Hugging Face's logo

Spaces:
sanjaystarc
/
voice-agent
Sleeping

App Files Community
sanjaystarc commited on
Commit
a2b7dcd
·
verified ·
1 Parent(s): bccadca

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +27 -18
app.py CHANGED
@@ -2,31 +2,31 @@ import base64
2
  import uvicorn
3
  from fastapi import FastAPI, WebSocket
4
  from fastapi.responses import HTMLResponse
5
- from dotenv import load_dotenv
6
- import os
7
  from voice_agent import CrewVoiceAgent
8
 
9
- load_dotenv()
10
- agent = CrewVoiceAgent()
11
-
12
  app = FastAPI()
 
13
 
 
 
 
14
  @app.get("/")
15
  async def home():
16
  return HTMLResponse("""
17
  <html>
18
  <body>
19
- <h2>🎤 Gemini CrewAI Voice Agent</h2>
20
  <button onclick="start()">Start</button>
21
  <button onclick="stop()">Stop</button>
22
  <p>Status: <span id='status'>Idle</span></p>
23
-
24
  <script>
25
  let ws;
26
  let recorder;
27
-
28
  async function start() {
29
- ws = new WebSocket("wss://" + location.host + "/ws");
 
30
  ws.onmessage = (event) => {
31
  const audio = new Audio("data:audio/wav;base64," + event.data);
32
  audio.play();
@@ -36,12 +36,12 @@ async def home():
36
  recorder = new MediaRecorder(stream, { mimeType: "audio/webm" });
37
 
38
  recorder.ondataavailable = async (e) => {
39
- const b = await e.data.arrayBuffer();
40
- ws.send(b);
41
  };
42
-
43
- recorder.start(300);
44
- document.getElementById("status").textContent = "Listening…";
45
  }
46
 
47
  function stop() {
@@ -54,15 +54,24 @@ async def home():
54
  </html>
55
  """)
56
 
 
 
 
57
  @app.websocket("/ws")
58
  async def websocket_endpoint(ws: WebSocket):
59
  await ws.accept()
60
 
61
- async for audio_chunk in ws.iter_bytes():
62
- response_audio = await agent.handle_audio(audio_chunk)
63
- b64_audio = base64.b64encode(response_audio).decode()
 
 
 
 
 
 
 
64
 
65
- await ws.send_text(b64_audio)
66
 
67
  if __name__ == "__main__":
68
  uvicorn.run(app, host="0.0.0.0", port=7860)
 
2
  import uvicorn
3
  from fastapi import FastAPI, WebSocket
4
  from fastapi.responses import HTMLResponse
 
 
5
  from voice_agent import CrewVoiceAgent
6
 
 
 
 
7
  app = FastAPI()
8
+ agent = CrewVoiceAgent()
9
 
10
+ # --------------------------
11
+ # FRONTEND
12
+ # --------------------------
13
  @app.get("/")
14
  async def home():
15
  return HTMLResponse("""
16
  <html>
17
  <body>
18
+ <h1>🎤 Gemini CrewAI Voice Agent</h1>
19
  <button onclick="start()">Start</button>
20
  <button onclick="stop()">Stop</button>
21
  <p>Status: <span id='status'>Idle</span></p>
22
+
23
  <script>
24
  let ws;
25
  let recorder;
26
+
27
  async function start() {
28
+ ws = new WebSocket("ws://" + location.host + "/ws"); // HF Spaces requires ws://
29
+
30
  ws.onmessage = (event) => {
31
  const audio = new Audio("data:audio/wav;base64," + event.data);
32
  audio.play();
 
36
  recorder = new MediaRecorder(stream, { mimeType: "audio/webm" });
37
 
38
  recorder.ondataavailable = async (e) => {
39
+ const arrayBuf = await e.data.arrayBuffer();
40
+ ws.send(arrayBuf);
41
  };
42
+
43
+ recorder.start(300); // send audio chunks
44
+ document.getElementById("status").textContent = "Listening...";
45
  }
46
 
47
  function stop() {
 
54
  </html>
55
  """)
56
 
57
+ # --------------------------
58
+ # WEBSOCKET
59
+ # --------------------------
60
  @app.websocket("/ws")
61
  async def websocket_endpoint(ws: WebSocket):
62
  await ws.accept()
63
 
64
+ while True:
65
+ try:
66
+ audio_bytes = await ws.receive_bytes() # Receive raw binary audio
67
+ audio_reply = await agent.handle_audio(audio_bytes)
68
+
69
+ b64_audio = base64.b64encode(audio_reply).decode()
70
+ await ws.send_text(b64_audio)
71
+
72
+ except Exception:
73
+ break
74
 
 
75
 
76
  if __name__ == "__main__":
77
  uvicorn.run(app, host="0.0.0.0", port=7860)