A plushie that talks back, recorded

# ESP32 Toy Backend 123
from flask import Flask, request, send_file, jsonify
import os, datetime, time
import struct
import wave

from whisper_stt import transcribe_audio
from gpt_reply import get_gpt_reply
from tts import synthesize_speech

app = Flask(__name__)
TEMP_DIR = os.path.join(os.path.dirname(os.path.dirname(__file__)), "temp")
AUDIO_DIR = os.path.join(os.path.dirname(__file__), "audio")
os.makedirs(TEMP_DIR, exist_ok=True)
os.makedirs(AUDIO_DIR, exist_ok=True)

def decompress_adpcm_to_wav(adpcm_data, output_path, sample_rate=16000):
    """
    Decompress ADPCM data to WAV format
    Assumes IMA ADPCM format commonly used by ESP32
    """
    try:
        # IMA ADPCM step size table
        step_table = [
            7, 8, 9, 10, 11, 12, 13, 14, 16, 17,
            19, 21, 23, 25, 28, 31, 34, 37, 41, 45,
            50, 55, 60, 66, 73, 80, 88, 97, 107, 118,
            130, 143, 157, 173, 190, 209, 230, 253, 279, 307,
            337, 371, 408, 449, 494, 544, 598, 658, 724, 796,
            876, 963, 1060, 1166, 1282, 1411, 1552, 1707, 1878, 2066,
            2272, 2499, 2749, 3024, 3327, 3660, 4026, 4428, 4871, 5358,
            5894, 6484, 7132, 7845, 8630, 9493, 10442, 11487, 12635, 13899,
            15289, 16818, 18500, 20350, 22385, 24623, 27086, 29794, 32767
        ]
        
        # Index table for IMA ADPCM (matches ESP32 implementation)
        index_table = [-1, -1, -1, -1, 2, 4, 6, 8, -1, -1, -1, -1, 2, 4, 6, 8]
        
        # Initialize decoder state
        predicted_sample = 0
        step_index = 0
        decoded_samples = []
        
        # Process ADPCM data
        for byte in adpcm_data:
            # Each byte contains two 4-bit ADPCM samples
            for nibble in [(byte & 0x0F), (byte >> 4)]:
                step = step_table[step_index]
                
                # Decode the nibble (matching ESP32 encoder logic)
                diffq = step >> 3
                if nibble & 4:
                    diffq += step
                if nibble & 2:
                    diffq += step >> 1
                if nibble & 1:
                    diffq += step >> 2
                
                if nibble & 8:
                    predicted_sample -= diffq
                else:
                    predicted_sample += diffq
                
                # Clamp to 16-bit range
                predicted_sample = max(-32768, min(32767, predicted_sample))
                decoded_samples.append(predicted_sample)
                
                # Update step index
                step_index += index_table[nibble]
                step_index = max(0, min(88, step_index))

/*
 * PlushieAI ESP32-S3 WiFi Manager & Backend Connection with Audio Playback + Voice Recording
 * 
 * Required Libraries (install via Arduino Library Manager or PlatformIO):
 * - ESPAsyncWebServer: https://github.com/me-no-dev/ESPAsyncWebServer
 * - AsyncTCP: https://github.com/me-no-dev/AsyncTCP
 * - DNSServer: Built-in with ESP32 core
 * - Preferences: Built-in with ESP32 core
 * - HTTPClient: Built-in with ESP32 core
 * - WiFi: Built-in with ESP32 core
 * - FFat: Built-in with ESP32 core
 * - I2S: Built-in with ESP32 core
 * 
 * Features:
 * - WiFi management with captive portal
 * - Audio playback on backend connection
 * - Press-to-record voice capture with PSRAM buffering
 * - ADPCM compression and HTTP upload
 * - 
 */

#include <WiFi.h>
#include <ESPAsyncWebServer.h>
#include <DNSServer.h>
#include <Preferences.h>
#include <HTTPClient.h>
#include <FFat.h>
#include <driver/i2s.h>
#include <vector>
#include <esp_heap_caps.h>
#include <esp_psram.h>
#define USE_ADPCM_COMPRESSION true

// Compile-time check for ESP32-S3
#if !CONFIG_IDF_TARGET_ESP32S3
#error "This code is designed specifically for ESP32-S3"
#endif

// Hardware pin definitions
#define BUTTON_PIN          21
#define MIC_POWER_PIN       2

// I2S ports
#define MIC_I2S_PORT        I2S_NUM_0
#define SPK_I2S_PORT        I2S_NUM_1

// Microphone pins
#define I2S_MIC_WS          15
#define I2S_MIC_SCK         18
#define I2S_MIC_SD          6

// Speaker pins
#define I2S_SPK_WS          4
#define I2S_SPK_SCK         5
#define I2S_SPK_DIN         17

// Constants
#define LED_BUILTIN         RGB_BUILTIN
#define WIFI_TIMEOUT        10000  // 10 seconds
#define SERVER_RETRY_DELAY  5000   // 5 seconds
#define AP_NAME             "PlushieAI-Setup"
#define BACKEND_HOST        "http://18.190.135.144"
#define WAKE_FILE_PATH      "/wake.wav"
#define MIN_FILE_SIZE       1024   // 1KB minimum file size

// Voice recording constants
#define SAMPLE_RATE         16000
#define BITS_PER_SAMPLE     16
#define DEBOUNCE_TIME_MS    50
#define INITIAL_BUFFER_SIZE (1024 * 512)  // 512KB initial
#define BUFFER_GROW_SIZE    (1024 * 256)  // 256KB growth chunks
#define MAX_PSRAM_USAGE     0.9f          // 90% of available PSRAM
#define RECORDING_GAIN      2.5f          // Volume boost multiplier
#define SOFT_LIMIT_THRESH   0.5f          // Threshold for soft limiting (50% of max)
#define LOW_PASS_ALPHA      0.15f         // Low-pass filter coefficient (0.1-0.3, lower = more filtering)

// Audio compression constants
#ifdef USE_ADPCM_COMPRESSION
#define ADPCM_BLOCK_SIZE    256           // Process in 256 sample blocks
#else