Create main.py

main.py

@@ -0,0 +1,66 @@
+# main.py
+
+from fastapi import FastAPI, UploadFile, File
+from fastapi.responses import JSONResponse
+from transformers import pipeline
+import traceback
+import re
+import uvicorn
+
+app = FastAPI(title="Tacab ASR Somali API")
+
+# Load ASR model
+asr = pipeline(
+    "automatic-speech-recognition",
+    model="tacab/ASR_SOMALI",
+    tokenizer="tacab/ASR_SOMALI",
+    chunk_length_s=30,
+    stride_length_s=6,
+    return_timestamps="word",
+    device=-1
+)
+
+# Auto punctuation
+def auto_punctuate(text):
+    text = text.strip()
+    # Capitalize sentences
+    def capitalize_sentences(text):
+        sentences = re.split(r'(?<=[.?!])\s+', text)
+        return '. '.join(s.strip().capitalize() for s in sentences if s)
+
+    if '.' not in text and len(text.split()) > 5:
+        text += '.'
+
+    words = text.split()
+    new_text = ""
+    for i in range(0, len(words), 10):
+        segment = " ".join(words[i:i+10])
+        new_text += segment.strip().capitalize() + ". "
+
+    return capitalize_sentences(new_text.strip())
+
+@app.post("/transcribe")
+async def transcribe(file: UploadFile = File(...)):
+    try:
+        # Save the uploaded file temporarily
+        temp_path = f"/tmp/{file.filename}"
+        with open(temp_path, "wb") as f:
+            f.write(await file.read())
+
+        # Transcribe
+        result = asr(temp_path)
+        raw_text = result.get("text", "").strip()
+        if not raw_text:
+            return JSONResponse({"error": "No transcription result."}, status_code=400)
+
+        # Punctuate
+        cleaned_text = auto_punctuate(raw_text)
+
+        return {"transcription": cleaned_text}
+
+    except Exception as e:
+        traceback.print_exc()
+        return JSONResponse({"error": str(e)}, status_code=500)
+
+if __name__ == "__main__":
+    uvicorn.run(app, host="0.0.0.0", port=7860)