◉ANTIGRAVITY LABJP
Articles/Agents & Manager
◈ Agents & Manager/2026-06-16Advanced

Generating Multilingual Release Notes with the Managed Antigravity Agent via the Gemini API

A hands-on record of building a pipeline that turns git commit logs into multilingual App Store and Google Play release notes using the Managed Antigravity Agent, now in public preview through the Gemini API.

antigravity476agents152gemini-api8automation96app-dev54

✦ Premium Article

When you ship four apps in parallel to the App Store and Google Play, the quietly time-consuming part of every release is localizing the release notes. When I rolled out v2.1.0 in a staged release recently, rewriting the English draft into each language, fitting each store's character limit, and keeping the tone consistent ate up the better part of an hour. The work around the code often feels longer than the code itself.

On June 15, antigravity-preview-05-2026 (the Managed Antigravity Agent) entered public preview through the Gemini API, so I built a small pipeline to hand this tedious step to an agent. Here is what differs from an ordinary generate_content call, and where I got stuck.

Why a Managed Agent instead of a one-shot generation call

At first I assumed I could just pass the commit log to generate_content and have it write the notes in each language. In practice the prose was fine, but Google Play's "What's new" field caps at roughly 500 characters, and whenever the output ran over, the trimming came right back to me. I ended up rewriting the prompt and regenerating every time something exceeded the limit.

What makes a Managed Agent fundamentally different from one-shot generation is that it can plan, reason, run code, touch files, and browse the web autonomously inside a sandbox. My task has several stages — classify commits, translate to each language, count characters and fit the limit — and the part that really paid off was letting the agent count the characters itself and trim when it went over. The job of watching the limit and bouncing things back moves cleanly onto the agent.

The way I think about dividing labor between "one-shot generation" and "an agent" is a continuation of what I laid out in the cloud-vs-local boundary of the Managed Agents API. Treat this article as its implementation companion.

The overall flow

The pipeline I built has four stages. Only the actual submission to the stores happens after I review everything by hand. I keep the agent's autonomy to text generation and press the publish button myself.

  1. Extract a structured git log on the caller side
  2. Hand the agent the commit summary plus constraints, and let it plan, classify, generate per-language, and validate length
  3. Receive JSON back from the agent
  4. Re-validate that JSON on the caller side and bounce anything that breaks a limit
✦

Thank you for reading this far.

Continue Reading

What follows includes implementation code, benchmarks, and practical content we hope you'll find useful. This site runs without ads — server and development costs are supported entirely by members like you. If it's been helpful, we'd be truly grateful for your support.

WHAT YOU'LL LEARN
✦Solve the 'translate while staying under the character limit' step that a single generate_content call kept choking on, using the agent's own self-checking loop
✦You'll get concrete code that calls the Managed Agent from the google-genai SDK and uses Function Calling to validate length, ready to drop into your own release flow
✦Sidestep the three traps — sandbox file access, over-translation, and cost — and cut a manual 50-minute chore down to about 6 minutes
Secure payment via Stripe · Cancel anytime
✦

Unlock This Article

Get full access to the rest of this article. Buy once, read anytime. This site is ad-free — your support goes directly toward keeping it running.

or
Unlock all articles with Membership →
Share

Thank You for Reading

Antigravity Lab is ad-free, supported entirely by members like you. We publish practical guides daily with implementation code, benchmarks, and production-ready patterns. If you've found it useful, we'd love to have you on board.

  • ✦Copy-paste ready implementation code
  • ✦New advanced guides published daily
  • ✦$5/mo or $10 for lifetime access
View Membership →

Related Articles

◈ Agents & Manager2026-04-20
Building a Business Automation Agent with Antigravity × Google Workspace API — Gmail, Drive, Sheets and Docs in Production
Learn how to build AI agents that automate Gmail responses, generate documents from Drive templates, and create intelligent Sheets reports using Antigravity, Google Workspace APIs, and Gemini.
◈ Agents & Manager2026-10-13
Turning Gut-Feeling Judgments into Rules an Agent Can Actually Follow
A repeatable way to turn the judgment calls that live only in your head into rules an Antigravity agent can apply: build conditions from past cases, choose where in .agents/rules they live and which trigger to use, catch rules that are silently ignored, and replay old cases before trusting them.
◈ Agents & Manager2026-09-04
Snapshot Your Agents' Effective Scope So You Notice the Day a Default Changes
Antigravity 1.1.25 made Markdown-defined custom agents inherit the surrounding skills, rules, and subagents by default. Here is the snapshot-and-diff routine I now use to check whether an agent I meant to keep narrow quietly got wider after an update.