Adolfo Reyna
249 lines
9.8 KiB
Python
249 lines
9.8 KiB
Python
import time
|
|
import struct
|
|
import machine
|
|
from machine import Pin, SPI, I2C, I2S
|
|
import urequests
|
|
import ili9341
|
|
from ft6336u import FT6336U
|
|
from audio_util import ES8311
|
|
|
|
# --- CONFIGURATION ---
|
|
HERMES_API_URL = "http://192.168.68.126:8642/api/esp32/voice"
|
|
DEVICE_ID = "kitchen-button"
|
|
|
|
# Placeholder for API Key. Since the key is stored on the Hermes server,
|
|
# please copy-paste the token string here.
|
|
HERMES_API_KEY = "mcT1YA1vOr9wXSiHpCYalweEGGZKX-PIfZv2drp8BSg"
|
|
|
|
def create_wav_header(data_size):
|
|
# Generates a 44-byte WAV header for 16kHz, 16-bit mono PCM
|
|
riff = b'RIFF'
|
|
file_size = data_size + 36
|
|
wave = b'WAVE'
|
|
fmt = b'fmt '
|
|
chunk_size = 16
|
|
audio_format = 1 # PCM
|
|
channels = 1 # Mono
|
|
sample_rate = 16000
|
|
bits_per_sample = 16
|
|
byte_rate = sample_rate * channels * (bits_per_sample // 8)
|
|
block_align = channels * (bits_per_sample // 8)
|
|
data_label = b'data'
|
|
|
|
return struct.pack('<4sI4s4sIHHIIHH4sI',
|
|
riff, file_size, wave, fmt, chunk_size,
|
|
audio_format, channels, sample_rate, byte_rate,
|
|
block_align, bits_per_sample, data_label, data_size)
|
|
|
|
def main():
|
|
print("--- Starting Hermes Voice Assistant Demo ---")
|
|
|
|
# 1. Initialize display
|
|
spi = SPI(1, baudrate=40000000, polarity=0, phase=0, sck=Pin(12), mosi=Pin(11), miso=Pin(13))
|
|
display = ili9341.ILI9341(spi, cs=Pin(10), dc=Pin(46), bl=Pin(45), rst=None)
|
|
display.clear(0)
|
|
display.text("Hermes Assistant", 10, 10, 1)
|
|
display.show()
|
|
|
|
# 2. Initialize touch
|
|
i2c = I2C(0, sda=Pin(16), scl=Pin(15))
|
|
touch = FT6336U(i2c, rst_pin=18, int_pin=17, width=320, height=240, swap_xy=True, invert_x=False, invert_y=True)
|
|
|
|
# 3. Configure audio codec ES8311
|
|
# We need MCLK pin (GPIO 4) active at 6.144 MHz
|
|
mclk_pin = Pin(4, Pin.OUT)
|
|
mclk_pwm = machine.PWM(mclk_pin)
|
|
mclk_pwm.freq(6144000)
|
|
mclk_pwm.duty_u16(32768)
|
|
|
|
codec = ES8311(i2c)
|
|
if not codec.init(sample_rate=16000):
|
|
print("ES8311 init failed!")
|
|
return
|
|
|
|
codec.set_volume(90)
|
|
|
|
# Configure microphone registers on ES8311
|
|
try:
|
|
codec._write(0x14, 0x1A) # Enable analog mic input & set PGA gain
|
|
codec._write(0x16, 0x01) # Boost MIC digital gain to +6dB
|
|
codec._write(0x17, 0xC8) # Set ADC digital volume
|
|
print("Microphone registers initialized.")
|
|
except Exception as e:
|
|
print("Failed to write microphone registers:", e)
|
|
|
|
amp_pin = Pin(1, Pin.OUT, value=1) # start with amp disabled (1 = disabled)
|
|
|
|
buffer = bytearray(1024)
|
|
|
|
while True:
|
|
display.clear(0)
|
|
display.text("Hermes Assistant", 10, 10, 1)
|
|
display.line(10, 22, 310, 22, 1)
|
|
display.text("Hold screen & ask", 50, 70, 1)
|
|
display.text("a question...", 50, 90, 1)
|
|
display.text("Status: Idle", 10, 220, 1)
|
|
display.show()
|
|
|
|
# Wait for touch
|
|
while not touch.is_touched():
|
|
time.sleep_ms(30)
|
|
|
|
print("Touch detected! Starting recording to RAM...")
|
|
display.clear(0)
|
|
display.text("Hermes Assistant", 10, 10, 1)
|
|
display.line(10, 22, 310, 22, 1)
|
|
display.text("Listening...", 80, 80, 1)
|
|
display.fill_rect(130, 110, 30, 30, 1)
|
|
display.text("Status: Recording", 10, 220, 1)
|
|
display.show()
|
|
|
|
# 4. Open I2S RX for recording (Mono 16kHz for Hermes STT)
|
|
i2s_rx = I2S(1,
|
|
sck=Pin(5),
|
|
ws=Pin(7),
|
|
sd=Pin(6),
|
|
mode=I2S.RX,
|
|
ibuf=8000,
|
|
rate=16000,
|
|
bits=16,
|
|
format=I2S.MONO)
|
|
|
|
audio_chunks = []
|
|
total_data_bytes = 0
|
|
|
|
try:
|
|
# Record loop - as long as touch is held
|
|
while touch.is_touched():
|
|
bytes_read = i2s_rx.readinto(buffer)
|
|
if bytes_read > 0:
|
|
audio_chunks.append(bytes(buffer[:bytes_read]))
|
|
total_data_bytes += bytes_read
|
|
|
|
print(f"Touch released! Recorded {total_data_bytes} bytes.")
|
|
|
|
except Exception as e:
|
|
print("Error recording:", e)
|
|
finally:
|
|
i2s_rx.deinit()
|
|
|
|
if total_data_bytes < 1000:
|
|
print("Recording too short, ignoring.")
|
|
continue
|
|
|
|
# 5. Connect and POST to Hermes voice endpoint
|
|
display.clear(0)
|
|
display.text("Hermes Assistant", 10, 10, 1)
|
|
display.line(10, 22, 310, 22, 1)
|
|
display.text("Sending audio...", 50, 80, 1)
|
|
display.text("Waiting for agent...", 50, 100, 1)
|
|
display.text("Status: Processing", 10, 220, 1)
|
|
display.show()
|
|
|
|
# Join chunks and prepend WAV header
|
|
raw_pcm = b"".join(audio_chunks)
|
|
wav_data = create_wav_header(len(raw_pcm)) + raw_pcm
|
|
|
|
# Headers
|
|
headers = {
|
|
"Authorization": f"Bearer {HERMES_API_KEY}",
|
|
"Content-Type": "audio/wav",
|
|
"X-Device-ID": DEVICE_ID
|
|
}
|
|
|
|
print(f"Posting raw WAV ({len(wav_data)} bytes) to {HERMES_API_URL}...")
|
|
try:
|
|
res = urequests.post(HERMES_API_URL, headers=headers, data=wav_data)
|
|
print(f"HTTP Status: {res.status_code}")
|
|
|
|
if res.status_code == 200:
|
|
response_data = res.content
|
|
print(f"Received {len(response_data)} bytes of audio response.")
|
|
|
|
# Check response format
|
|
if response_data[0:4] == b'RIFF' and response_data[8:12] == b'WAVE':
|
|
# Parse WAV header
|
|
idx = response_data.find(b'fmt ')
|
|
if idx != -1:
|
|
fmt_chunk = response_data[idx:idx+24]
|
|
audio_format, channels, sample_rate, byte_rate, block_align, bits = struct.unpack('<HHIIHH', fmt_chunk[8:24])
|
|
print(f"Playing WAV: {sample_rate}Hz, {channels}ch, {bits}bit")
|
|
|
|
# Find data chunk
|
|
data_idx = response_data.find(b'data')
|
|
if data_idx != -1:
|
|
audio_start = data_idx + 8
|
|
|
|
display.clear(0)
|
|
display.text("Hermes Assistant", 10, 10, 1)
|
|
display.line(10, 22, 310, 22, 1)
|
|
display.text("Playing response...", 50, 80, 1)
|
|
display.text("Status: Speaking", 10, 220, 1)
|
|
display.show()
|
|
|
|
# Enable amp & play
|
|
amp_pin.value(0) # Active Low Enable
|
|
i2s_format = I2S.MONO if channels == 1 else I2S.STEREO
|
|
i2s_tx = I2S(1,
|
|
sck=Pin(5),
|
|
ws=Pin(7),
|
|
sd=Pin(8),
|
|
mode=I2S.TX,
|
|
ibuf=4096,
|
|
rate=sample_rate,
|
|
bits=bits,
|
|
format=i2s_format)
|
|
try:
|
|
pos = audio_start
|
|
chunk_size = 2048
|
|
while pos < len(response_data):
|
|
chunk = response_data[pos:pos+chunk_size]
|
|
i2s_tx.write(chunk)
|
|
pos += chunk_size
|
|
finally:
|
|
time.sleep_ms(150)
|
|
amp_pin.value(1) # Disable
|
|
i2s_tx.deinit()
|
|
else:
|
|
print("Error: Could not parse WAV format.")
|
|
else:
|
|
# Check if it is MP3
|
|
if response_data[0:3] == b'ID3' or (len(response_data) > 2 and response_data[0] == 0xFF and (response_data[1] & 0xE0) == 0xE0):
|
|
print("Error: Server returned MP3 format. ESP32 requires WAV.")
|
|
display.clear(0)
|
|
display.text("Error: Got MP3", 10, 10, 1)
|
|
display.line(10, 22, 310, 22, 1)
|
|
display.text("Hermes returned MP3.", 30, 80, 1)
|
|
display.text("Configure server to", 30, 100, 1)
|
|
display.text("convert MP3 to WAV.", 30, 120, 1)
|
|
display.show()
|
|
time.sleep(4)
|
|
else:
|
|
print(f"Error: Unknown response format. Starts with: {response_data[0:16]}")
|
|
else:
|
|
print(f"Request failed with status {res.status_code}: {res.text}")
|
|
display.clear(0)
|
|
display.text("HTTP Error", 10, 10, 1)
|
|
display.line(10, 22, 310, 22, 1)
|
|
display.text(f"Status: {res.status_code}", 50, 80, 1)
|
|
display.show()
|
|
time.sleep(3)
|
|
|
|
except Exception as e:
|
|
print("Failed to contact Hermes server:", e)
|
|
display.clear(0)
|
|
display.text("Server Error", 10, 10, 1)
|
|
display.line(10, 22, 310, 22, 1)
|
|
display.text("Could not connect", 50, 80, 1)
|
|
display.text("to API gateway.", 50, 100, 1)
|
|
display.show()
|
|
time.sleep(3)
|
|
|
|
# Debounce touch release
|
|
while touch.is_touched():
|
|
time.sleep_ms(30)
|
|
time.sleep_ms(300)
|
|
|
|
if __name__ == "__main__":
|
|
main()
|