Update app.py

app.py

@@ -1,49 +1,70 @@
-from fastapi import FastAPI, UploadFile, File, WebSocket
-from fastapi.responses import JSONResponse
+from fastapi import FastAPI, WebSocket, UploadFile, File
 from pywhispercpp.model import Model
+import uvicorn
 import tempfile
 import os
+from time import time
 
-app = FastAPI(title="Whisper STT Server")
+app = FastAPI(title="pyWhisperCPP API")
 
-# Load tiny model once at startup
-#model = Model("tiny.en-q5_1")
 model = Model('base.en')
+
+
 @app.get("/")
-async def home():
-    return {"status": "Whisper STT is running"}
+def root():
+    return {"status": "Whisper.cpp API is running!"}
+
 
-# --------- File Upload Transcription ---------
 @app.post("/transcribe")
-async def transcribe_audio(file: UploadFile = File(...)):
+async def transcribe(file: UploadFile = File(...)):
     with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as temp:
         temp.write(await file.read())
         temp.flush()
         audio_path = temp.name
 
+    start = time()
     segments = model.transcribe(audio_path)
+    text = " ".join([seg.text for seg in segments])
+    elapsed = round(time() - start, 3)
     os.remove(audio_path)
 
-    text = " ".join(s.text for s in segments)
-    return {"text": text}
+    return {"text": text, "processing_time_seconds": elapsed}
 
-# --------- LIVE Streaming WebSocket ---------
-@app.websocket("/ws")
-async def websocket_stt(ws: WebSocket):
-    await ws.accept()
+
+# ==========================================================
+# 🔥 Real-time Speech Recognition (WebSocket)
+# ==========================================================
+@app.websocket("/ws/transcribe_stream")
+async def websocket_transcription(websocket: WebSocket):
+    await websocket.accept()
+    buffer = b""
 
     while True:
-        try:
-            data = await ws.receive_bytes()
+        chunk = await websocket.receive_bytes()
+        if chunk == b"__END__":
+            break
 
+        buffer += chunk
+
+        # When buffer > 1 sec of audio, transcribe
+        if len(buffer) >= 16000 * 2:  # 16kHz * 2 bytes = 1 second PCM16
+            # Save buffer temporarily
             with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as temp:
-                temp.write(data)
+                temp.write(buffer)
                 temp.flush()
-                segments = model.transcribe(temp.name)
+                audio_path = temp.name
 
-            text = " ".join(s.text for s in segments)
+            segments = model.transcribe(audio_path)
+            text = " ".join([seg.text for seg in segments])
 
-            await ws.send_json({"text": text})
-        except Exception:
-            await ws.close()
-            break
+            await websocket.send_text(text)
+            buffer = b""  # clear for next batch
+            
+            os.remove(audio_path)
+
+    await websocket.send_text("stream_end")
+    await websocket.close()
+
+
+if __name__ == "__main__":
+    uvicorn.run(app, host="0.0.0.0", port=7860)