Push-To-Talk AI Assistant

#include <Arduino.h>
#include <WiFi.h>
#include <HTTPClient.h>
#include <ArduinoJson.h>
#include <driver/i2s.h>

#define WIFI_SSID "YOUR_WIFI_SSID"
#define WIFI_PASS "YOUR_WIFI_PASSWORD"
#define OPENAI_KEY "sk-YOUR_OPENAI_API_KEY"
#define BTN_PIN 4
#define LED_R 16
#define LED_G 17
#define LED_B 18
#define MIC_SCK 32
#define MIC_WS 25
#define MIC_SD 33
#define DAC_BCK 26
#define DAC_LCK 27
#define DAC_DIN 14
#define SAMPLE_RATE 16000
#define RECORD_SECONDS 5
#define PCM_BYTES (SAMPLE_RATE * 2 * RECORD_SECONDS)


// Forward declarations
void led(bool r,bool g,bool b);
void initMic();
void initDac();
uint8_t* wavFromMic(size_t &wavLen);
String postJson(const char* url,const String &body);
String transcribe(uint8_t* wav,size_t n);
String askAI(const String& q);
void speak(const String& text);

void led(bool r,bool g,bool b){digitalWrite(LED_R,r);digitalWrite(LED_G,g);digitalWrite(LED_B,b);}
void initMic(){
  i2s_config_t c={.mode=(i2s_mode_t)(I2S_MODE_MASTER|I2S_MODE_RX),.sample_rate=SAMPLE_RATE,.bits_per_sample=I2S_BITS_PER_SAMPLE_32BIT,.channel_format=I2S_CHANNEL_FMT_ONLY_LEFT,.communication_format=I2S_COMM_FORMAT_STAND_I2S,.intr_alloc_flags=ESP_INTR_FLAG_LEVEL1,.dma_buf_count=8,.dma_buf_len=512,.use_apll=false,.tx_desc_auto_clear=false,.fixed_mclk=0};
  i2s_pin_config_t p={.bck_io_num=MIC_SCK,.ws_io_num=MIC_WS,.data_out_num=I2S_PIN_NO_CHANGE,.data_in_num=MIC_SD};
  i2s_driver_install(I2S_NUM_0,&c,0,NULL); i2s_set_pin(I2S_NUM_0,&p);
}
void initDac(){
  i2s_config_t c={.mode=(i2s_mode_t)(I2S_MODE_MASTER|I2S_MODE_TX),.sample_rate=24000,.bits_per_sample=I2S_BITS_PER_SAMPLE_16BIT,.channel_format=I2S_CHANNEL_FMT_RIGHT_LEFT,.communication_format=I2S_COMM_FORMAT_STAND_I2S,.intr_alloc_flags=ESP_INTR_FLAG_LEVEL1,.dma_buf_count=8,.dma_buf_len=512,.use_apll=false,.tx_desc_auto_clear=true,.fixed_mclk=0};
  i2s_pin_config_t p={.bck_io_num=DAC_BCK,.ws_io_num=DAC_LCK,.data_out_num=DAC_DIN,.data_in_num=I2S_PIN_NO_CHANGE};
  i2s_driver_install(I2S_NUM_1,&c,0,NULL); i2s_set_pin(I2S_NUM_1,&p);
}
uint8_t* wavFromMic(size_t &wavLen){
  static int16_t pcm[PCM_BYTES/2]; size_t used=0;
  while(used<PCM_BYTES){int32_t raw[128];size_t n=0;i2s_read(I2S_NUM_0,raw,sizeof(raw),&n,portMAX_DELAY);for(size_t i=0;i<n/4&&used<PCM_BYTES;i++) {pcm[used/2]=(int16_t)(raw[i]>>14);used+=2;}}
  wavLen=44+used; uint8_t* w=(uint8_t*)malloc(wavLen); if(!w)return nullptr;
  uint32_t ds=used, cs=36+ds, br=SAMPLE_RATE*2; uint16_t one=1, ba=2, bits=16, ch=1;
  memcpy(w,"RIFF",4);memcpy(w+4,&cs,4);memcpy(w+8,"WAVEfmt ",8);uint32_t fmt=16;memcpy(w+16,&fmt,4);memcpy(w+20,&one,2);memcpy(w+22,&ch,2);uint32_t sr=SAMPLE_RATE;memcpy(w+24,&sr,4);memcpy(w+28,&br,4);memcpy(w+32,&ba,2);memcpy(w+34,&bits,2);memcpy(w+36,"data",4);memcpy(w+40,&ds,4);memcpy(w+44,pcm,used);return w;
}
String postJson(const char* url,const String &body){HTTPClient h;h.begin(url);h.addHeader("Authorization",String("Bearer ")+OPENAI_KEY);h.addHeader("Content-Type","application/json");int rc=h.POST(body);String r=rc==200?h.getString():"";h.end();return r;}
String transcribe(uint8_t* wav,size_t n){
  HTTPClient h;h.begin("https://api.openai.com/v1/audio/transcriptions");h.addHeader("Authorization",String("Bearer ")+OPENAI_KEY);String b="----VoiceButton";String a="--"+b+"\r\nContent-Disposition: form-data; name=\"model\"\r\n\r\nwhisper-1\r\n--"+b+"\r\nContent-Disposition: form-data; name=\"file\"; filename=\"voice.wav\"\r\nContent-Type: audio/wav\r\n\r\n",z="\r\n--"+b+"--\r\n";uint8_t* body=(uint8_t*)malloc(a.length()+n+z.length());if(!body){h.end();return "";}memcpy(body,a.c_str(),a.length());memcpy(body+a.length(),wav,n);memcpy(body+a.length()+n,z.c_str(),z.length());h.addHeader("Content-Type","multipart/form-data; boundary="+b);int rc=h.POST(body,a.length()+n+z.length());free(body);String reply=rc==200?h.getString():"";h.end();JsonDocument d;deserializeJson(d,reply);return d["text"].as<String>();
}
String askAI(const String& q){JsonDocument d;d["model"]="gpt-4o-mini";JsonArray m=d["messages"].to<JsonArray>();JsonObject s=m.add<JsonObject>();s["role"]="system";s["content"]="You are a helpful voice assistant. Keep answers concise and natural for speech.";JsonObject u=m.add<JsonObject>();u["role"]="user";u["content"]=q;d["max_tokens"]=256;String b,r;serializeJson(d,b);r=postJson("https://api.openai.com/v1/chat/completions",b);deserializeJson(d,r);return d["choices"][0]["message"]["content"].as<String>();}
void speak(const String& text){JsonDocument d;d["model"]="tts-1";d["input"]=text;d["voice"]="nova";d["response_format"]="pcm";String b;serializeJson(d,b);HTTPClient h;h.begin("https://api.openai.com/v1/audio/speech");h.addHeader("Authorization",String("Bearer ")+OPENAI_KEY);h.addHeader("Content-Type","application/json");if(h.POST(b)!=200){h.end();return;}led(0,1,0);WiFiClient* s=h.getStreamPtr();uint8_t mono[512],stereo[1024];while(h.connected()||s->available()){int a=s->available();if(a<=0){delay(1);continue;}int n=s->readBytes(mono,min(a,512));int16_t* x=(int16_t*)mono;int16_t* y=(int16_t*)stereo;for(int i=0;i<n/2;i++){y[2*i]=x[i];y[2*i+1]=x[i];}size_t done;i2s_write(I2S_NUM_1,stereo,n*2,&done,portMAX_DELAY);}h.end();}
void setup(){Serial.begin(115200);pinMode(BTN_PIN,INPUT_PULLUP);pinMode(LED_R,OUTPUT);pinMode(LED_G,OUTPUT);pinMode(LED_B,OUTPUT);led(0,0,1);WiFi.begin(WIFI_SSID,WIFI_PASS);while(WiFi.status()!=WL_CONNECTED)delay(300);initMic();initDac();led(0,0,0);Serial.println("Ready — hold button to talk.");}
void loop(){if(digitalRead(BTN_PIN)!=LOW)return;delay(30);if(digitalRead(BTN_PIN)!=LOW)return;led(1,0,0);size_t n;uint8_t* w=wavFromMic(n);if(!w){led(0,0,0);return;}led(0,0,1);String q=transcribe(w,n);free(w);if(q.length()){Serial.println("You: "+q);String r=askAI(q);Serial.println("AI: "+r);if(r.length())speak(r);}led(0,0,0);while(digitalRead(BTN_PIN)==LOW)delay(10);}