homeassistant 2025.9.3__py3-none-any.whl → 2025.10.0__py3-none-any.whl

homeassistant/components/google_generative_ai_conversation/ai_task.py

@@ -3,6 +3,10 @@
 from __future__ import annotations
 
 from json import JSONDecodeError
+from typing import TYPE_CHECKING
+
+from google.genai.errors import APIError
+from google.genai.types import GenerateContentConfig, Part, PartUnionDict
 
 from homeassistant.components import ai_task, conversation
 from homeassistant.config_entries import ConfigEntry
@@ -11,8 +15,17 @@ from homeassistant.exceptions import HomeAssistantError
 from homeassistant.helpers.entity_platform import AddConfigEntryEntitiesCallback
 from homeassistant.util.json import json_loads
 
-from .const import LOGGER
-from .entity import ERROR_GETTING_RESPONSE, GoogleGenerativeAILLMBaseEntity
+from .const import CONF_CHAT_MODEL, CONF_RECOMMENDED, LOGGER, RECOMMENDED_IMAGE_MODEL
+from .entity import (
+    ERROR_GETTING_RESPONSE,
+    GoogleGenerativeAILLMBaseEntity,
+    async_prepare_files_for_prompt,
+)
+
+if TYPE_CHECKING:
+    from homeassistant.config_entries import ConfigSubentry
+
+    from . import GoogleGenerativeAIConfigEntry
 
 
 async def async_setup_entry(
@@ -37,10 +50,22 @@ class GoogleGenerativeAITaskEntity(
 ):
     """Google Generative AI AI Task entity."""
 
-    _attr_supported_features = (
-        ai_task.AITaskEntityFeature.GENERATE_DATA
-        | ai_task.AITaskEntityFeature.SUPPORT_ATTACHMENTS
-    )
+    def __init__(
+        self,
+        entry: GoogleGenerativeAIConfigEntry,
+        subentry: ConfigSubentry,
+    ) -> None:
+        """Initialize the entity."""
+        super().__init__(entry, subentry)
+        self._attr_supported_features = (
+            ai_task.AITaskEntityFeature.GENERATE_DATA
+            | ai_task.AITaskEntityFeature.SUPPORT_ATTACHMENTS
+        )
+
+        if subentry.data.get(CONF_RECOMMENDED) or "-image" in subentry.data.get(
+            CONF_CHAT_MODEL, ""
+        ):
+            self._attr_supported_features |= ai_task.AITaskEntityFeature.GENERATE_IMAGE
 
     async def _async_generate_data(
         self,
@@ -79,3 +104,89 @@ class GoogleGenerativeAITaskEntity(
             conversation_id=chat_log.conversation_id,
             data=data,
         )
+
+    async def _async_generate_image(
+        self,
+        task: ai_task.GenImageTask,
+        chat_log: conversation.ChatLog,
+    ) -> ai_task.GenImageTaskResult:
+        """Handle a generate image task."""
+        # Get the user prompt from the chat log
+        user_message = chat_log.content[-1]
+        assert isinstance(user_message, conversation.UserContent)
+
+        model = self.subentry.data.get(CONF_CHAT_MODEL, RECOMMENDED_IMAGE_MODEL)
+        prompt_parts: list[PartUnionDict] = [user_message.content]
+        if user_message.attachments:
+            prompt_parts.extend(
+                await async_prepare_files_for_prompt(
+                    self.hass,
+                    self._genai_client,
+                    [(a.path, a.mime_type) for a in user_message.attachments],
+                )
+            )
+
+        try:
+            response = await self._genai_client.aio.models.generate_content(
+                model=model,
+                contents=prompt_parts,
+                config=GenerateContentConfig(
+                    response_modalities=["TEXT", "IMAGE"],
+                ),
+            )
+        except (APIError, ValueError) as err:
+            LOGGER.error("Error generating image: %s", err)
+            raise HomeAssistantError(f"Error generating image: {err}") from err
+
+        if response.prompt_feedback:
+            raise HomeAssistantError(
+                f"Error generating content due to content violations, reason: {response.prompt_feedback.block_reason_message}"
+            )
+
+        if (
+            not response.candidates
+            or not response.candidates[0].content
+            or not response.candidates[0].content.parts
+        ):
+            raise HomeAssistantError("Unknown error generating image")
+
+        # Parse response
+        response_text = ""
+        response_image: Part | None = None
+        for part in response.candidates[0].content.parts:
+            if (
+                part.inline_data
+                and part.inline_data.data
+                and part.inline_data.mime_type
+                and part.inline_data.mime_type.startswith("image/")
+            ):
+                if response_image is None:
+                    response_image = part
+                else:
+                    LOGGER.warning("Prompt generated multiple images")
+            elif isinstance(part.text, str) and not part.thought:
+                response_text += part.text
+
+        if response_image is None:
+            raise HomeAssistantError("Response did not include image")
+
+        assert response_image.inline_data is not None
+        assert response_image.inline_data.data is not None
+        assert response_image.inline_data.mime_type is not None
+
+        image_data = response_image.inline_data.data
+        mime_type = response_image.inline_data.mime_type
+
+        chat_log.async_add_assistant_content_without_tools(
+            conversation.AssistantContent(
+                agent_id=self.entity_id,
+                content=response_text,
+            )
+        )
+
+        return ai_task.GenImageTaskResult(
+            image_data=image_data,
+            conversation_id=chat_log.conversation_id,
+            mime_type=mime_type,
+            model=model.partition("/")[-1],
+        )

homeassistant/components/google_generative_ai_conversation/const.py

@@ -23,6 +23,7 @@ CONF_CHAT_MODEL = "chat_model"
 RECOMMENDED_CHAT_MODEL = "models/gemini-2.5-flash"
 RECOMMENDED_STT_MODEL = RECOMMENDED_CHAT_MODEL
 RECOMMENDED_TTS_MODEL = "models/gemini-2.5-flash-preview-tts"
+RECOMMENDED_IMAGE_MODEL = "models/gemini-2.5-flash-image-preview"
 CONF_TEMPERATURE = "temperature"
 RECOMMENDED_TEMPERATURE = 1.0
 CONF_TOP_P = "top_p"

homeassistant/components/google_generative_ai_conversation/entity.py

@@ -3,12 +3,13 @@
 from __future__ import annotations
 
 import asyncio
+import base64
 import codecs
 from collections.abc import AsyncGenerator, AsyncIterator, Callable
-from dataclasses import replace
+from dataclasses import dataclass, replace
 import mimetypes
 from pathlib import Path
-from typing import TYPE_CHECKING, Any, cast
+from typing import TYPE_CHECKING, Any, Literal, cast
 
 from google.genai import Client
 from google.genai.errors import APIError, ClientError
@@ -27,6 +28,7 @@ from google.genai.types import (
     PartUnionDict,
     SafetySetting,
     Schema,
+    ThinkingConfig,
     Tool,
     ToolListUnion,
 )
@@ -201,6 +203,30 @@ def _create_google_tool_response_content(
     )
 
 
+@dataclass(slots=True)
+class PartDetails:
+    """Additional data for a content part."""
+
+    part_type: Literal["text", "thought", "function_call"]
+    """The part type for which this data is relevant for."""
+
+    index: int
+    """Start position or number of the tool."""
+
+    length: int = 0
+    """Length of the relevant data."""
+
+    thought_signature: str | None = None
+    """Base64 encoded thought signature, if available."""
+
+
+@dataclass(slots=True)
+class ContentDetails:
+    """Native data for AssistantContent."""
+
+    part_details: list[PartDetails]
+
+
 def _convert_content(
     content: (
         conversation.UserContent
@@ -209,32 +235,91 @@ def _convert_content(
     ),
 ) -> Content:
     """Convert HA content to Google content."""
-    if content.role != "assistant" or not content.tool_calls:
-        role = "model" if content.role == "assistant" else content.role
+    if content.role != "assistant":
         return Content(
-            role=role,
-            parts=[
-                Part.from_text(text=content.content if content.content else ""),
-            ],
+            role=content.role,
+            parts=[Part.from_text(text=content.content if content.content else "")],
         )
 
     # Handle the Assistant content with tool calls.
     assert type(content) is conversation.AssistantContent
     parts: list[Part] = []
+    part_details: list[PartDetails] = (
+        content.native.part_details
+        if isinstance(content.native, ContentDetails)
+        else []
+    )
+    details: PartDetails | None = None
 
     if content.content:
-        parts.append(Part.from_text(text=content.content))
+        index = 0
+        for details in part_details:
+            if details.part_type == "text":
+                if index < details.index:
+                    parts.append(
+                        Part.from_text(text=content.content[index : details.index])
+                    )
+                    index = details.index
+                parts.append(
+                    Part.from_text(
+                        text=content.content[index : index + details.length],
+                    )
+                )
+                if details.thought_signature:
+                    parts[-1].thought_signature = base64.b64decode(
+                        details.thought_signature
+                    )
+                index += details.length
+        if index < len(content.content):
+            parts.append(Part.from_text(text=content.content[index:]))
+
+    if content.thinking_content:
+        index = 0
+        for details in part_details:
+            if details.part_type == "thought":
+                if index < details.index:
+                    parts.append(
+                        Part.from_text(
+                            text=content.thinking_content[index : details.index]
+                        )
+                    )
+                    parts[-1].thought = True
+                    index = details.index
+                parts.append(
+                    Part.from_text(
+                        text=content.thinking_content[index : index + details.length],
+                    )
+                )
+                parts[-1].thought = True
+                if details.thought_signature:
+                    parts[-1].thought_signature = base64.b64decode(
+                        details.thought_signature
+                    )
+                index += details.length
+        if index < len(content.thinking_content):
+            parts.append(Part.from_text(text=content.thinking_content[index:]))
+            parts[-1].thought = True
 
     if content.tool_calls:
-        parts.extend(
-            [
+        for index, tool_call in enumerate(content.tool_calls):
+            parts.append(
                 Part.from_function_call(
                     name=tool_call.tool_name,
                     args=_escape_decode(tool_call.tool_args),
                 )
-                for tool_call in content.tool_calls
-            ]
-        )
+            )
+            if details := next(
+                (
+                    d
+                    for d in part_details
+                    if d.part_type == "function_call" and d.index == index
+                ),
+                None,
+            ):
+                if details.thought_signature:
+                    parts[-1].thought_signature = base64.b64decode(
+                        details.thought_signature
+                    )
 
     return Content(role="model", parts=parts)
 
@@ -243,14 +328,20 @@ async def _transform_stream(
     result: AsyncIterator[GenerateContentResponse],
 ) -> AsyncGenerator[conversation.AssistantContentDeltaDict]:
     new_message = True
+    part_details: list[PartDetails] = []
     try:
         async for response in result:
             LOGGER.debug("Received response chunk: %s", response)
-            chunk: conversation.AssistantContentDeltaDict = {}
 
             if new_message:
-                chunk["role"] = "assistant"
+                if part_details:
+                    yield {"native": ContentDetails(part_details=part_details)}
+                    part_details = []
+                yield {"role": "assistant"}
                 new_message = False
+                content_index = 0
+                thinking_content_index = 0
+                tool_call_index = 0
 
             # According to the API docs, this would mean no candidate is returned, so we can safely throw an error here.
             if response.prompt_feedback or not response.candidates:
@@ -284,23 +375,62 @@ async def _transform_stream(
                 else []
             )
 
-            content = "".join([part.text for part in response_parts if part.text])
-            tool_calls = []
             for part in response_parts:
-                if not part.function_call:
-                    continue
-                tool_call = part.function_call
-                tool_name = tool_call.name if tool_call.name else ""
-                tool_args = _escape_decode(tool_call.args)
-                tool_calls.append(
-                    llm.ToolInput(tool_name=tool_name, tool_args=tool_args)
-                )
+                chunk: conversation.AssistantContentDeltaDict = {}
+
+                if part.text:
+                    if part.thought:
+                        chunk["thinking_content"] = part.text
+                        if part.thought_signature:
+                            part_details.append(
+                                PartDetails(
+                                    part_type="thought",
+                                    index=thinking_content_index,
+                                    length=len(part.text),
+                                    thought_signature=base64.b64encode(
+                                        part.thought_signature
+                                    ).decode("utf-8"),
+                                )
+                            )
+                        thinking_content_index += len(part.text)
+                    else:
+                        chunk["content"] = part.text
+                        if part.thought_signature:
+                            part_details.append(
+                                PartDetails(
+                                    part_type="text",
+                                    index=content_index,
+                                    length=len(part.text),
+                                    thought_signature=base64.b64encode(
+                                        part.thought_signature
+                                    ).decode("utf-8"),
+                                )
+                            )
+                        content_index += len(part.text)
+
+                if part.function_call:
+                    tool_call = part.function_call
+                    tool_name = tool_call.name if tool_call.name else ""
+                    tool_args = _escape_decode(tool_call.args)
+                    chunk["tool_calls"] = [
+                        llm.ToolInput(tool_name=tool_name, tool_args=tool_args)
+                    ]
+                    if part.thought_signature:
+                        part_details.append(
+                            PartDetails(
+                                part_type="function_call",
+                                index=tool_call_index,
+                                thought_signature=base64.b64encode(
+                                    part.thought_signature
+                                ).decode("utf-8"),
+                            )
+                        )
+
+                yield chunk
 
-            if tool_calls:
-                chunk["tool_calls"] = tool_calls
+        if part_details:
+            yield {"native": ContentDetails(part_details=part_details)}
 
-            chunk["content"] = content
-            yield chunk
     except (
         APIError,
         ValueError,
@@ -448,12 +578,13 @@ class GoogleGenerativeAILLMBaseEntity(Entity):
         assert isinstance(user_message, conversation.UserContent)
         chat_request: list[PartUnionDict] = [user_message.content]
         if user_message.attachments:
-            files = await async_prepare_files_for_prompt(
-                self.hass,
-                self._genai_client,
-                [a.path for a in user_message.attachments],
+            chat_request.extend(
+                await async_prepare_files_for_prompt(
+                    self.hass,
+                    self._genai_client,
+                    [(a.path, a.mime_type) for a in user_message.attachments],
+                )
             )
-            chat_request = [*chat_request, *files]
 
         # To prevent infinite loops, we limit the number of iterations
         for _iteration in range(MAX_TOOL_ITERATIONS):
@@ -521,11 +652,12 @@ class GoogleGenerativeAILLMBaseEntity(Entity):
                     ),
                 ),
             ],
+            thinking_config=ThinkingConfig(include_thoughts=True),
         )
 
 
 async def async_prepare_files_for_prompt(
-    hass: HomeAssistant, client: Client, files: list[Path]
+    hass: HomeAssistant, client: Client, files: list[tuple[Path, str | None]]
 ) -> list[File]:
     """Upload files so they can be attached to a prompt.
 
@@ -534,10 +666,11 @@ async def async_prepare_files_for_prompt(
 
     def upload_files() -> list[File]:
         prompt_parts: list[File] = []
-        for filename in files:
+        for filename, mimetype in files:
             if not filename.exists():
                 raise HomeAssistantError(f"`{filename}` does not exist")
-            mimetype = mimetypes.guess_type(filename)[0]
+            if mimetype is None:
+                mimetype = mimetypes.guess_type(filename)[0]
             prompt_parts.append(
                 client.files.upload(
                     file=filename,

homeassistant/components/google_generative_ai_conversation/manifest.json

@@ -1,6 +1,6 @@
 {
   "domain": "google_generative_ai_conversation",
-  "name": "Google Generative AI",
+  "name": "Google Gemini",
   "after_dependencies": ["assist_pipeline", "intent"],
   "codeowners": ["@tronikos", "@ivanlh"],
   "config_flow": true,
@@ -8,5 +8,5 @@
   "documentation": "https://www.home-assistant.io/integrations/google_generative_ai_conversation",
   "integration_type": "service",
   "iot_class": "cloud_polling",
-  "requirements": ["google-genai==1.29.0"]
+  "requirements": ["google-genai==1.38.0"]
 }

homeassistant/components/google_generative_ai_conversation/services.yaml

@@ -5,10 +5,6 @@ generate_content:
       selector:
         text:
           multiline: true
-    image_filename:
-      required: false
-      selector:
-        object:
     filenames:
       required: false
       selector:

homeassistant/components/google_generative_ai_conversation/strings.json

@@ -150,21 +150,22 @@
       }
     }
   },
+  "issues": {
+    "deprecated_generate_content": {
+      "title": "Deprecated 'generate_content' action",
+      "description": "Action 'google_generative_ai_conversation.generate_content' is deprecated and will be removed in the 2026.4.0 release. Please use the 'ai_task.generate_data' action instead"
+    }
+  },
   "services": {
     "generate_content": {
-      "name": "Generate content",
-      "description": "Generate content from a prompt consisting of text and optionally images",
+      "name": "Generate content (deprecated)",
+      "description": "Generate content from a prompt consisting of text and optionally images (deprecated)",
       "fields": {
         "prompt": {
           "name": "Prompt",
           "description": "The prompt",
           "example": "Describe what you see in these images"
         },
-        "image_filename": {
-          "name": "Image filename",
-          "description": "Deprecated. Use filenames instead.",
-          "example": "/config/www/image.jpg"
-        },
         "filenames": {
           "name": "Attachment filenames",
           "description": "Attachments to add to the prompt (images, PDFs, etc)",
@@ -172,11 +173,5 @@
         }
       }
     }
-  },
-  "issues": {
-    "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."
-    }
   }
 }

homeassistant/components/google_generative_ai_conversation/translations/bg.json

@@ -102,15 +102,10 @@
                 "filenames": {
                     "example": "/config/www/image.jpg"
                 },
-                "image_filename": {
-                    "example": "/config/www/image.jpg",
-                    "name": "\u0418\u043c\u0435 \u043d\u0430 \u0444\u0430\u0439\u043b\u0430 \u0441 \u0438\u0437\u043e\u0431\u0440\u0430\u0436\u0435\u043d\u0438\u0435\u0442\u043e"
-                },
                 "prompt": {
                     "example": "\u041e\u043f\u0438\u0448\u0435\u0442\u0435 \u043a\u0430\u043a\u0432\u043e \u0432\u0438\u0436\u0434\u0430\u0442\u0435 \u043d\u0430 \u0442\u0435\u0437\u0438 \u0438\u0437\u043e\u0431\u0440\u0430\u0436\u0435\u043d\u0438\u044f"
                 }
-            },
-            "name": "\u0413\u0435\u043d\u0435\u0440\u0438\u0440\u0430\u043d\u0435 \u043d\u0430 \u0441\u044a\u0434\u044a\u0440\u0436\u0430\u043d\u0438\u0435"
+            }
         }
     }
 }

homeassistant/components/google_generative_ai_conversation/translations/ca.json

@@ -84,16 +84,11 @@
     "services": {
         "generate_content": {
             "fields": {
-                "image_filename": {
-                    "example": "/config/www/imatge.jpg",
-                    "name": "Nom de la imatge"
-                },
                 "prompt": {
                     "description": "Instrucci\u00f3",
                     "name": "Instrucci\u00f3"
                 }
-            },
-            "name": "Genera contingut"
+            }
         }
     }
 }

homeassistant/components/google_generative_ai_conversation/translations/cs.json

@@ -151,32 +151,27 @@
         }
     },
     "issues": {
-        "deprecated_image_filename_parameter": {
-            "description": "Parametr \"image_filename\" v \u200bakc\u00edch Google Generative AI je zastaral\u00fd. Upravte skripty a automatizace, aby m\u00edsto toho pou\u017e\u00edvaly \"filenames\".",
-            "title": "Zastaral\u00fd parametr \"image_filename\""
+        "deprecated_generate_content": {
+            "description": "Akce \"google_generative_ai_conversation.generate_content\" je zastaral\u00e1 a bude oebr\u00e1na ve verzi 2026.4.0. Pou\u017eijte m\u00edsto n\u00ed akci \"ai_task.generate_data\".",
+            "title": "Zastaral\u00e1 akce \"generate_content\""
         }
     },
     "services": {
         "generate_content": {
-            "description": "Vygeneruje obsah z dotazu sest\u00e1vaj\u00edc\u00ed z textu a voliteln\u011b obr\u00e1zk\u016f.",
+            "description": "Vygeneruje obsah z dotazu sest\u00e1vaj\u00edc\u00ed z textu a voliteln\u011b obr\u00e1zk\u016f (zastaral\u00e9).",
             "fields": {
                 "filenames": {
                     "description": "P\u0159\u00edlohy, kter\u00e9 se maj\u00ed p\u0159idat do dotazu (obr\u00e1zky, soubory PDF atd.)",
                     "example": "/config/www/image.jpg",
                     "name": "N\u00e1zvy soubor\u016f p\u0159\u00edloh"
                 },
-                "image_filename": {
-                    "description": "Zastaral\u00e9. M\u00edsto toho pou\u017eijte n\u00e1zvy soubor\u016f.",
-                    "example": "/config/www/image.jpg",
-                    "name": "N\u00e1zev souboru obr\u00e1zku"
-                },
                 "prompt": {
                     "description": "Dotaz",
                     "example": "Popi\u0161te, co vid\u00edte na t\u011bchto obr\u00e1zc\u00edch",
                     "name": "Dotaz"
                 }
             },
-            "name": "Vygenerovat obsah"
+            "name": "Vygenerovat obsah (zastaral\u00e9)"
         }
     }
 }

homeassistant/components/google_generative_ai_conversation/translations/de.json

@@ -151,32 +151,27 @@
         }
     },
     "issues": {
-        "deprecated_image_filename_parameter": {
-            "description": "Der Parameter \u201eimage_filename\u201c in Aktionen von Google Generative AI ist veraltet. Bitte bearbeite Skripte und Automationen, um stattdessen \u201efilenames\u201c zu verwenden.",
-            "title": "Veralteter Parameter \u201eimage_filename\u201c"
+        "deprecated_generate_content": {
+            "description": "Die Aktion 'google_generative_ai_conversation.generate_content' ist veraltet und wird in der Version 2026.4.0 entfernt. Verwende stattdessen die Aktion 'ai_task.generate_data'.",
+            "title": "Veraltete 'generate_content'-Aktion"
         }
     },
     "services": {
         "generate_content": {
-            "description": "Erzeugt Inhalte aus einem Prompt, der aus Text und optional aus Bildern besteht",
+            "description": "Erzeugt Inhalte aus einem Prompt, der aus Text und optional aus Bildern besteht (veraltet)",
             "fields": {
                 "filenames": {
                     "description": "Anh\u00e4nge, die zum Prompt hinzugef\u00fcgt werden sollen (Bilder, PDFs usw.)",
                     "example": "/config/www/image.jpg",
                     "name": "Dateinamen der Anh\u00e4nge"
                 },
-                "image_filename": {
-                    "description": "Veraltet. Verwende stattdessen Dateinamen.",
-                    "example": "/config/www/image.jpg",
-                    "name": "Dateiname eines Bildes"
-                },
                 "prompt": {
                     "description": "Die Textanweisung",
                     "example": "Beschreibe, was du auf diesen Bildern siehst",
                     "name": "Prompt"
                 }
             },
-            "name": "Inhalte erzeugen"
+            "name": "Inhalte erzeugen (veraltet)"
         }
     }
 }