Google Gen AI Python SDK releases
Gemini 3.8 Flash model added to SDKs in version 2.22.0
The 2.22.0 release includes the Gemini 3.8 Flash model in SDKs.
The complete AI news desk
The AI world, in one verified news room.
Models, agents, robotics, research, business and policy, with the original source attached. Search and move beyond the latest thirty updates.
Open the 30-day AI Change IndexLatest and archive
Open the fast flashes streamRobotics
Google Gen AI JavaScript SDK releases
SDKs gain Gemini 3.8 Flash model support
The 2.21.0 release includes the Gemini 3.8 Flash model in SDKs.
Google Gen AI Java SDK official releases
Gemini 3.8 Flash model added to SDKs
The 2026-09-02 release includes the Gemini 3.8 Flash model in SDKs.
Prefect official releases
Prefect updates deployment schedule dialog with parameter overrides.
The latest nightly release includes a fix for the create-flow form item gap and updates to the Snowflake connector.
LiteLLM official releases
LiteLLM 1.99.1 Docker images now include OpenTelemetry v2 cache token counts.
LiteLLM 1.99.1 adds OpenTelemetry v2 span cache token counts. Docker images are signed with cosign.
Cline official releases
Cline v3.0.61 now prompts users before replacing an older running Hub.
Cline v3.0.61 now prompts users before replacing an older running Hub, showing active sessions and allowing for a graceful shutdown. Windows binaries are now Authenticode-signed.
OpenAI Java SDK releases
OpenAI Java 4.56.0 updates usage APIs, documentation and OkHttp website links.
OpenAI Java 4.56.0 updates usage APIs and documentation, and refreshes links to the OkHttp website.
OpenAI Python library releases
New API Usage Documentation Released in Version 3.7.0
The 3.7.0 release includes updated API documentation and fixes repeated numpy checks.
OpenAI Go SDK releases
Version 3.55.0 adds compute_units to API responses.
The 3.55.0 release includes new API features for tracking compute units in responses and chat completions. The update also improves usage APIs and documentation.
OpenAI Node SDK official releases
Version 7.9.0 of the software raises custom-code budget to 4,000 lines.
The latest release includes support for Jest 28 CommonJS package imports. The custom-code budget has been raised to 4,000 lines.
Google Gemini CLI official releases
Version 0.58.0 of the gemini-cli-robot software is released with Docker isolation on macOS.
Version 0.58.0 improves Docker isolation on macOS and fixes symlink evaluation in ignore paths.
Google Gen AI Java SDK official releases
The 1.69.0 release of the SDK adds translation_config support for GEAP.
The 1.69.0 release includes a new README.md with example usage instructions.
Agno official releases
v3.0.5: Ingestion now reports embedding failures instead of success.
v3.0.5: Content ingestion now reports failures, adding a partial status. The ContentStatus enum is updated to include partial.
LangChain.js official releases
LangChain AWS Adds Claude Model Variants to Supported Tool Choices
The AWS LangChain library v1.4.5 now supports Claude-Sonnet-5, Claude-Opus-5, and Claude-Haiku-5.
Pydantic AI official releases
New version 2.37.0 adds glm-5.3-flash and updates Z.AI test suite.
The update includes a new glm-5.3-flash feature and fixes for span queries.
Google Gen AI JavaScript SDK releases
The Interactions API now supports video understanding.
The 2.20.0 release includes audio/webm support and GEAP translation_config SDK.
Google Gen AI Python SDK releases
Version 2.21.0 adds support for audio/webm MIME type
The 2.21.0 release adds support for streaming downloads and the audio/webm MIME type.
OpenAI Java SDK releases
Version 4.55.0 adds compute units to API responses.
v4.55.0 adds compute units to API responses and supports X.509 token exchange.
NVIDIA Triton Inference Server official releases
NVIDIA Triton Inference Server 2.72.0 fixes dynamic-batcher starvation.
Triton Inference Server 2.72.0 fixes dynamic-batcher starvation and improves model-readiness reporting.
Docling official releases
New LaTeX output format added to CLI in v2.124.0
The new version 2.124.0 includes a LaTeX output format for the CLI and handles Retry-After HTTP-date with unknown timezone.
Triton language official releases
AMD updates gfx950 tutorial to LLVM pin b010a18d, drops warp-pipeline barrier.
The tutorial now supports out-of-tree LLIR-scheduler and amdgcnas plugins. The warp-pipeline barrier commit from v2.0 was intentionally removed.
Agno official releases
KnowledgeManagementTools flags renamed, ingest_path now opt-in
The KnowledgeManagementTools constructor flags have been renamed to specify the tools they register. The ingest_path flag is now opt-in and defaults to off.
Agno official releases
Website ingestion now loads one content row per page.
v3.0.3: Website ingestion now processes each page individually, allowing for selective refreshes and deletions.
Agno official releases
New MCP configuration rejects unknown fields and raises errors at boot.
The Synthorai model provider integrates with the Synthorai gateway via an OpenAI-compatible endpoint. The WaveSpeed integration supports image and video generation through its API.