File size: 8,526 Bytes
18092b2
3b4968e
b64bb23
3b4968e
 
 
 
 
 
 
7db8704
b64bb23
630fe1c
3b4968e
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
c88bfae
3b4968e
b64bb23
 
 
 
1b4cab5
3ba759c
39d90db
b64bb23
3b4968e
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5dda573
3b4968e
c5a6d28
7db8704
 
 
 
 
 
 
 
3b4968e
7db8704
 
 
 
 
 
3b4968e
 
 
 
 
 
bf48613
09351a6
 
3b4968e
09351a6
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
39d90db
 
 
 
 
 
 
 
 
 
 
 
 
bf48613
39d90db
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
from fastapi import FastAPI, UploadFile, File, Response, Request, Form, Body
from fastapi.staticfiles import StaticFiles
from fastapi.responses import FileResponse
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
import json
from typing import List, Dict, Optional

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/conv.html", "r") as f:
        return Response(content=f.read(), media_type="text/html")




@app.get("/conv/")
async def serve_convpage():
    """Serve the chat interface HTML."""
    return FileResponse("static/index.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."""
    global instance
    # 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
            )
@app.post("/continuous-chat/")
async def continuous_chat(
    file: UploadFile = File(...),
    chat_history: Optional[str] = Form(None)
):
    """Process input WAV with chat history, send text to LLM, and return response as WAV."""
    global instance
    
    # Parse chat history if provided
    messages = [{"role": "system", "content": "you are a helpful assistant. answer always in one sentence"}]
    
    if chat_history:
        try:
            history = json.loads(chat_history)
            for msg in history:
                if msg["role"] in ["user", "assistant"]:
                    messages.append(msg)
        except Exception as e:
            print(f"Error parsing chat history: {str(e)}")
    
    # Read the file content into memory
    file_content = await file.read()
    
    # Process the audio file
    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
                )
            
            decoded_message = user_message.decode("utf-8")
            print("user_message: " + decoded_message)
            
            # Add user message to messages
            messages.append({"role": "user", "content": decoded_message})
            
            # Send to LLM with full chat history
            chat_completion = client.chat.completions.create(
                messages=messages,
                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)
            waveform_float32 = np.frombuffer(encoded_waveform, dtype=np.float32)
            waveform_int16 = np.int16(waveform_float32 * 32767)
            
            # Save to buffer
            buffer = io.BytesIO()
            with wave.open(buffer, "wb") as wf:
                wf.setnchannels(1)
                wf.setsampwidth(2)
                wf.setframerate(48000)
                wf.writeframes(waveform_int16.tobytes())
            
            buffer.seek(0)
            
            return Response(
                content=buffer.getvalue(),
                media_type="audio/wav",
                headers={
                    "X-User-Message": decoded_message,
                    "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
            )