fix: handle long text translation by chunking

app.py

@@ -57,8 +57,35 @@ async def translate_text(request: TranslationRequest):
     if not state["model_loaded"]:
         return JSONResponse(content={"message": "Model is not loaded yet"}, status_code=503)
     
-    result = state["translator"](request.text)
-    return {"translated_text": result[0]["translation_text"]}
+    # Split the text into chunks
+    text_chunks = split_text(request.text)
+    
+    # Translate each chunk
+    translated_chunks = []
+    for chunk in text_chunks:
+        # The translator returns a list of dictionaries
+        translated_chunk = state["translator"](chunk, max_length=512)
+        translated_chunks.append(translated_chunk[0]['translation_text'])
+    
+    # Join the translated chunks
+    translated_text = "".join(translated_chunks)
+    
+    return {"translated_text": translated_text}
+
+def split_text(text: str, max_length: int = 512):
+    # A simple way to split text by chunks of max_length
+    # A more sophisticated approach could split by sentences.
+    text_chunks = []
+    while len(text) > max_length:
+        # Find the last space to avoid splitting words
+        split_at = text.rfind(' ', 0, max_length)
+        if split_at == -1:
+            # No space found, split at max_length
+            split_at = max_length
+        text_chunks.append(text[:split_at])
+        text = text[split_at:].lstrip()
+    text_chunks.append(text)
+    return text_chunks
 
 @app.get("/health")
 async def health_check():