syntax = "proto3";
option java_multiple_files = true;
package br.com.cpqd.asr.grpc;

message RecognitionConfig {
    repeated LanguageModel lm = 1;
    AudioEncoding audio_encoding = 2;
    optional int32 sample_rate = 3;

    optional EndpointerConfig endpointer = 4;

    optional string word_hints = 5;
    optional bool textify_enabled = 6;
    optional int32 noinput_timeout = 7;
    optional bool noinput_timeout_enabled = 8;
    optional int32 recognition_timeout = 9;
    optional bool recognition_timeout_enabled = 10;
    optional int32 confidence_threshold = 11;
    optional int32 max_sentences = 12;
    optional bool continuous_mode = 13;
    optional WordDetails word_details = 14;
    optional string logging_tag = 15;
    optional string account_tag = 16;
    optional bool age_scores_enabled = 17;
    optional bool gender_scores_enabled = 18;
    optional bool emotion_scores_enabled = 19;
    optional string channel_identifier = 20;
    optional bool verify_buffer_utterance = 21;
    optional bool start_input_timers = 22;
    optional bool endpointer_enabled = 23;

    message LanguageModel {
        optional string content_type = 1;
        optional string uri = 2;
        optional string definition = 3;
        optional string id = 4;
    }
    
    enum AudioEncoding {
        UNSPECIFIED = 0;
        LINEAR16 = 1;
        WAV = 2;
        ALAW = 3;
        MULAW = 4;
        MP3 = 5;
        OGG_OPUS = 6;
        OGG_VORBIS = 7;
        FLAC = 8;
    }

    message EndpointerConfig {
        optional bool use_tone_detectors = 1;
        optional int32 head_margin = 2;
        optional int32 tail_margin = 3;
        optional int32 wait_end = 4;
        optional LevelMode level_mode = 5;
        optional int32 level_threshold = 6;
        optional int32 auto_level_len = 7;

        enum LevelMode {
            IGNORE = 0;
            AUTO = 1;
            FIXED = 2;
        }
    }

    enum WordDetails {
        NONE = 0;
        FIRST = 1;
        ALL = 2;
    }
}

enum RecognitionEvent {
    START_OF_SPEECH = 0;
    END_OF_SPEECH = 1;
    RECOGNITION_RESULT = 2;
    FAILURE = 3;
    TIMEOUT = 4;
    INVALID_REQUEST = 5;
    LISTENING = 6;
}

message RecognitionResult {
    RecognitionStatus status = 1;
    optional int32 segment_index = 2;
    optional bool last_segment = 3;
    optional bool final_result = 4;
    optional float start_time = 5;
    optional float end_time = 6;
    repeated RecognitionAlternative alternatives = 7;

    enum RecognitionStatus {
        NONE = 0;
        PROCESSING = 1;
        RECOGNIZED = 2;
        NO_MATCH = 3;
        NO_INPUT_TIMEOUT = 4;
        MAX_SPEECH = 5;
        EARLY_SPEECH = 6;
        RECOGNITION_TIMEOUT = 7;
        NO_SPEECH = 8;
        CANCELED = 9;
        FAILURE = 10;
    }

    message RecognitionAlternative {
        string text = 1;
        int32 score = 2;
        string lm = 3;
        repeated string interpretations = 4;
        repeated int32 interpretation_scores = 5;
        repeated Words words = 6;

        message Words {
            string text = 1;
            int32 score = 2;
            float startTime = 3;
            float endTime = 4;
        }
    }
}

message RecognizeRequest {
    RecognitionConfig config = 1;
    optional int32 timeout = 2;
    bytes media = 3;
}

message RecognizeResponse {
    optional RecognitionEvent event = 1;
    optional string error_code = 2;
    optional string error_message = 3;
    repeated RecognitionResult result = 4;
    optional int64 audio_duration = 5;
}

message StreamingRecognizeRequest {
    optional RecognitionConfig config = 1;
    optional int32 timeout = 2;
    optional bool last_packet = 3;
    optional bytes media = 4;
}

message StreamingRecognizeResponse {
    RecognitionEvent event = 1;
    optional string error_code = 2;
    optional string error_message = 3;
    repeated RecognitionResult result = 4;
    optional int64 audio_duration = 5;
}

service RecognizeService {
    rpc Recognize(RecognizeRequest) returns (RecognizeResponse);
    rpc StreamingRecognize(stream StreamingRecognizeRequest) returns (stream StreamingRecognizeResponse);
}