Jaycee Li

This commit introduces **new capabilities** to the **Gemini API data models**, extending key response and content structures. It adds `model_status` to `GenerateContentResponse` and `part_metadata` to `Part` objects, alongside supporting `rendered_parts` within `GroundingSupport`. These **API extensions** required significant updates to the **type definitions** in `google/genai/types.py` and corresponding **conversion logic** across multiple modules (`_live_converters.py`, `_tokens_converters.py`, `batches.py`, `caches.py`, `models.py`) to correctly handle these new fields for mldev and Vertex AI formats. Additionally, it enhances **model tuning** by adding `distillation_sampling_spec` and `tuning_job_metadata` to `TuningJob` objects, providing richer information for advanced features like **grounding** and **model lifecycle management**.

This commit performs a **refactoring** to **maintain backward-compatible field ordering** within the `google.genai` library's **generated data types**. It reorders the `history_config` field definition in `LiveClientSetup`, `LiveClientSetupDict`, `LiveConnectConfig`, and `LiveConnectConfigDict` classes within `google/genai/types.py`. Additionally, conditional blocks are reordered in conversion functions such as `_LiveClientSetup_to_mldev`, `_LiveClientSetup_to_vertex`, and `_LiveConnectConfig_to_mldev` across `_live_converters.py` and `_tokens_converters.py`. This ensures consistent serialization and deserialization behavior, preventing potential breaking changes for downstream consumers relying on the exact structure of these data types. The work is a **maintenance chore** to preserve API stability.

This commit performs a significant **internal refactoring** within the `google/genai` module to improve code structure and organization. It primarily involves **reordering conditional checks** within numerous internal conversion functions, such as `_FunctionDeclaration_to_vertex`, `_Tool_to_mldev`, and `_FunctionCallingConfig_to_mldev`, across files like `_live_converters.py`, `batches.py`, `caches.py`, and `models.py`. Additionally, `types.py` sees class members reordered and descriptions updated for various type definitions. This **maintenance** effort enhances the readability and maintainability of the codebase without introducing any functional changes or affecting external API behavior.

This commit introduces a **new feature** by adding the `enable_enhanced_civic_answers` field to the **`GenerateContentConfig`** configuration object within the `google.genai` library. This allows users to specify whether enhanced civic answers should be enabled during content generation requests. The change integrates this new configuration into the `_GenerateContentConfig_to_mldev` function for general content generation and batch processing. Crucially, it also adds **validation to explicitly disallow** the use of `enable_enhanced_civic_answers` when targeting **Vertex AI**, ensuring proper behavior and preventing unsupported configurations.

This commit performs **maintenance** by updating the **testing infrastructure** to utilize the `gemini-2.5-flash` model. Specifically, it modifies `google/genai/tests/models/test_generate_content.py` and `google/genai/tests/models/test_generate_content_tools.py` to use `gemini-2.5-flash` for content generation and RAG model tests, including updating generation configuration parameters. This change ensures that **model-related tests** are run against the latest available Gemini model. Additionally, it removes references to deprecated `gemini-1.x` models from the Local Tokenizer, improving the relevance and forward compatibility of the test suite and overall codebase.

This commit performs **maintenance updates** to the **batch processing tests** within the `google.genai` library. It primarily **updates model identifiers** across various batch creation, cancellation, and retrieval tests to utilize the newer **`gemini-2.5` models** and corresponding MLDev batch operation names. Additionally, a minor **bug fix** corrects the header key case from `content-type` to `Content-Type` in the `test_list.py` module. This ensures the **test suite remains aligned** with current API specifications and model versions, improving the reliability of batch operation validation.

This commit performs significant **internal refactoring and cleanup** across the `google/genai` library. It standardizes the handling of `google_maps` within the `_Tool_to_mldev` and `_Tool_to_vertex` conversion functions across modules like `_live_converters.py`, `_tokens_converters.py`, `batches.py`, `caches.py`, and `models.py`. Concurrently, the `google/genai/types.py` module receives a **feature enhancement** by introducing the `HttpElementLocation` enum and new configuration fields to `ApiKeyConfig`, alongside reordered and updated type definitions. This work improves consistency in internal tool conversion logic and expands the available configuration types for the `google.genai` client.

This commit performs a **refactoring** of the `google.genai` library's internal code. It **reorders conditional checks** within various data conversion functions, such as `_Blob_to_mldev` and `_Part_to_mldev`, across modules like `_live_converters.py`, `_tokens_converters.py`, `batches.py`, `caches.py`, and `models.py` for improved consistency. Additionally, it **reorganizes enum, class, and field definitions** and updates descriptions for core data types including `FileData`, `Blob`, `VideoMetadata`, `Part`, `Content`, and `SafetySetting` within `types.py`. This **maintenance** work enhances the internal code structure and readability, with no direct impact on external functionality.

This commit performs **internal refactoring** to clean up configurations within the **`google.genai` library**. It reorders the assignment of `url_context_metadata` within the `_Candidate_from_mldev` function in both `batches.py` and `models.py`. Additionally, it reorders definitions of `UrlRetrievalStatus`, `UrlMetadata`, `UrlContextMetadata`, and the `url_context_metadata` field within the `Candidate` class and `CandidateDict` in `types.py`, also updating their descriptions. This **maintenance** work primarily affects the **internal structure and readability** of the `genai` package's type definitions and candidate processing logic. The changes are purely organizational and do not alter any external functionality or API behavior.

This commit performs an internal **refactoring** to clean up and improve the clarity of **speech configuration** within the `google.genai` library. It reorders the `language_code` check within the `_SpeechConfig_to_vertex` function in both `_live_converters.py` and `models.py` for better logical flow. Additionally, the `types.py` module sees a **reorganization and update of descriptions** for several `SpeechConfig`-related type definitions, including `SpeechConfig`, `PrebuiltVoiceConfig`, and `VoiceConfig`. This **maintenance** work enhances code readability and maintainability without altering external API behavior.

This commit provides a **documentation update** to the **`google.genai` library**, specifically clarifying the retry behavior for HTTP requests. It updates the docstrings for the `attempts` field within the `HttpRetryOptions` and `HttpRetryOptionsDict` types in `google/genai/types.py`. This enhancement ensures that developers can easily understand the **default retry count** for HTTP operations, improving the usability and clarity of the API's retry configuration. The change is purely informational, with no impact on runtime behavior.

This commit **enhances the documentation** for **API request retry options** within the `google/genai/types.py` module. It specifically updates the descriptions for fields in `HttpRetryOptions` and `HttpRetryOptionsDict` to clearly specify their default values. This **feature-related documentation update** improves developer understanding and usability of the retry configuration, supporting the broader **new capability** of **retries in API requests**. The changes provide essential context for configuring retry behavior without introducing any functional code modifications.

This commit performs a **refactoring** of internal code within the **`google.genai` library** to enhance consistency and readability. It primarily involves **reordering conditional blocks** within various conversion functions, such as `_GoogleSearch_to_mldev` and `_Tool_to_mldev`, across modules like `_live_converters.py`, `_tokens_converters.py`, `batches.py`, `caches.py`, and `models.py`. Additionally, data type definitions in `types.py` have been updated and reordered for better clarity. This **maintenance** work improves the internal structure and maintainability of the library without altering external functionality.

This commit **enhances documentation** within the `google/genai/types.py` module by adding specific docstrings. These updates clarify that certain **enum classes and fields** are **not supported** when used with the **Gemini API** or **Vertex API**. This **documentation improvement** provides crucial guidance to developers, preventing potential errors or confusion when integrating with these AI services.

This commit **improves documentation** within the `google.genai.types.py` module by adding specific docstring notes. These notes clarify the support status of various **enums, fields, and classes**, indicating whether they are compatible with the **Gemini API** or **Vertex AI**. This **documentation enhancement** provides crucial information for developers, helping them understand API compatibility and avoid potential integration issues when working with different Google AI platforms. The **maintenance work** ensures that the type definitions accurately reflect their operational scope.

This commit performs **maintenance** on the project's testing infrastructure by updating a **replay test** for the `GoogleMaps.enableWidget` functionality. The update ensures the test accurately reflects current behavior and compatibility requirements, specifically in the context of the **Gemini API**. This work helps maintain the robustness and reliability of the `GoogleMaps` integration by ensuring its core widget enabling mechanism is correctly validated against evolving API dependencies.

This commit introduces a **new capability** by adding the `enable_enhanced_civic_answers` field to the **`GenerationConfig`** within the **`google.genai` library**, allowing users to specify this parameter for text generation. It includes crucial **validation logic** in `_GenerationConfig_to_vertex` to prevent the use of `enable_enhanced_civic_answers` with **Vertex AI**, raising an error if attempted. This ensures proper configuration handling and prevents unsupported feature combinations across different AI backends. Additionally, the commit refactors `GenerationConfigThinkingConfig` inheritance and updates various docstrings and enum orderings, enhancing the overall **configuration API** and its robustness.

This commit delivers a **bug fix** to prevent incorrect key re-casing within **Live API tool responses**. The core issue, located in the `convert_to_dict` utility function within `google/genai/_common.py`, involved the improper propagation of the `convert_keys` argument during recursive dictionary conversions. By correcting this propagation, the fix ensures that **nested dictionary keys in tool outputs maintain their original casing**, preventing unintended data transformations. Associated **live tests** were updated to include more comprehensive key structures and utilize robust assertion methods, verifying the correct handling of function responses and their keys. This change significantly improves the reliability of **data processing for tool outputs** within the `google.genai` library.

This commit introduces a **new capability** by adding comprehensive support for the `enableWidget` feature within the **Google Maps tool integration** of the `google.genai` library. It updates the internal conversion logic (`_GoogleMaps_to_mldev`) across various modules to correctly process and set this parameter. Crucially, the `google/genai/types.py` file is enhanced to include the `enable_widget` field in Google Maps types and introduces new types for **grounding metadata**, such as source flagging URIs and review details. This enables users to control the display of Google Maps widgets and provides richer contextual information for generated content. New test cases have been added to verify the functionality of this `enable_widget` parameter.

This commit **refactors** the **`APIError` constructor** within the **`google.genai.errors` module** by **removing no-op and redundant error conversion logic**. Specifically, the `__init__` method now directly handles specific response types, streamlining the internal error construction process. This **internal cleanup** improves code clarity and efficiency in the library's error handling subsystem without altering the public API or error behavior.