feat: speech to text

src/ai-engine/stt.client.py

@@ -0,0 +1,11 @@
+import asyncio
+import websockets
+
+async def main():
+    uri = "ws://localhost:8000/ws"
+    async with websockets.connect(uri) as ws:
+        while True:
+            text = await ws.recv()
+            print("Live:", text)
+
+asyncio.run(main())

src/ai-engine/stt_backend.py

@@ -0,0 +1,58 @@
+import asyncio
+import numpy as np
+import sounddevice as sd
+import whisper
+from collections import deque
+from fastapi import FastAPI, WebSocket
+
+app = FastAPI()
+model = whisper.load_model("base")
+
+SAMPLE_RATE = 16000
+queue = asyncio.Queue()
+buffer = deque(maxlen=SAMPLE_RATE * 2)
+loop = None
+
+def audio_callback(indata, frames, time, status):
+    loop.call_soon_threadsafe(queue.put_nowait, indata.copy())
+
+async def audio_collector():
+    print("Microphone started...")
+    while True:
+        chunk = await queue.get()
+        for sample in chunk:
+            buffer.append(sample)
+
+@app.on_event("startup")
+async def start_mic():
+    global loop
+    loop = asyncio.get_running_loop()
+
+    stream = sd.InputStream(
+        samplerate=SAMPLE_RATE,
+        channels=1,
+        callback=audio_callback
+    )
+    stream.start()
+
+    asyncio.create_task(audio_collector())
+
+@app.websocket("/ws")
+async def websocket_endpoint(ws: WebSocket):
+    await ws.accept()
+    print("Client connected")
+
+    while True:
+        await asyncio.sleep(0.5)
+
+        if len(buffer) < SAMPLE_RATE:
+            continue
+
+        audio = np.array(buffer).flatten()
+        result = await loop.run_in_executor(
+            None, lambda: model.transcribe(audio, fp16=False)
+        )
+
+        text = result["text"].strip()
+        if text:
+            await ws.send_text(text)