Generate smart replies with ML Kit on Android

ML Kit can generate short replies to messages using an on-device model.

To generate smart replies, you pass ML Kit a log of recent messages in a conversation. If ML Kit determines the conversation is in English, and that the conversation doesn't have potentially sensitive subject matter, ML Kit generates up to three replies, which you can suggest to your user.

See the ML Kit quickstart sample on GitHub for an example of this API in use.

Before you begin

  1. In your project-level build.gradle file, make sure to include Google's Maven repository in both your buildscript and allprojects sections.
  2. Add the dependencies for the ML Kit Android libraries to your module's app-level gradle file, which is usually app/build.gradle:
    dependencies {
      // ...
    
      implementation 'com.google.mlkit:smart-reply:16.1.0'
    }
    
  3. Also in your app-level build.gradle file, disable compression of tflite files:
    android {
        // ...
        aaptOptions {
            noCompress "tflite"
        }
    }
    

1. Create a conversation history object

To generate smart replies, you pass ML Kit a chronologically-ordered List of TextMessage objects, with the earliest timestamp first.

Whenever the user sends a message, add the message and its timestamp to the conversation history:

Java

conversation.add(TextMessage.createForLocalUser(
        "heading out now", System.currentTimeMillis()));

Kotlin

conversation.add(TextMessage.createForLocalUser(
        "heading out now", System.currentTimeMillis()))

Whenever the user receives a message, add the message, its timestamp, and the sender's user ID to the conversation history. The user ID can be any string that uniquely identifies the sender within the conversation. The user ID doesn't need to correspond to any user data, and the user ID doesn't need to be consistent between conversation or invocations of the smart reply generator.

Java

conversation.add(TextMessage.createForRemoteUser(
        "Are you coming back soon?", System.currentTimeMillis(), userId));

Kotlin

conversation.add(TextMessage.createForRemoteUser(
        "Are you coming back soon?", System.currentTimeMillis(), userId))

A conversation history object looks like the following example:

Timestamp userID isLocalUser Message
Thu Feb 21 13:13:39 PST 2019 true are you on your way?
Thu Feb 21 13:15:03 PST 2019 FRIEND0 false Running late, sorry!

ML Kit suggests replies to the last message in a conversation history. The last message should be from a non-local user. In the example above, the last message in the conversation is from the non-local user FRIEND0. When you use pass ML Kit this log, it suggests replies to FRIENDO's message: "Running late, sorry!"

2. Get message replies

To generate smart replies to a message, get an instance of SmartReplyGenerator and pass the conversation history to its suggestReplies() method:

Java

SmartReplyGenerator smartReply = SmartReply.getClient();
smartReply.suggestReplies(conversation)
        .addOnSuccessListener(new OnSuccessListener<SmartReplySuggestionResult>() {
            @Override
            public void onSuccess(SmartReplySuggestionResult result) {
                if (result.getStatus() == SmartReplySuggestionResult.STATUS_NOT_SUPPORTED_LANGUAGE) {
                    // The conversation's language isn't supported, so
                    // the result doesn't contain any suggestions.
                } else if (result.getStatus() == SmartReplySuggestionResult.STATUS_SUCCESS) {
                    // Task completed successfully
                    // ...
                }
            }
        })
        .addOnFailureListener(new OnFailureListener() {
            @Override
            public void onFailure(@NonNull Exception e) {
                // Task failed with an exception
                // ...
            }
        });

Kotlin

val smartReplyGenerator = SmartReply.getClient()
smartReply.suggestReplies(conversation)
        .addOnSuccessListener { result ->
            if (result.getStatus() == SmartReplySuggestionResult.STATUS_NOT_SUPPORTED_LANGUAGE) {
                // The conversation's language isn't supported, so
                // the result doesn't contain any suggestions.
            } else if (result.getStatus() == SmartReplySuggestionResult.STATUS_SUCCESS) {
                // Task completed successfully
                // ...
            }
        }
        .addOnFailureListener {
            // Task failed with an exception
            // ...
        }

If the operation succeeds, a SmartReplySuggestionResult object is passed to the success handler. This object contains a list of up to three suggested replies, which you can present to your user:

Java

for (SmartReplySuggestion suggestion : result.getSuggestions()) {
    String replyText = suggestion.getText();
}

Kotlin

for (suggestion in result.suggestions) {
    val replyText = suggestion.text
}

Note that ML Kit might not return results if the model isn't confident in the relevance of the suggested replies, the input conversation isn't in English, or if the model detects sensitive subject matter.

Next steps

See the ML Kit quickstart sample on GitHub for an example of this API in use.