#pragma once

#include <stdbool.h>
#include <stddef.h>
#include <stdint.h>

#include "esp_err.h"

typedef enum {
    VOICE_DIALOG_STATE_IDLE = 0,
    VOICE_DIALOG_STATE_LISTENING,
    VOICE_DIALOG_STATE_THINKING,
    VOICE_DIALOG_STATE_RESPONDING,
} voice_dialog_state_t;

typedef struct {
    bool ready;
    bool session_active;
    bool ws_connected;
    bool started;
    bool tap_active;
    voice_dialog_state_t dialog_state;
    char task_id[40];
    char dialog_id[40];
    char last_error[128];
    int64_t last_event_ms;
    uint32_t upstream_packets;
    uint32_t downstream_packets;
} voice_interaction_status_t;

esp_err_t voice_interaction_init(void);
esp_err_t voice_interaction_start(char *err, size_t err_len);
esp_err_t voice_interaction_stop(char *err, size_t err_len);

esp_err_t voice_interaction_tap_start(char *err, size_t err_len);
esp_err_t voice_interaction_tap_cancel(char *err, size_t err_len);

void voice_interaction_get_status(voice_interaction_status_t *out_status);
const char *voice_interaction_dialog_state_str(voice_dialog_state_t state);