Gemini API Interactions breaking changes (May 2026): outputs → steps, new response_format, and migration timeline
Google’s Gemini API Interactions endpoint is changing its response schema (outputs → steps), consolidating output format controls into response_format, and updating streaming event names, with a May 26 default flip and a June 8 legacy sunset.
Editorial Team
The AI Coding Tools Directory editorial team researches and reviews AI-powered development tools to help developers find the best solutions for their workflows.
Google published a migration guide for upcoming breaking changes to the Gemini API Interactions endpoint, including a response schema change from outputs to steps, plus a firm timeline for rollout. (https://ai.google.dev/gemini-api/docs/interactions-breaking-changes-may-2026)
1) Response schema: outputs → steps
Google says responses move from a flat outputs array to a structured steps array:
POST /interactionsreturns only output steps.GET /interactions/{id}returns the full step timeline (including the initialuser_inputstep).
(https://ai.google.dev/gemini-api/docs/interactions-breaking-changes-may-2026)
2) Output format changes: response_mime_type removed
Google says response_mime_type is removed and output control moves into a polymorphic response_format field.
The guide also notes that image output configuration previously under generation_config.image_config moves into a response_format entry with type: "image". (https://ai.google.dev/gemini-api/docs/interactions-breaking-changes-may-2026)
3) Streaming event names change
Google’s guide lists new server-sent-event types (e.g. interaction.created, interaction.completed, step.start, step.delta, step.stop) replacing older event names. (https://ai.google.dev/gemini-api/docs/interactions-breaking-changes-may-2026)
For streaming function calls, Google says function arguments arrive as partial JSON string deltas (arguments_delta), and clients must accumulate deltas to reconstruct full arguments. (https://ai.google.dev/gemini-api/docs/interactions-breaking-changes-may-2026)
Migration timeline (key dates)
The guide lists a staged rollout:
- May 7, 2026: opt-in starts (REST API opt-in via
Api-Revision: 2026-05-20; SDK users upgrade to Python ≥2.0.0 / JS ≥2.0.0). (https://ai.google.dev/gemini-api/docs/interactions-breaking-changes-may-2026) - May 26, 2026: new schema becomes the default; REST API users can temporarily opt out via
Api-Revision: 2026-05-07. (https://ai.google.dev/gemini-api/docs/interactions-breaking-changes-may-2026) - June 8, 2026: legacy schema is removed for Interactions;
Api-Revisionis ignored. (https://ai.google.dev/gemini-api/docs/interactions-breaking-changes-may-2026)
Sources
- Google AI for Developers migration guide: https://ai.google.dev/gemini-api/docs/interactions-breaking-changes-may-2026
Free Resource
2026 AI Coding Tools Comparison Chart
Side-by-side comparison of features, pricing, and capabilities for every major AI coding tool.
No spam, unsubscribe anytime.
Frequently Asked Questions
What is Gemini API Interactions breaking changes (May 2026): outputs → steps, new response_format, and migration timeline?
Related Articles
Claude Fable 5 and Mythos 5: Anthropic launches a new frontier model with safeguards and trusted access
Anthropic announced Claude Fable 5 for broad availability and Claude Mythos 5 for a restricted trusted-access program, with safety fallbacks to Opus 4.8 and new rollout/pricing details.
Read more →Product UpdatesWarp 2026.06.03: queued prompts for Agent Mode, configurable line numbers, and PR metadata for better context
Warp’s June 3, 2026 release adds queued prompts for Agent Mode, configurable absolute/relative line numbers in code editors, and richer agent request context via PR/repository metadata.
Read more →Product UpdatesWarp 2026.05.27: Agent Mode adds inference endpoint aliases, Bedrock model indicators, and image viewing
Warp’s May 27, 2026 release improves Agent Mode with token-usage alias names for custom inference endpoints, AWS Bedrock indicators in the model picker, and the ability to view attached images in agent conversations.
Read more →