update

Dockerfile

@@ -1,8 +1,17 @@
-# Use the official LibreTranslate image
-FROM libretranslate/libretranslate:latest
+# Use an official Python runtime as a parent image
+FROM python:3.9-slim
 
-# Expose the default LibreTranslate port
-EXPOSE 5000
+# Set the working directory in the container
+WORKDIR /code
 
-# Set the command to run the LibreTranslate server
-CMD ["./run.sh", "--host", "0.0.0.0"]
+# Copy the dependencies file to the working directory
+COPY requirements.txt .
+
+# Install any needed packages specified in requirements.txt
+RUN pip install --no-cache-dir --upgrade -r requirements.txt
+
+# Copy the rest of the application code to the working directory
+COPY . .
+
+# Command to run the API
+CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "7860"]

app.py

@@ -0,0 +1,46 @@
+from fastapi import FastAPI, HTTPException
+from pydantic import BaseModel
+import ctranslate2
+import sentencepiece as spm
+import os
+
+app = FastAPI()
+
+# Load the CTranslate2 model and the SentencePiece tokenizer
+try:
+    model_path = "en_ar_ct2_model/"
+    sp_model_path = os.path.join(model_path, "source.spm")
+
+    translator = ctranslate2.Translator(model_path, device="cpu") # Use "cuda" if on a GPU Space
+    sp = spm.SentencePieceProcessor()
+    sp.load(sp_model_path)
+except Exception as e:
+    # This helps in debugging if the model files are not found
+    raise RuntimeError(f"Error loading model: {e}")
+
+
+class TranslationRequest(BaseModel):
+    text: list[str] # Expect a list of strings for batching
+
+class TranslationResponse(BaseModel):
+    translations: list[str]
+
+@app.get("/")
+def read_root():
+    return {"message": "English to Arabic Translation API is running."}
+
+@app.post("/translate", response_model=TranslationResponse)
+def translate_text(request: TranslationRequest):
+    if not request.text:
+        raise HTTPException(status_code=400, detail="Input text list cannot be empty.")
+
+    source_sentences = request.text
+    source_tokenized = [sp.encode(sentence, out_type=str) for sentence in source_sentences]
+
+    # Translate the batch of sentences
+    translations_tokenized = translator.translate_batch(source_tokenized)
+
+    # Decode the translated sentences
+    translations = [sp.decode(translation.hypotheses[0]) for translation in translations_tokenized]
+
+    return {"translations": translations}

requirements.txt

@@ -0,0 +1,5 @@
+fastapi
+uvicorn
+ctranslate2
+sentencepiece
+transformers