File size: 5,172 Bytes
3b4968e 7db8704 3b4968e 5dda573 3b4968e c5a6d28 7db8704 3b4968e 7db8704 3b4968e dad5570 09351a6 3b4968e 09351a6 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 |
from fastapi import FastAPI, UploadFile, File, Response, Request
from fastapi.staticfiles import StaticFiles
import ggwave
import scipy.io.wavfile as wav
import numpy as np
import os
from pydantic import BaseModel
from groq import Groq
import io
import wave
app = FastAPI()
# Serve static files
app.mount("/static", StaticFiles(directory="static"), name="static")
# Initialize ggwave instance
instance = ggwave.init()
# Initialize Groq client
client = Groq(api_key=os.environ.get("GROQ_API_KEY"))
class TextInput(BaseModel):
text: str
@app.get("/")
async def serve_homepage():
"""Serve the chat interface HTML."""
with open("static/index.html", "r") as f:
return Response(content=f.read(), media_type="text/html")
@app.post("/stt/")
async def speech_to_text(file: UploadFile = File(...)):
"""Convert WAV audio file to text using ggwave."""
with open("temp.wav", "wb") as audio_file:
audio_file.write(await file.read())
# Load WAV file
fs, recorded_waveform = wav.read("temp.wav")
os.remove("temp.wav")
# Convert to bytes and decode
waveform_bytes = recorded_waveform.astype(np.uint8).tobytes()
decoded_message = ggwave.decode(instance, waveform_bytes)
return {"text": decoded_message}
@app.post("/tts/")
def text_to_speech(input_text: TextInput):
"""Convert text to a WAV audio file using ggwave and return as response."""
encoded_waveform = ggwave.encode(input_text.text, protocolId=1, volume=100)
# Convert byte data into float32 array
waveform_float32 = np.frombuffer(encoded_waveform, dtype=np.float32)
# Normalize float32 data to the range of int16
waveform_int16 = np.int16(waveform_float32 * 32767)
# Save to buffer instead of a file
buffer = io.BytesIO()
with wave.open(buffer, "wb") as wf:
wf.setnchannels(1) # Mono audio
wf.setsampwidth(2) # 2 bytes per sample (16-bit PCM)
wf.setframerate(48000) # Sample rate
wf.writeframes(waveform_int16.tobytes()) # Write waveform as bytes
buffer.seek(0)
return Response(content=buffer.getvalue(), media_type="audio/wav")
@app.post("/chat/")
async def chat_with_llm(file: UploadFile = File(...)):
"""Process input WAV, send text to LLM, and return generated response as WAV."""
# Initialize ggwave instance
instance = ggwave.init()
# Read the file content into memory without saving to disk
file_content = await file.read()
# Create a BytesIO object to use with wav.read
with io.BytesIO(file_content) as buffer:
try:
fs, recorded_waveform = wav.read(buffer)
recorded_waveform = recorded_waveform.astype(np.float32) / 32767.0
waveform_bytes = recorded_waveform.tobytes()
user_message = ggwave.decode(instance, waveform_bytes)
if user_message is None:
return Response(
content="No message detected in audio",
media_type="text/plain",
status_code=400
)
print("user_message: " + user_message.decode("utf-8"))
# Send to LLM
chat_completion = client.chat.completions.create(
messages=[
{"role": "system", "content": "you are a helpful assistant. answer always in one sentence"},
{"role": "user", "content": user_message.decode("utf-8")}
],
model="llama-3.3-70b-versatile",
)
llm_response = chat_completion.choices[0].message.content
print(llm_response)
# Convert response to audio
encoded_waveform = ggwave.encode(llm_response, protocolId=1, volume=100)
# Convert byte data into float32 array
waveform_float32 = np.frombuffer(encoded_waveform, dtype=np.float32)
# Normalize float32 data to the range of int16
waveform_int16 = np.int16(waveform_float32 * 32767)
# Save to buffer instead of a file
buffer = io.BytesIO()
with wave.open(buffer, "wb") as wf:
wf.setnchannels(1) # Mono audio
wf.setsampwidth(2) # 2 bytes per sample (16-bit PCM)
wf.setframerate(48000) # Sample rate
wf.writeframes(waveform_int16.tobytes()) # Write waveform as bytes
buffer.seek(0)
return Response(
content=buffer.getvalue(),
media_type="audio/wav",
headers={
"X-User-Message": user_message.decode("utf-8"),
"X-LLM-Response": llm_response
}
)
except Exception as e:
print(f"Error processing audio: {str(e)}")
return Response(
content=f"Error processing audio: {str(e)}",
media_type="text/plain",
status_code=500
) |