ADK-Rust v0.5.0 — The Journey to a Major Release
From v0.3.1 to v0.5.0: what changed, what we learned, and where we're headed.
v0.5.0 is the biggest ADK-Rust release yet. It touches every layer of the framework — from new crates and model providers to a completely refreshed developer experience. Here's the story.
New Crates
v0.5.0 introduces several new crates that expand what agents can do:
- •
adk-rag— Retrieval-Augmented Generation with pluggable chunkers, embedding providers (Gemini, OpenAI), and vector stores (in-memory, Qdrant, pgvector, LanceDB) - •
adk-code— Sandboxed code execution for agents that need to run code - •
adk-deploy— Shared manifest and bundling contract for the upcoming ADK Platform - •
adk-rust-macros— Derive macros for cleaner tool and agent definitions
Model Provider Updates
Two new dedicated provider docs landed: Anthropic (with extended thinking support) and OpenAI Responses API (the successor to Chat Completions). The provider trait now supports context caching lifecycle management — set a CacheCapable model and the runner handles cache creation, refresh, and TTL automatically.
Breaking Changes
The upgrade from 0.3.1 required adapting to a few API changes:
- •
RunnerConfiggained four new fields for context caching, cancellation tokens, and request context - •
Runner::run()now takesUserIdandSessionIdnewtypes instead of raw strings - •
AdkError::ToolbecameAdkError::tool()(lowercase constructor)
cargo-adk CLI
The new cargo-adk CLI replaces manual project setup. Five templates cover the most common starting points: basic Gemini agent, custom tools, RAG with vector search, REST server, and OpenAI-powered agent.
cargo install cargo-adk cargo adk new my-agent --template rag cd my-agent && cargo run
Website Refresh
The website got a visual overhaul: deeper navy palette, refined code blocks with glow effects, a v0.5.0 release badge in the hero, and the Get Started section now uses the cargo-adk CLI flow. All 47 docs were synced fresh from the upstream repo, and 5 new doc pages were added (RAG, Code Execution, Payments, Anthropic, OpenAI Responses).
What's Next
The roadmap for the rest of 2026 includes ADK Platform (managed deployment), Super Agents (Job Hunter, Customer Voices), and deeper integrations with the VS Code extension. The framework continues to push toward making Rust the best language for building production AI agents.
Ready to try v0.5.0?
Get started with the most powerful Rust framework for building AI agents.