Comeback — on-device smart replies for Kotlin Multiplatform

Why Comeback

Smart reply without the baggage

No server, no native libraries, no Google Play services. It is a small Kotlin library that you can read, test and ship.

Safe by design

Replies come from a fixed, reviewed list of about 1,300 short replies and are never generated, so it can't say something rude or odd. Unhelpful chips like "What?" or "Huh" are never shown.

Chips that differ

At most one reply per intent group, so you won't get "Thanks", "Thank you" and "Thanks!" side by side.

No dependencies

Pure Kotlin inference. There is no ONNX Runtime, TFLite, ML Kit or native code, and it runs on Android (minSdk 21), iOS, macOS, JVM desktop, JavaScript and WebAssembly, from Kotlin or Java.

Private & offline

The model ships inside the library on every platform. It makes no network calls, needs no permissions and sends no telemetry, so messages never leave the phone.

~0.5ms

Per message

Android emulator, once warm

~5MB

On-device

model + tokenizer + replies

1,303

Replies

reviewed, in intent groups

Dependencies

no native code

On device

Tap a chip, send a reply

Real output from the sample app on an Android emulator. These are not mockups.

Comeback replies to Are you coming tonight?

Plans

"Are you coming tonight?" → I'm not sure yet · I haven't decided yet · Yes

Thanks

"Thanks for your help" → No problem, have a great day! · You're welcome · Anytime

Sent

"Running 10 minutes late" → tapped "OK"

Same library on every platform

The Kotlin Multiplatform sample app, built from the published artifacts, gives the same result on Android, iOS, desktop and the web.

Android

Are you coming tonight? → I'm not sure yet 70%

iOS

Are you coming tonight? → I'm not sure yet 70%

Desktop (JVM)

Are you coming tonight? → I'm not sure yet 70%

Web (Wasm)

Are you coming tonight? → I'm not sure yet 70%

Install

Add it to your build

One line in your Gradle file, from Maven Central. Works in commonMain for Android, iOS, macOS, JVM desktop, JavaScript and WebAssembly. No API key, no account.

Maven Central live

// build.gradle.kts: commonMain, or any platform source set
dependencies {
    implementation("io.github.rajumark:comeback:2.0.0")
}

// Gradle picks the right artifact per platform:
// comeback-android · comeback-jvm · comeback-iosarm64 · comeback-iossimulatorarm64
// comeback-macosarm64 · comeback-js · comeback-wasm-js

Build with a prompt

Add Comeback from rajumark to this Kotlin Multiplatform or Android project.

What it does: on-device smart replies. Given a received message, it suggests 3 short replies to tap, picked from a fixed reviewed list. English, ~5 MB, no dependencies, no network.

Install (Maven Central, no extra repository):
  build.gradle.kts (commonMain or any source set): implementation("io.github.rajumark:comeback:2.0.0")
  Platforms: Android, iOS, macOS, JVM desktop, JavaScript, WebAssembly

Usage:
  val comeback = Comeback()      // off the main thread, keep one instance
  val chips = comeback.replies("Are you coming tonight?").map { it.text }

Repo: https://github.com/rajumark/comeback#readme
Add the dependency, then follow the README for the exact API and current version.

Usage

Three lines to reply chips

Load once and reuse the instance. replies() is thread-safe.

Kotlin

val comeback = Comeback()

val chips = comeback.replies("Let's meet at 7")
chips.map { it.text }  // [OK, OK, see you then, See you there!]

comeback.close()

Java

try (Comeback comeback = new Comeback()) {
    List<SmartReply> r =
        comeback.replies("Are you coming tonight?");
}

Use cases

Where Comeback fits

Anywhere a user answers a message and a single tap would do.

Chat & messaging apps

Show reply chips above the composer for the last received message.

"I miss you" → I miss you too

Notification quick replies

Offer reply actions straight from a notification, with no server round trip.

"Can you call me?" → Sure

Wear OS & small screens

Typing on a watch is painful; one tap isn't.

"See you tomorrow" → See you

Community & marketplace inboxes

Help buyers and sellers answer quickly on logistics.

"Let's meet at 7" → OK, see you then

Accessibility

Let people with motor or typing difficulties answer with one tap.

"Good morning!" → Hello!

Offline-first apps

Works in airplane mode, on patchy networks and in regions where cloud AI isn't an option.

"Running 10 minutes late" → OK

App ideas

20 apps you could build with Comeback

Real problems Comeback solves today. Pick one and ship it.

Chat app reply chips · Show three one-tap replies above the keyboard for the last received message.

Notification quick replies · Reply straight from the notification shade, with no server round trip.

Wear OS messaging · Typing on a watch is painful; tapping "On my way" isn't.

Marketplace seller inbox · Answer "Is this still available?" with "Yes, it is" in one tap.

Driver & delivery apps · Let drivers send "Reached" or "5 minutes away" without typing on the road.

Rental & property chat · Owners reply to "Can I visit today?" in one tap instead of typing each time.

Team chat · Quick "Sounds good", "On it" and "Thanks!" acknowledgements in work messengers.

Senior-friendly messenger · Big one-tap reply buttons for parents and grandparents who find typing slow.

Accessibility · People with motor or typing difficulties answer any message with a single tap.

Email client · Short suggested replies for quick "Yes, works for me" type emails.

In-game chat · Reply to teammates without leaving the match or opening a keyboard.

Appointment reminders · Patients reply "Confirmed" or "Can we reschedule?" in one tap.

Delivery customer chat · Customers answer the delivery person with "Leave it at the door" instantly.

Family safety app · Kids reply "Reached school" or "Coming home" to parents with one tap.

Classroom messaging · Students and parents quickly acknowledge teacher announcements.

Sales CRM · Reps respond to lead messages fast so no enquiry waits for hours.

Housing society app · Residents reply "Noted, thanks" to society notices without typing.

Chatbot suggestions · Show likely user responses as buttons in a support bot conversation.

Offline-first messaging · Suggestions keep working in airplane mode and on patchy rural networks.

Dating & social apps · Help users keep conversations moving with quick, friendly, reviewed replies.

Quality

Honest numbers

Measured on 335 hand-written messages across 16 categories. A reply counts as good when a person would plausibly send it.

	Comeback	Google ML Kit Smart Reply
A good reply in the top 3	63.6%	72.8%
Top reply is good	42.1%	52.8%
Answers every message	Yes	95%
Size	4.6 MB	6.6 MB + 4.4 MB native libs
Latency (Android emulator)	~0.5 ms	~16.6 ms
Dependencies	None	ML Kit + native code

ML Kit gives better replies. Comeback is smaller and much faster, needs no native code or Google Play services, and you can read and change its reply list. It is strongest on thanks (100% good in the top 3), yes/no questions (90%), affection and goodbyes. It is weakest on opinions and open questions, where you can hide the chips with minConfidence.

Reference

API at a glance

Member	Description
`Comeback()`	Loads the bundled model. Implements `AutoCloseable`.
`replies(message, limit = 3, minConfidence = 0f)`	The best replies first, one per intent group. Empty for a blank message, or when the best reply scores below `minConfidence`.
`SmartReply(text, confidence)`	One reply.

Pricing & license

Free until you're big.

No API key, no sign-up, no license file. Add Comeback to your project and ship. It stays free until your product passes 10,000 monthly active devices, and there's never a limit on how often each device runs it.

Start here

Community

Free

For indie developers, startups and growing apps, up to 10,000 monthly active devices.

Commercial use allowed
No API key, account or license file
Unlimited calls on every device
Runs fully on-device, with no tracking or phoning home

Get started

Commercial

Let's talk

For products above 10,000 monthly active devices on any platform.

One license per product, per model
Pricing that fits your scale
Direct support from the people who built it
Early access to model updates

Email WhatsApp Call

Custom models

Built for you

A small, fast on-device model trained for your language, domain or task.

Moderation, classification, detection, replies and more
Tuned on your use case and your users' languages
Shipped as a plain Kotlin library, like Comeback
Private: your model runs on your users' devices

Email WhatsApp Call

How devices are counted

A monthly active device is a phone, tablet or other device that runs Comeback at least once in a calendar month.

The 10,000 limit applies separately to each product, each platform (Android, iOS, web…) and each Hoverfly model. Crossed it? You have 30 days to get a commercial license.

What the license does not allow

Selling or redistributing Comeback or its model on its own, or inside another SDK
Extracting, modifying or fine-tuning the model weights
Using the model or its outputs to train or distill another model
Reverse engineering the model or its file format
Offering it as a hosted API for others

Questions about licensing? Email raju348636@gmail.com or call / WhatsApp +91 63533 21951. Full terms: Hoverfly Community License.

FAQ

Questions

What is Comeback?

A Kotlin Multiplatform library (Android, iOS, macOS, JVM desktop, JavaScript, WebAssembly) that reads a message and returns short replies to show as tappable chips, like the smart replies in messaging apps.

Does it generate text?

No. It scores a fixed list of about 1,300 reviewed replies and returns the best ones, so every possible output is known in advance.

Which languages does it support?

English in v1. To detect the language first and show chips only for English, use Beacon.

Does it run on device?

Yes. Inference runs entirely on the device in plain Kotlin. It makes no server calls, needs no network permission and sends no telemetry.

Is the Kotlin port exact?

Features, the top 5 replies and the 3 shown chips match the Python reference on all 354 test vectors, both on the JVM and on an Android device.

How much does Comeback cost?

Nothing for products with up to 10,000 monthly active devices per platform, including commercial apps. There is no API key or account, and no limit on how often it runs. Above that, a commercial license is needed. See Pricing.

What happens when my app passes 10,000 monthly active devices?

Nothing breaks: Comeback keeps working and never checks in with a server. You have 30 days to get a commercial license. Email raju348636@gmail.com or call / WhatsApp +91 63533 21951.

Can you build a model for my app?

Yes. Hoverfly trains custom on-device models for your language, domain or task and ships them as a small library like this one. Get in touch on the same email or number.

Try the real model, right here

Smart reply without the baggage

Safe by design

Chips that differ

No dependencies

Private & offline

Tap a chip, send a reply

Plans

Thanks

Sent

Same library on every platform

Android

iOS

Desktop (JVM)

Web (Wasm)

Add it to your build

Three lines to reply chips

Where Comeback fits

Chat & messaging apps

Notification quick replies

Wear OS & small screens

Community & marketplace inboxes

Accessibility

Offline-first apps

20 apps you could build with Comeback

Honest numbers

API at a glance

Free until you're big.

How devices are counted

What the license does not allow

Questions

What is Comeback?

Does it generate text?

Which languages does it support?

Does it run on device?

Is the Kotlin port exact?

How much does Comeback cost?

What happens when my app passes 10,000 monthly active devices?

Can you build a model for my app?

Comeback is one of eight.