v2.6.0

vectorize-io/hindsightv2.6.0Oct 20, 2025by cguiguet

Upgrading?

Release notes summary

Version 2.6.0 is a substantial feature release for the mobile-use project, introducing task cancellation, exclusive execution, ADB direct commands, and a screen analyzer agent that decouples Cortex reasoning from VLM requirements. It also includes numerous fixes for iOS detection, long press behavior, and agent initialization.

Key Highlights

  • Added task cancellation and exclusive execution support
  • Added ADB direct commands for core actions
  • Introduced screen analyzer agent so Cortex reasoning agent no longer needs to be a VLM
  • Added fallback mechanisms for all agents
  • Fixed iOS device detection by checking for Darwin instead of Linux
  • Reworked planner prompt and replanning logic
  • Fixed long press with variable time on Android

New Features

  • Task cancellation and exclusive execution
  • ADB direct commands for core actions
  • Screen analyzer agent
  • Fallback mechanisms for all agents
  • Streaming status endpoint
  • Minitap default LLM profile preference over OpenAI config
  • Renamed input and clear text tools for clarity
  • Revamped wait-for-animation-to-end tool

Full Release Notes

## What's Changed
* feat: bump to v2.5.0 by @cguiguet in https://github.com/minitap-ai/mobile-use/pull/79
* Fix iOS device detection by checking for Darwin instead of Linux by @yulingtianxia in https://github.com/minitap-ai/mobile-use/pull/80
* feat: prefer minitap default llm profile over openai config by @plfavreau in https://github.com/minitap-ai/mobile-use/pull/81
* feat: get streaming status endpoint by @plfavreau in https://github.com/minitap-ai/mobile-use/pull/82
* fix(sdk): adb host forwarded to maestro server by @KYN4 in https://github.com/minitap-ai/mobile-use/pull/83
* fix: correct typo in Cortex documentation from 'Exemple' to 'Example' by @yulingtianxia in https://github.com/minitap-ai/mobile-use/pull/84
* feat: adding task cancellation and exclusive execution by @cguiguet in https://github.com/minitap-ai/mobile-use/pull/88
* feat: adding adb direct commands for core actions by @cguiguet in https://github.com/minitap-ai/mobile-use/pull/87
* fix: ignore known benign warnings in device hardware bridge logs by @plfavreau in https://github.com/minitap-ai/mobile-use/pull/86
* feat: screen analyzer agent - cortex reasoning agent does not need to be a VLM anymore by @plfavreau in https://github.com/minitap-ai/mobile-use/pull/85
* fix: Agent initialization fails when adb is not installed, but xcrun is by @plfavreau in https://github.com/minitap-ai/mobile-use/pull/90
* fix: planner prompt rework for planning - replanning logic by @cguiguet in https://github.com/minitap-ai/mobile-use/pull/89
* fix: long press is now working correctly with variable time for android by @cguiguet in https://github.com/minitap-ai/mobile-use/pull/94
* feat: add fallback mechanisms for all agents by @plfavreau in https://github.com/minitap-ai/mobile-use/pull/95
* docs(env): add default values and clarifying comments to environment variables by @plfavreau in https://github.com/minitap-ai/mobile-use/pull/91
* Feat: Renaming input and clear text tools to be more explicit by @cguiguet in https://github.com/minitap-ai/mobile-use/pull/93
* feat: Revamp wait for animation to end tool + tiny fixes by @plfavreau in https://github.com/minitap-ai/mobile-use/pull/92

## New Contributors
* @yulingtianxia made their first contribution in https://github.com/minitap-ai/mobile-use/pull/80

**Full Changelog**: https://github.com/minitap-ai/mobile-use/compare/v2.5.0...v2.6.0

Related Repositories