---
type: "post"
id: "mika-continues-rapid-release-cadence-with-new-android-8a9e47"
slug: "mika-continues-rapid-release-cadence-with-new-android-8a9e47"
title: "Mika continues rapid release cadence with new Android agent features"
description: "Mika’s v0.1.45 Android release adds runtime limits for scheduled tasks, more flexible schedule editing, and an editable WeChat follow-up draft after AI calls."
retrieval_nugget: "Mika’s v0.1.45 Android release adds runtime limits for scheduled tasks, more flexible schedule editing, and an editable WeChat follow-up draft after AI calls."
published_at: "2026-11-16"
updated_at: "2026-11-16"
record_date: "2026-11-16"
date_kind: "published_at"
topics: ["agents","automation"]
entities: ["Mika","wangzan101","Android","WeChat"]
related_patterns: ["harness-architecture-outlives-model-choice"]
novelty_label: "new_fact"
what_changed: "Mika v0.1.45 adds fixes and usability improvements, including correct quota-error handling and copying assistant replies, while reverting system-level voice wake because ordinary apps cannot access the required Android hotword interface."
prior_refs: ["https://newruntime.com/posts/mika-an-apache-2-0-on-device-android-27884e/","https://newruntime.com/posts/mika-v0-1-45-rolls-back-system-voice-9fce85/"]
source_url: "https://github.com/wangzan101/Mika/releases"
source_urls: ["https://github.com/wangzan101/Mika/releases"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:mika-continues-rapid-release-cadence-with-new-android-8a9e47"
status: "published"
visuals: []
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/posts/mika-continues-rapid-release-cadence-with-new-android-8a9e47/","markdown":"https://newruntime.com/posts/mika-continues-rapid-release-cadence-with-new-android-8a9e47.md","json":"https://newruntime.com/posts/mika-continues-rapid-release-cadence-with-new-android-8a9e47.json"}
markdown_url: "https://newruntime.com/posts/mika-continues-rapid-release-cadence-with-new-android-8a9e47.md"
---

# Mika continues rapid release cadence with new Android agent features

## Retrieval answer

Mika’s v0.1.45 Android release adds runtime limits for scheduled tasks, more flexible schedule editing, and an editable WeChat follow-up draft after AI calls.

## Summary

Mika’s v0.1.45 Android release adds runtime limits for scheduled tasks, more flexible schedule editing, and an editable WeChat follow-up draft after AI calls. It also rolls back system voice wake-up because an ordinary app cannot access Android’s system hotword facility without root.

For teams building mobile agents, the update is a useful reminder that task limits, failure visibility, and explicit human review can be more deployable than an interaction pattern that the operating system does not permit.

## What happened

Mika’s GitHub release notes identify v0.1.45 as the latest release, published on November 16. The update changes the Android agent’s scheduled-task and voice workflows, adds an AI-call follow-up feature, and contains a collection of reliability and interface fixes.

For scheduled tasks, users can now set a maximum runtime. The release notes also say that notifications after consecutive failures include the failure count, giving users more context when a recurring task repeatedly fails. The schedule editor adds a time picker and supports editing multiple times per day or week. Together, these features give an operator three controls over recurring work: when it should start, how many scheduled times it has, and how long an individual execution may run.

Those controls matter because a recurring mobile task is not simply a one-off prompt. A task can run when a person is not watching it, encounter an unavailable app state, or repeat a failure across several scheduled attempts. A maximum runtime creates a stated boundary around a run. Showing the number of consecutive failures makes an alert more informative than a generic failure message. The release notes do not frame these additions as autonomous recovery; they are controls and notifications for the person responsible for the schedule.

The release also adds a post-call workflow: after an AI call, Mika can generate an editable WeChat follow-up draft. Mika’s release notes describe a draft rather than automatic sending. That distinction keeps the final wording and decision to send with the user, while reducing the work of turning a call into a follow-up action. The feature sits alongside the scheduling changes as another attempt to carry agent output into a bounded, reviewable operational step.

The most consequential rollback concerns the “Hello Mika” voice wake-up feature. Mika’s developers say that a regular app without root access cannot obtain Android’s system hotword entry point. Keeping a microphone recording continuously at the application layer would not reproduce system-level wake-word behavior, and the developers cite power consumption and background restrictions as problems that cannot be solved this way. The rollback does not remove all voice interaction: push-to-talk and the existing capsule long-press-to-talk behavior remain.

Mika’s release notes also correct the classification of a `403 + insufficient_quota` response so it is shown as exhausted quota rather than an invalid key. They add text selection and one-tap copying in the personal assistant, validation for API keys containing invalid characters, and improved guidance where the speech-recognition service lacks microphone permission. A scheduled-task fallback fix prevents an exhausted daily budget from incorrectly jumping to another model from the same provider. Session exports retain the message snapshot selected at export time.

GitHub lists two APK assets for v0.1.45: an arm64 package of 111 MB and a universal package of 123 MB, each with a SHA-256 checksum. The release page also includes source-code archives for the tag. The changes are a versioned release rather than a stated pricing or access change.

## What the archive adds

New Runtime covered Mika’s earlier additions of scheduling controls in [v0.1.45 coverage](https://newruntime.com/posts/mika-v0-1-45-rolls-back-system-voice-9fce85/) on November 16, and previously described its on-device Android agent in [an earlier note](https://newruntime.com/posts/mika-an-apache-2-0-on-device-android-27884e/) on November 2. This is a new update to a known story: it makes schedules more configurable and adds runtime limits, while the wake-word rollback clarifies a platform boundary.

## Trend signal

This update supports the pattern that agent workflows depend on the surrounding harness and operating environment, not just model capability. The scheduled-task controls change the execution layer: they constrain runs and make repeated failures legible to an operator. The wake-word rollback shows how Android permissions, microphone behavior, and background limits can shape the product boundary. See [Harness architecture outlives model choice](https://newruntime.com/patterns/harness-architecture-outlives-model-choice/).

## Why now

The release notes connect the rollback directly to the limits of ordinary Android apps: without root access, Mika cannot use the system hotword entry point, and continuous app-level recording brings power and background constraints. The update retains voice controls that work within those boundaries rather than presenting an always-listening feature as equivalent to system-level wake-word support.

## Why it matters

For teams building mobile agents, this release is a reminder that reliability features often matter as much as adding new model capabilities. A maximum runtime provides a clear boundary for scheduled work; failure-count notifications make repeated problems more visible. Multiple daily or weekly times reduce friction for routine automation, while an editable follow-up draft can turn a call outcome into a next action without surrendering final message approval.

The rollback is equally instructive. An app-level implementation cannot necessarily recreate a privileged operating-system affordance, even when the feature appears to be a simple voice trigger. Product teams should distinguish between a capability the platform exposes and one that would require persistent background behavior the platform restricts. Mika preserves push-to-talk rather than implying that the hotword feature is still available.

For an operator, the practical design lesson is to expose the failure boundary. A scheduled agent should reveal that it has exceeded its allowed run time or failed repeatedly, rather than only promising that it will execute. Likewise, a communications workflow can prepare a message but should retain an edit point when the content represents the user to another person.

## Limits and uncertainty

Mika’s release notes describe intended behavior, but this note has not independently tested the new schedule controls, the generated WeChat drafts, or the rollback on Android devices. The notes do not quantify schedule completion rates, battery effects, or the quality of generated drafts. They also do not establish, through independent measurement, how consistently the maximum runtime prevents problematic long-running tasks across devices and Android versions.

The published APK sizes and SHA-256 values are release metadata, not independent verification of the builds. The explanation of the wake-word rollback is Mika’s developers’ account of Android access, power, and background constraints. Its applicability may vary with device manufacturers, Android releases, permissions, and rooted versus ordinary application environments.

## Skill to build now

For one recurring mobile-agent workflow this week, define an explicit maximum runtime before enabling the schedule. Decide what an operator should see after one failed run and after repeated failures, then make the alert include the count and enough task context to investigate. For any follow-up message created from a call, keep it as a reviewable draft and require a person to approve or edit it before sending.

## Reader takeaway

Mika v0.1.45 improves recurring-task controls and adds a reviewable follow-up draft, but removes system voice wake-up in response to Android’s access and background constraints. For mobile-agent builders, the durable lesson is to pair bounded execution and human review with interaction modes the platform can actually support.

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- topic hub: [Agents](https://newruntime.com/topics/agents.md): Explore the agents topic hub.
- topic hub: [Automation](https://newruntime.com/topics/automation.md): Explore the automation topic hub.
- collection: [Field Notes](https://newruntime.com/posts.json): Open the latest editorial analysis.
- collection: [Sources](https://newruntime.com/sources/): See the original sources this archive cites.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)
