, safety_settings=[SafetySetting(method=None, category=<HarmCategory.HARM_CATEGORY_HATE_SPEECH: 'HARM_CATEGORY_HATE_SPEECH'>, threshold=<HarmBlockThreshold.BLOCK_MEDIUM_AND_ABOVE: 'BLOCK_MEDIUM_AND_ABOVE'>), SafetySetting(method=None, category=<HarmCategory.HARM_CATEGORY_HARASSMENT: 'HARM_CATEGORY_HARASSMENT'>, threshold=<HarmBlockThreshold.BLOCK_MEDIUM_AND_ABOVE: 'BLOCK_MEDIUM_AND_ABOVE'>), SafetySetting(method=None, category=<HarmCategory.HARM_CATEGORY_DANGEROUS_CONTENT: 'HARM_CATEGORY_DANGEROUS_CONTENT'>, threshold=<HarmBlockThreshold.BLOCK_MEDIUM_AND_ABOVE: 'BLOCK_MEDIUM_AND_ABOVE'>), SafetySetting(method=None, category=<HarmCategory.HARM_CATEGORY_SEXUALLY_EXPLICIT: 'HARM_CATEGORY_SEXUALLY_EXPLICIT'>, threshold=<HarmBlockThreshold.BLOCK_MEDIUM_AND_ABOVE: 'BLOCK_MEDIUM_AND_ABOVE'>)], tools=None, tool_config=None, labels=None, cached_content=None, response_modalities=None, media_resolution=None, speech_config=None, audio_timestamp=None, automatic_function_calling=None, thinking_config=None),
         'contents': list([
           'Write an opening speech for a Home Assistant release party',
         ]),

tests/components/google_generative_ai_conversation/test_init.py

@@ -6,6 +6,7 @@
 from requests.exceptions import Timeout
 from syrupy.assertion import SnapshotAssertion
 
+from homeassistant.components.google_generative_ai_conversation import CONF_CHAT_MODEL
 from homeassistant.config_entries import ConfigEntryState
 from homeassistant.core import HomeAssistant
 from homeassistant.exceptions import HomeAssistantError
@@ -20,6 +21,23 @@ async def test_generate_content_service_without_images(
     hass: HomeAssistant, snapshot: SnapshotAssertion
 ) -> None:
     """Test generate content service."""
+    SAMPLE_MODEL = "fake-test-model"
+
+    second_entry = MockConfigEntry(
+        domain="google_generative_ai_conversation",
+        title="Google Generative AI Conversation",
+        data={
+            "api_key": "bla",
+        },
+        options={
+            CONF_CHAT_MODEL: SAMPLE_MODEL,
+        },
+    )
+    second_entry.add_to_hass(hass)
+    with patch("google.genai.models.AsyncModels.get"):
+        await hass.config_entries.async_setup(second_entry.entry_id)
+        await hass.async_block_till_done()
+
     stubbed_generated_content = (
         "I'm thrilled to welcome you all to the release "
         "party for the latest version of Home Assistant!"
@@ -36,14 +54,18 @@ async def test_generate_content_service_without_images(
         response = await hass.services.async_call(
             "google_generative_ai_conversation",
             "generate_content",
-            {"prompt": "Write an opening speech for a Home Assistant release party"},
+            {
+                "config_entry": second_entry.entry_id,
+                "prompt": "Write an opening speech for a Home Assistant release party",
+            },
             blocking=True,
             return_response=True,
         )
 
     assert response == {
         "text": stubbed_generated_content,
     }
+    assert mock_generate.call_args.kwargs["model"] == SAMPLE_MODEL
     assert [tuple(mock_call) for mock_call in mock_generate.mock_calls] == snapshot