Sample: Using Whisper API with M5 CoreS3

main.cpp

// Copyright (c) 2023 Inaba (@hollyhockberry)
// This software is released under the MIT License.
// http://opensource.org/licenses/mit-license.php

#include <SD.h>
#include <M5Unified.h>
#include <WiFi.h>
#include <HTTPClient.h>
#include <ArduinoJson.h>
#include <gob_unifiedButton.hpp>

constexpr char* ROOT_CA = "openai.comの証明書";
constexpr char* API_KEY = "API KEY";

gob::UnifiedButton unfiedButton;

bool transcriptions(File& file, String* result) {
  char boundary[64] = "------------------------";
  for (auto i = 0; i < 2; ++i) {
    ltoa(random(0x7fffffff), boundary + strlen(boundary), 16);
  }

  const char* API_HOST = "api.openai.com";
  const uint16_t API_PORT = 443;
  const char* API_PATH = "/v1/audio/transcriptions";
  WiFiClientSecure client;
  client.setCACert(ROOT_CA);

  if (!client.connect(API_HOST, API_PORT)) {
    Serial.println("connection failed");
    return false;
  }

  const String header = "--" + String(boundary) + "\r\n"
    "Content-Disposition: form-data; name=\"model\"\r\n\r\nwhisper-1\r\n"
    "--" + String(boundary) + "\r\n"
    "Content-Disposition: form-data; name=\"file\"; filename=\"" + file.name() + "\"\r\n"
    "Content-Type: application/octet-stream\r\n\r\n";
  const String footer = "\r\n--" + String(boundary) + "--\r\n";

  // header
  client.printf("POST %s HTTP/1.1\n", API_PATH);
  client.printf("Host: %s\n", API_HOST);
  client.println("Accept: */*");
  client.printf("Authorization: Bearer %s\n", API_KEY);
  client.printf("Content-Length: %d\n", header.length() + file.size() + footer.length());
  client.printf("Content-Type: multipart/form-data; boundary=%s\n", boundary);
  client.println();
  client.print(header.c_str());
  client.flush();

  // body
  uint8_t buf[512];
  while (file.available()) {
    const auto sz = file.read(buf, sizeof(buf));
    client.write(buf, sz);
    client.flush();
  }
  client.flush();
  file.close();

  // footer
  client.print(footer.c_str());
  client.flush();

  // wait response
  while (client.available() == 0) {
  }

  bool isBody = false;
  String body = "";
  while(client.available()){
    const auto line = client.readStringUntil('\r');
    if (isBody) {
      body += line;
    } else if (line.equals("\n")) {
      isBody = true;
    }
  }
  client.stop();

  StaticJsonDocument<200> doc;
  deserializeJson(doc, body);
  const auto text = doc["text"].as<String>();
  if (text.length() <= 0) {
    return false;
  } 
  *result = text;
  return true;
}

void setup() {
  M5.begin();
  unfiedButton.begin(&M5.Display);
  SD.begin(GPIO_NUM_4, SPI, 25000000);

  WiFi.begin();
  while (WiFi.status() != WL_CONNECTED) {
    ::delay(500);
  }
  Serial.println("WiFi connected");
  ::randomSeed(::analogRead(G6));
}

void loop() {
  unfiedButton.update();
  M5.update();
  if (!M5.BtnA.wasPressed()) {
    return;
  }

  File f = SD.open("/sample.wav");
  if (!f) {
    Serial.println("file open error");
    return;
  }
  String text;
  if (transcriptions(f, &text)) {
    Serial.println(text);
  }
}