ReleaseBytes
gemini Genkit Releases · · 0.8.0

Genkit Python SDK v0.8.0: Expanded Google AI support, multi-region routing

sdkaideprecationmedia
feature breaking patch

Genkit Python SDK v0.8.0 introduces expanded model support and multi-region routing for Google GenAI and Vertex AI integrations, along with a modernized plugin packaging architecture. This release impacts Python developers integrating with Google's AI services, offering new capabilities like multimodal embeddings and generative video models. A staged rollout maintains compatibility for most plugins while core and Google GenAI packages are updated, with a clear migration path for the renamed Google GenAI plugin.

  • →Expanded Google AI & Vertex AI capabilities
  • →Google GenAI plugin renamed and namespace modernized
  • →Refined middleware lifecycle hooks
  • →Staged rollout for Genkit core and Google GenAI plugins
  • →Gemini tool role compatibility and streaming fixes
Breaking changes (2) ›
  • Google GenAI plugin renamed and namespace modernized

    The `genkit-plugin-google-genai` package has been renamed to `genkit-google-genai`, and its module import path changed from `genkit.plugins.google_genai` to `genkit_google_genai`. A tombstone package ensures zero-downtime transition with a `DeprecationWarning` for existing applications.

  • Refined middleware lifecycle hooks

    The `GenerateHookParams` for `wrap_generate` no longer include the transport `request`, providing direct access to `options`, `iteration`, and `message_index`. Custom middleware inspecting `params.request` should migrate to `params.options` or use `wrap_model` or `wrap_tool` for raw HTTP payload access.

Features (1) ›
  • Expanded Google AI & Vertex AI capabilities

    The `genkit-google-genai` package now supports Vertex AI multi-region routing, multimodal embeddings via `:predict`, Gemini embedding-2, various Gemini 3.x text and image generation models, and Veo 3.x generative video models.

Fixes (2) ›
  • Gemini tool role compatibility and streaming fixes

    Resolved `400 Bad Request` API rejections for Gemini 3.6 by mapping tool execution history turns to the correct role, and fixed accurate finish reason and token usage reporting for `generate_stream` executions.

  • Model discovery, embedder registration, and option type updates

    Fixed runtime discovery for Vertex AI models referenced by name, ensured only callable Vertex embedders are registered, and updated the `output_instructions` parameter in `ai.generate()` to accept boolean flags.

Notes (1) ›
  • Staged rollout for Genkit core and Google GenAI plugins

    Genkit Python SDK v0.8.0 is released in stages, with foundational packages `genkit`, `genkit-google-genai`, and a tombstone for `genkit-plugin-google-genai` updated. This approach allows critical Gemini updates to be released promptly while awaiting internal naming review for other plugins, ensuring backward compatibility for remaining plugins at v0.7.0.

Read the original announcement →

https://github.com/genkit-ai/genkit/releases/tag/py/v0.8.0

Related releases