When you’re learning to build with AI APIs, the last thing you want is for your tools to suddenly stop working. That nearly happened to Google’s engineering teams earlier this year when their proprietary SDK generator shut down unexpectedly. Today’s announcement means you’ll have more stable tools to work with.
What changed in plain terms
Google open sourced the tools that turn API specifications into ready-to-use code libraries (SDKs). These tools previously lived inside private companies, creating risk if those companies disappeared. Now they’re publicly available under the AGPLv3 license, meaning:
- You can generate SDKs for 7 languages (Python, TypeScript, Go, Java, C#, PHP, Ruby)
- The generated code belongs to you (MIT/Apache licensed)
- Improvements to the core generator stay open source
Tip
SDKs are pre-built code packages that save you from writing raw HTTP requests. They handle authentication, retries, and data formatting automatically.
Why this matters for beginners
When you’re starting out with AI APIs, SDKs do three critical things:
- Reduce boilerplate - No more copying curl commands or debugging JSON formatting
- Provide type safety - Your editor can autocomplete and catch errors early
- Handle edge cases - Streaming, pagination, and errors work consistently
The old system meant Google (and you) relied on closed tools that could vanish overnight. Now the core generator is community-maintained infrastructure.
What you can do today
If you’re working with Google’s Gemini APIs, update to the new SDKs:
from google import genai
client = genai.Client()
interaction = client.interactions.create(
model="gemini-3.8-flash",
input="Analyze this commit log",
)
Key improvements in the new versions:
| Feature | Before | Now |
|---|---|---|
| Streaming | Manual HTTP handling | Built-in SSE support |
| Errors | Generic exceptions | Detailed error hierarchy |
| Types | Basic validation | Full OpenAPI type checking |
The bigger picture: AI + determinism
Google’s approach here reveals an important lesson about where AI fits in development workflows:
- Non-deterministic tasks (like writing original code) use AI agents
- Deterministic tasks (like SDK generation) use traditional compilers
This hybrid approach prevents “shifting sands” under your projects while still leveraging AI where it excels.
Frequently asked questions
Do I need to rewrite my existing code?
No. The new SDKs maintain backward compatibility with existing Google GenAI API calls.
Can I use this for my own APIs?
Yes. The Speakeasy generator works with any OpenAPI 3.0 specification, not just Google’s.
Read next
- What OpenAI’s Australian Youth Safety Blueprint Means for Beginners
- How Cooley Uses ChatGPT to Streamline IPO Preparation for Lawyers
Want to try all of this hands-on? Start with the free Claude Code from Zero course.
Source
Based on Google Developers' announcement, “Why client SDK generation belongs in the open”. Written for people learning to build with these tools.